Changelog

What changed, newest first. Anything that affects how the assistant behaves on a call is called out.

9 October 2026 — The mobile app#

In the app#

• AnswerStack on your phone. Open app.getanswerstack.com/m and add it to your home screen: My next tour, your tours with their briefs, the Visit section with photos for the family, tasks, calls, your schedule and booking pages. See The mobile app for how to install it.
• Unlock with Face ID or a fingerprint through a passkey, and it locks like a banking app. Add your passkey on your phone under Personal settings → Login security. Passkeys added from now on can unlock the mobile app; add an older one again if it doesn't.

8 October 2026 — A smarter Stack#

In the app#

• Stack answers far more. An AI model now decides what to look up in your account (calls, bookings, callbacks, playbooks and what they say, knowledge, usage, setup and, with Insights Ask, conversation results), reads it and answers. Code still checks every number before you see it.
• Chat history. Your chats, named by the question each started with, and your latest ones when you open Stack. The ask bar now fills the top bar.
• Was this helpful? Rate an answer, and say what was wrong; it helps us make Stack better.
• Daily limits keep AI answers within an allowance per question, person and account; past it, Stack answers in its simpler mode and says so.

For developers#

• New MCP tool ask_stack: one plain-English question to Stack, answered within your token's scopes.
• New routes POST /v1/account/insights/ask/messages/{messageId}/feedback and …/opened.

8 October 2026 — Stack, the assistant in the top bar#

In the app#

• Stack. Ask in plain English from the top bar of every page: Ask your Stack anything…, or ⌘K / Ctrl K. Stack answers how your account is set up on every plan (which number goes where, what is live on each playbook, which communities have no backup plan, whether the CRM and calendar are connected, who is on the team), with a table and a link to the page that changes it. With Insights Ask it also answers about your conversations, as Ask did. See Stack, your assistant.
• The Ask button in Insights is now Ask Stack and opens the same panel with the view's filters.

For developers#

• The Ask routes (/v1/account/insights/ask*) now take any member. A question about conversations from an account without insights.ask (or a role without Insights) is answered with kind: unavailable instead of a 402. Answers can carry a new page_link block.

7 October 2026 — Tour notes, next steps and photos#

In the app#

• Next steps as CRM tasks. A step on your list can also create a task in your CRM, for its owner, on the person's record. Done here completes it there; done there closes it here.
• Finish tour. Send a tour's notes and next steps to your CRM as one entry, with how many photos you shared; it happens on its own a day after the tour if you do not. Changes after that go as an update. The calls scorecard gains a Tour notes, next steps and photos card: how many tours were written up, how next steps ended, and whether families looked at their photos.
• Photos for the family. Take photos during a tour from the Notes tab, after confirming the family agreed. They upload in the background on a weak connection, and location and phone details are removed from every photo. Add captions, set the order, and choose which ones to share. The family sees them on their follow-up page after you press Show on their page, and anyone they forward it to does too. A tour with no page gets one, handed over as a QR code or a link you copy. On the page, photos open full screen and download. The follow-up page's sections gain Photos from your tour.
• Next steps after a tour. A tap on Send pricing or Schedule a second tour under a tour's notes makes a task with an owner and a due date. Admins set the list in the playbook (Next steps after a visit); communities add their own. Steps show on My tasks and Today, with the tour they came from, and managers see their communities' open ones on Today.
• Tour notes. Open a tour in My tours and choose Notes to write what happened, for your team only. Notes save as you type, show who wrote them, and stay with a tour when it is moved. Staff see them on the call's page under Tour notes. On a phone, an open tour now fills the screen.

For developers#

• New routes for a visit's notes: GET /v1/me/meetings/{bookingId}/visit, and POST, PATCH and DELETE under /v1/me/meetings/{bookingId}/visit/notes. A create takes a clientKey, so a retried request returns the same note; an edit takes the version it read and is refused with 409 when it is stale.
• Next steps: the playbook's new visitActions block; POST /v1/me/meetings/{bookingId}/visit/actions; GET, PATCH, DELETE and POST .../take under /v1/me/visit-actions; and GET /v1/account/visit-actions.
• Photos: POST .../visit/photo-consent; POST .../visit/photos (start, returning a one-time upload link), POST .../visit/photos/{photoId}/complete, PATCH and DELETE under /v1/me/meetings/{bookingId}/visit/photos/{photoId}; PUT .../visit/photos/order.
• The family's page: POST /v1/me/meetings/{bookingId}/visit/publish (returns a fresh link and QR code for a page made for the tour) and .../unpublish. The page payload gains a visit_photos module, and page events gain visit_photos_viewed and visit_photo_downloaded (with position). A follow-up page now has an origin (call or visit), and callId may be null.
• Finish: POST /v1/me/meetings/{bookingId}/visit/finish; the visit gains a crm state. Metrics: GET /v1/account/analytics/visits.
• CRM tasks from next steps: the visitActions block gains typedCrmTask, steps gain crmTask and closedInCrm, and connectors may declare completeTask. MCP gains get_visit and list_visit_actions, on GET /v1/account/visits/{bookingId} and GET /v1/account/visit-actions.

6 October 2026 — Hand callers to a ring group#

On calls#

• A hand-off can ring a whole ring group. When a location's human backup names a ring group, the AI agent rings its people (in its own order, with press 1) when a caller asks for a person, instead of the backup numbers. The first to accept is put through; if nobody does, the agent takes a callback. See When the AI agent hands a caller over.

In the app#

• Human backup → Ring a ring group instead.
• A ring group that takes turns says who the next call starts with.
• A call's page shows Rang the backup ring group for a hand-off that rang one.
• Test the ringing: the test number's call flow has Ring your team on test calls, off by default. When on, test calls to it ring your team as its call flow says.

6 October 2026 — A community's profile in sections, photos only from Media, test emails to anyone, and Microsoft 365 Calendar#

In the app#

• A community's profile is in sections: Details, Location, Hours and closures, Staff and Highlights and values, under a Profile heading in the list beside the page. Each has its own Save to unpublished changes button and saves only what is in it, and you are asked before leaving a section with changes you have not saved. The other sections sit under What callers hear and After the call. See Communities.
• A section with a problem or an unfilled value the playbook uses has a red or amber dot, and each value on What the account playbook uses links to the section that fills it in.
• Photos and the virtual tour come only from Media. The profile's own photos and tour link are gone; the follow-up page's Photos section shows what is linked to the community in general on its Media tab.
• Old links to a community's profile still work: they open Details.
• Send a test… asks who to send it to. In every email editor, the test button opens a dialog with your own address filled in. Change it, or add colleagues, up to five addresses. Each gets its own copy marked [Test], and if one could not be reached the app names it. See Emails.
• New connector: Microsoft 365 Calendar. Book meetings on an Outlook calendar, or on a shared mailbox such as tours@, the same way Google Calendar works: the assistant offers times when the staff member is free and invites them to the meeting. Sign in with a work or school Microsoft 365 account. Your platform administrator turns it on for your account. See Microsoft 365 Calendar.
• Reconnecting a calendar keeps its account when meetings go on the signed-in account's own calendar (Google Calendar's main calendar, or Microsoft 365 without a shared mailbox): signing in as someone else is refused, so meetings already booked can still be cancelled. Calendars shared by ID or a shared mailbox can be reconnected by anyone with access.

For developers#

• The group profile (GET and PUT /v1/account/groups/{groupId}/profile) no longer has media. The {{group.tourUrl}} detail is retired.
• The Microsoft 365 Calendar connector is microsoft_calendar, in the scheduler role. Its options are bookingMailbox, calendarId and inviteCaller.
• The sign-in completion's reason has two new values: personal_account and mailbox_access_denied, and its result a new one, other_account, for a refused reconnect.

5 October 2026 — Ring your team first, then the AI agent#

On calls#

• A number can ring your team before the AI agent answers. Each part of a number's schedule says who answers: the AI agent straight away, or a ring group for up to 60 seconds first. Callers hear a short notice (that the call is recorded, and transcribed when live notes are on) and normal ringing. A team member who picks up hears who the call is for and presses 1 to take it, so a voicemail never takes a caller. When nobody does, the AI agent answers, still saying it is an AI first.
• A caller who hangs up while it rings is a missed call: the location's notify list is told and your CRM gets a task to call them back. A location can turn this off.
• Calls your team answers are recorded like every other call, and the notetaker takes notes on them when the notice says the call is transcribed.

In the app#

• New: the visual call flow on every number (Phone numbers → Call flow), with What happens at to check any moment, and Ring groups on each location. See Call flows and ring groups.

• Taking calls or Away on My calls: Away skips you in every ring group.

• A call that rang your team shows Rang the team first on its page: each phone in order, how long it rang and what happened. The Calls list says Rep answered or Missed call.

• The playbook's Persona tab has the notice said before the phones ring and what the agent says when it answers after ringing. A location can write its own notice on its Profile.

• Direct lines: on Team, give a team member a number that rings them first, then the AI agent.

• The Scorecard's Calls part shows Your team answering first: per ring group and per person, calls offered, taken, missed and talk minutes. The AI agent's own rates now leave out calls your team answered.

• My calls marks calls a team member answered when the team was rung.

Billing#

• Minutes your team talks on calls it answers are a new meter, rep minutes, priced apart from the AI agent's. Plans include an allowance. On a call the agent takes after ringing, its minutes start when it answers.

For developers#

• GET /v1/account/calls/{callId} has ring: the ring group and revision that rang, how it ended, who answered and each leg (attempts). Calls have answeredBy (agent, rep or nobody).
• New ring group endpoints under /v1/account/groups/{groupId}/ring-groups and /v1/account/ring-groups/{id}, with /history; GET /v1/account/numbers/{id}/routing/preview.
• A new usage meter, rep_minutes, and a new notification, missed_call.
• GET /v1/account/analytics has ringing (byRingGroup, byRep, byAnswered); connected and transferred there now count the agent's own conversations only.
• New GET /v1/account/direct-lines, and PUT / DELETE /v1/account/numbers/{numberId}/direct-line.
• GET /v1/me/handoffs items have fromRingStep.

4 October 2026 — A menu in sections, a Settings list, and Personal settings#

In the app#

• The main menu is in sections. Overview is at the top, then Your work (Today, My tours, My calls, My tasks, My booking pages), Conversations (Calls, Insights), Business (Communities, Pricing, Booking pages) and AI agent (Playbooks, Knowledge, Phone numbers, Tools, Connections), with Settings at the bottom. My pages is now My booking pages. Who sees what has not changed.
• Settings is a list of pages: Account (General, Team, Billing, Usage, Login security), Language (Your terms, Caller vocabulary), Calls (Recording and notes, Insights) and Brand and channels (Brands, Emails, Website widget). Team, Billing and Usage have moved there from the main menu. See Settings.
• New: Personal settings, from the account menu in the top right, for everyone. Set your name and the time zone you see times in, change your email or password, and find your own Login security, Calendar, Availability and Developer pages. My calendar and My availability have moved here from the main menu. See Personal settings.
• Old links still work: they open the page in its new place.

3 October 2026 — Models from a list, and every call says which it used#

In the app#

• Advanced → Models on a playbook is now a list of the models we support, with Recommended first. Our team can also set your account's models for you to try a new one.
• A call's page shows the Language model, Speech recognition model and Voice model it ran on, and says when the backup language model stepped in.

For developers#

• New GET /v1/account/ai-models lists the models a playbook can choose and the recommended one of each.
• GET /v1/account/calls/{callId} has models: llm, stt, tts (null on calls from before this change, and stt and tts on chats) and llmFallbackUsed.

3 October 2026 — See what the model was sent#

In the app#

• Show prompts in a call's debugger shows exactly what the language model read on each step: its instructions, the turn's notes, the conversation, tool calls and results, and the tools it could use. Test calls and test chats always keep them; real calls keep them when our team turns prompt capture on for your account. Each look is written to the audit log.

For developers#

• New GET /v1/account/calls/{callId}/prompts returns the call's prompts, unmasked. Like the other reveal routes it is for signed-in people only (API tokens are refused), is rate limited, and is audited as call.reveal_sensitive with source: prompts.

1 October 2026 — One name for each thing#

In the app#

• One word per thing. The app now uses the names in the new Glossary:
  • What the rules put on Today and My tasks are tasks; what was promised to a caller is a commitment (no longer "items" or "promises").
  • A staff member's booking-rule field is their Job title; Role means only the sign-in role, and the group manager role shows your own word ("Community manager").
  • Settings → Your words is now Your terms, and Words your callers use is Caller vocabulary.
  • Invoice lines for a reduction we give you read Discount (they said "Promotion"), so they are never confused with your specials.
  • Show rate, held rate and "came to the visit" are now Attended and Attended rate.
  • Insights, Today, Ask's examples and the token and connected-app screens use your word for a location everywhere.
• Locked features say what turns them on. A locked screen now says the feature is an add-on that is not turned on for your account, instead of "not part of your plan". Talk to us still asks our team to turn it on.

For developers#

• No API paths, fields or values changed. The 402 feature_not_included message now reads "This account does not have … turned on.", then what turns it on (an add-on or a grant our team adds; Insights Ask names Conversation Intelligence, which it comes with). New invoice lines' descriptions start "Discount:".
• GET /v1/account/groups/{groupId}/draft no longer fails on a draft row that does not validate: the invalid parts are dropped and the response lists them in discarded. A read by someone who may save the draft also repairs the row, once; a viewer's read changes nothing.

1 October 2026 — Community changes are tested and published#

In the app#

• Community changes wait until you publish them. A community's profile, human backup plan and customizations now save to its unpublished changes, like a playbook draft. Callers keep hearing the live settings until someone presses Publish on the Unpublished changes bar at the top of the community, which says what changed and what publishing would hold back. Discard throws the changes away. See Test and publish a community's changes.
• Holidays and closures still apply at once. They save on their own, with Save holidays and closures, and do not wait for a publish.
• Test a community before callers hear it. The community page has Test call and Test chat, with the community's live settings or its unpublished changes, and the playbook editor's test panels can use a community's unpublished changes. Each panel says what it is testing.
• Group managers can run test calls and chats for their own communities. Tests are still never billed.

For developers#

• PUT /v1/account/groups/{groupId}/profile, …/human-backup and …/customizations, the customization reset and the bulk apply now save to the community's unpublished changes. They take baseHash (the draft hash you loaded, or null), and answer 409 when someone else saved first. The profile save still accepts expectedVersion (the live version you loaded) for now, so existing API-token integrations keep working; it is deprecated in favour of baseHash. Publish with POST /v1/account/groups/{groupId}/draft/publish, read the changes with GET …/draft, discard with DELETE …/draft, and save closures live with PUT …/closures. The MCP tool update_group_profile saves to the unpublished changes too.
• POST /v1/account/test-call and POST /v1/account/test-chat/messages take groupDraft.

1 October 2026 — Emails wear your brand#

In the app#

• One logo and one set of colours. Staff alerts, caller emails and meeting emails now wear the same brand as your booking and follow-up pages: the community's brand, or your default. The Colors and logo tab under Settings → Emails and the colors and logo on a community's Emails tab are gone. Change the look under Settings → Brands, or on a community's Brand tab.
• The brand editor has a new Emails section, where you set whether the logo sits on the left or in the middle of an email. It also shows a preview of an email. Emails show the logo only when it is a PNG or JPEG.
• Your email look was moved for you. Where your brand had no logo, it took your email logo, unless your pages already showed a logo from a playbook. Where both had one, your brand's was kept. An account or community with an email look but no brand got a brand made from it. It is in use only where your pages wore our standard look; otherwise it waits for you to pick it, so your pages did not change. Settings → Brands shows once what changed, with a link to download any email logo that is no longer used. See Colors and logo: your brand.
• Group managers who can change their community's brand now change its emails' look too.

1 October 2026 — Offers live in specials, and price points retire#

How the assistant behaves on calls#

• The assistant states an offer only when it is one of your running specials. An offer or discount found anywhere else is stopped before it is spoken, and the assistant says your fallback line instead. That covers your brief, FAQ, knowledge, a community's highlights or custom values, and a tool's answer. Without a running special it does not mention specials. See Specials are the only place for offers.
• Price points are retired. They can no longer be edited. A community that has not published its drafted prices yet keeps its old price points on calls until you publish them from Pricing → Prices; after that, only the price book is used. See Price points are retired.
• A community highlight that states a price or an offer is no longer said.

In the app#

• The profile's Talking points are now Local highlights, so they are not confused with the playbook's talking points under Before moving on.
• Talking points and local highlights cannot hold a price or an offer. Publishing such a talking point, or saving such a new highlight, is refused with a note saying where it belongs. Ones saved before are flagged, never deleted, and Move to a draft special turns one into a draft special for you to date and approve.
• The profile's Price points section is gone. The publish checklist names any community that still says its old price points because it has no published prices.

1 October 2026 — One booking rate, and Analytics in Insights#

In the app#

• One booking rate everywhere. Booking rate is tours booked ÷ connected conversations, one channel at a time, on the Overview, in Insights, in Ask and in the API. Each number says what it is a share of. The Overview's "% of today's calls" divided by every call, answered or not; it is now the booking rate of today's connected calls, and it no longer stops counting at 200 calls. See Call metrics.
• Analytics is now part of Insights. Every tile, chart and table is in the Calls and Booking pages parts of the Insights Scorecard, with the Insights filters, and still comes with every plan. The Analytics menu item is gone, and old links open the Calls part.
• Insights shows phone calls unless you choose another channel. Choose Every channel to see each channel's booking rate side by side.
• Ask answers "what was our booking rate?" with the booking rate. Rankings of communities still use meeting target attainment, and now say so.
• With the Conversation Intelligence add-on, the questions the assistant could not answer are in Knowledge gaps, linked from the Calls part.
• Calls → Happening now lists phone calls, with open chats under their own heading.
• Usage calls the booked meter Conversations that booked a tour, since chats count too. It also explains why its minutes and answered calls can differ from Insights and the invoice.
• Booking pages say Still booked for tours not cancelled.

1 October 2026 — One set of messages for every tour#

What callers notice#

• Every tour is confirmed. A tour booked on a call, in the website chat, on a booking page or a follow-up page gets a confirmation, reminders and word of any change, whether or not the follow-up page is on: by text to people who agreed to texts, and by email to the address they gave. A calendar invitation from your calendar connection counts as the confirmation email.
• The assistant only promises what will go. It knows at the start of the call whether the number replied STOP and whether the CRM says not to text, so it never says a text is coming when it is not.

In the app#

• Meeting emails replace the booking page emails in Settings → Emails, and are used for every tour. The follow-up page's reminder, moved and cancelled emails are now these. See Emails.
• A tour booked from a follow-up page's Book a time shows on that page.

30 September 2026 — One calendar per rep, and closures#

What callers notice#

• Tours land on the rep's own calendar. A rep who connected a calendar in My calendar is booked on it by calls and the website chat too, whatever calendar the playbook uses, and no channel offers a time another channel already booked.
• Closed means closed. A community's holidays and closures, set on its Profile, stop tours being offered everywhere and send callers who need a person to a callback, due when it reopens. See Communities.

In the app#

• Phone numbers can add a part for while the community is closed, and the weekly slot's example name is now Daytime (slots named Business hours were renamed). See Phone numbers.
• Booking rules call the staffed-hours condition Inside staffed hours. A rule that books through a page only offers the people on it who match the rule's person or role.
• Which hours apply is a new guide to tour windows, reps' hours, staffed hours and schedules. See Which hours apply.

30 September 2026 — Opt-outs, recording and spam#

What callers notice#

• "Stop texting me" opts out. A STOP, CANCEL, END, QUIT, UNSUBSCRIBE or OPT OUT reply with only "texting", "texts", "messages", "me" or "please" beside it now opts the number out, the same as a bare STOP. "Cancel my tour" is still a request for your team. See Texting and consent.
• Recording starts with the opening. A recorded call starts recording when the opening starts playing, never before the recording notice. The audio file may miss the opening's first words. See Call recording.

In the app#

• Booking page emails are yours to edit: the confirmation, reminders, a new time and a cancellation, on their own tab of Settings → Emails and in each community's Emails tab, starting from the words they have always had. See Emails.
• Phone numbers show Split test running on a number an experiment is splitting, with the versions, and warn before a change that would stop it.
• Insights marks the views that need the Conversation Intelligence add-on with an Add-on badge.
• Publishing a playbook that uses none of your account documents, while you have some, now shows a warning in the checklist. See Knowledge.
• Too many wrong sign-in codes now sends you an email, and our support team can unlock your sign-in. See Login security.
• A new guide: Website chat.

Billing#

• Calls answered no longer counts a caller who hung up before the opening finished playing, even if they said "hello?" over it, and it leaves out calls from numbers we have flagged as spam or robocalls (their minutes still count). See Usage and invoices.

30 September 2026 — Signed incoming events and safer sign-ins#

In the app#

• Incoming events secret on each connection: your system signs the events it sends AnswerStack (such as a cancelled tour) with its own secret and a timestamp, so a captured request cannot be replayed. The old signature keeps working until your first request signed the new way. See Incoming events.
• Only the person who started a sign-in can finish it, when connecting Google Calendar, Microsoft 365, your own calendar or a tool server. Finish in the browser where you are signed in to AnswerStack.

28 September 2026 — Emails you can edit#

In the app#

• Settings → Emails. Owners and Admins edit the staff alerts sent when a caller needs someone, in an email editor with the same detail pills as the playbook, a live preview and Send me a test. See Emails.

• Colors and logo for your emails, with Start from the follow-up page and a contrast check as you choose.

• Caller emails in the playbook. The Follow-up page step has What the emails say: the email sent for each message when the caller chose email.

• Each community's Emails tab for its own alerts, caller emails and colors.

• Standard emails, read-only: the sign-in, invitation and account emails your people receive.

• Every follow-up message by email for a caller who chose email: reminders, a moved or cancelled tour and the note the morning after, not only the first link.

• One-click unsubscribe in every email to a caller. See Texting and consent.

Behind the scenes#

• Email is now sent through Resend, with bounces and spam complaints recorded so a bad address is not tried again.

28 September 2026 — Brands#

In the app#

• Brands: your logo, cover photo, colours and typeface, set once for booking pages, follow-up pages and the website widget. Make brands for the whole account, and let each community use one or make its own; owners can turn community brands off. See Brands.
• Booking pages can pick a brand and a cover photo of their own.
• Follow-up pages of a community with a brand use the brand's look in place of the playbook's page style.

28 September 2026 — Booking pages everywhere#

In the app#

• Share on every booking page: its link, the page inside a box on your website, or a popup button. Links can carry UTM values and fill the form in (name, email, phone). See Booking pages.
• Texts about a booking: people who tick the box on the form get the confirmation, reminders and changes by text too. STOP stops them; START turns them back on.
• START only counts for your number. A START or YES sent to another business's number no longer turns your texts back on for someone who opted out of yours.
• Analytics → Booking pages: page views (no cookies), times picked, bookings, completion, cancellations, moves and no-shows per page and rep, and every booking by source.
• Book through a page on a booking rule. See Booking rules.

On calls and chat#

• A booking rule that names a community page offers that page's times and length, and the page chooses who takes the tour, as it does online. If the page cannot book, the rule books as before.
• On a follow-up page, Book a time opens the page the rule names, with the caller's first name already filled in.
• The website chat asks for the visitor's name when an email or phone they type matches a record, instead of asking "am I speaking with…?" with the name on it. Nothing from the record is used until the name they give matches.

27 September 2026 — Community booking pages#

In the app#

• Community booking pages: a community's team can share one link. Tours go round robin (with priority and weight, equal shares or whoever has the fewest that day), or people choose who to meet. Each tour lands on the rep's own calendar, inside their own hours. See Community booking pages.
• Booking pages in the main menu: every page in the account with anything that stops it taking bookings.
• Reps specialize by service on a community's staff, with what each person Handles. A page can ask the service first, and a booking rule can book Whoever handles a service.

On calls and chat#

• Returning leads get their own rep: once a caller confirms who they are, the assistant offers their CRM sales counselor's times first, when that person is on the community's staff. The website chat looks a visitor up by the email or phone number they give, and does the same.
• A caller who chose a service, at a community whose reps specialize, is offered only the people who handle it. If nobody does, the assistant offers a callback.

27 September 2026 — Reps' tasks and visit briefs#

In the app#

• My tasks (with Conversation Intelligence): a sales rep sees the tasks for their own leads, with who to call and why, and marks them done, snoozes, dismisses or hands them back. See My tasks. Tasks can now go to any active login, a sales rep included; high-priority emails to a rep link to My tasks.
• Visit briefs in My tours: an open tour has a Visit brief tab, and the day-before email opens the tour.
• Your next seven days on My availability: when you can actually be booked, with busy time taken out.
• Staff list warnings: an entry whose email matches no login, or a deactivated login with tours still to come, says so. Hours beside a rep opens their hours (read-only for group managers).
• Keep tours in the CRM on Team's Hours for a rep with no calendar yet.
• A rep's note when they cancel a tour now reaches the person, on the cancellation text or email.

27 September 2026 — Booking pages#

In the app#

• Booking pages: each rep can make pages people use to book a tour with them, like Calendly. People pick a time from the rep's open times, give their details and answer the page's questions, and the tour goes straight onto the rep's calendar. They get a confirmation, reminders the day before and two hours before, and a link to reschedule or cancel. See Booking pages.
• My pages: where each rep makes, edits, pauses and removes their pages and changes their link.
• Pages on Team: owners and admins pause or resume anyone's pages; the rep is emailed.
• My tours shows tours booked on a page, with what the person answered.

Booking pages are part of some plans. Bookings made on a page are included and not charged like calls.

27 September 2026 — Reps' hours, the CRM as a calendar, and My tours#

On the call#

• Each rep is offered only inside their own hours. The assistant offers a rep only when the time is inside the community's tour windows, inside the rep's own hours, and free on their calendar. Every rep's hours start as a copy of their community's tour windows, so nothing changes until someone edits them. See My availability.
• A rep's calendar can be your CRM. A rep who keeps their tours in the CRM is offered when the CRM shows them free, at any community, and their tours are booked there. See My calendar.

In the app#

• My tours: everyone's own upcoming, past and cancelled tours, with who is coming and what they told the assistant, and Cancel and Reschedule. A Sales rep lands here. See My tours.
• My availability: weekly hours, date overrides and named schedules.
• Hours on Team: owners and admins change a rep's hours; the rep is emailed.

27 September 2026 — Schedules for phone numbers#

On the call#

• A number can run different playbooks at different times. Business hours, evenings, weekends and holidays can each have their own playbook, so an after-hours caller can hear a different greeting and conversation, not only a callback offer. See Run other playbooks at set times.

In the app#

• Schedule, a new calendar icon on each row of Phone numbers and on a community's numbers: holidays and closures, weekly time slots and All other times, each with its own playbook and version, and a line saying which part answers right now.
• Experiments on part of a schedule. An experiment can split only the calls one part of a number's schedule answers, for example Business hours.

26 September 2026 — Price books#

On the call#

• Prices come from each community's published price book. The assistant quotes a community's starting prices, always as "starting at", in the wording under How prices are said on the playbook's Guardrails tab. It never gives a top price, a range, a total or a discount it worked out, and never another community's price. See Prices and care types.
• Nothing changes until you publish. Your communities' price points were copied into draft prices for you to check. Until a community publishes its prices, the assistant keeps using its price points exactly as before.
• It says plainly what a community does not offer. Asked about memory care at a community without it, the assistant says so. It can also name nearby communities that offer it, and offer to connect the caller, if you turn that on.
• It narrows a price without asking twice. For a long price list it looks up the price of the exact apartment or care level the caller asks about, and uses answers they have already given.
• Specials are said in your approved words, only while they run, with their end date. See Specials and price approval.
• Questions can take their choices from your prices. A Care type question can offer only what the community offers, and booking rules can act on the answer.

In the app#

• Pricing, a new item in the menu: every community against every care type, with filters, Compare with an earlier date, Change many prices, Export CSV and Import CSV, and one Review and publish for many communities at once.
• Care types and options: what you sell, once for the account, with up to two option sets (such as apartment and care level) per care type. Communities can add care types only they offer.
• A Pricing tab on every community: what it offers, its starting prices, extra words or a hidden price, scheduled changes, versions and Roll back, and Still correct for prices due for review.
• Price approval, off by default: price changes and specials by community managers, or by everyone, wait for an owner or admin.
• Nearby communities on the community profile, and Pricing insights: what callers asked about cost, and what they asked for that a community does not offer.
• A Pricing section for the follow-up page, off until you add it, showing the community's current prices with what the caller asked about first. See Follow-up page.
• CRM price sync, optional: take starting prices from WelcomeHome every night through a mapping you control, with review or automatic publishing within limits. See CRM price sync.
• Two new words under Settings → Your words: what you sell ("care type") and a dated offer ("special").

26 September 2026 — Talking points before the offer#

• Before moving on, a new section under each line of questioning, holds what the assistant says once the required answers are in and before it offers the tour. This changes what callers hear on any line you add talking points to. See Questions and flows.
• Say this points are said word for word. Talk about points are a topic the assistant puts in its own words, and can work in the caller's answers and values from a tool that runs once all required answers are in.
• A point whose answer or lookup is not in is left out, never said with a gap.
• Communities can change, turn off or add talking points. A save that would leave one using a tool or question the community no longer has is refused.
• The call debugger shows each talking point said, left out, or reworded.

25 September 2026 — Insights: Today, commitments and visit briefs#

• Today: a daily list of what needs a person, from twelve fixed rules: a Hot lead with no next step, a promise coming due, an urgent timeline, a visit tomorrow whose page was never opened, a cancellation not rebooked, a no-show, a conversation that ended badly, and more. Snooze, complete, dismiss with a note, or assign each item. Nothing on it contacts a lead. See Today.
• Commitments: callbacks, answers to confirm and information promised to a lead are recorded with an owner and a due time, closed when the callback is done, the page is sent or the CRM task is completed, and shown on the conversation and lead pages. See Commitments.
• Visit briefs before every booked visit, refreshed the day before and two hours before, with talking points quoted from your playbook and never a care, health or money topic. See Visit briefs.
• Your CRM gets a task for each promise, for the lead's rep, and the brief as an activity, where your CRM supports them (WelcomeHome does). High-priority items and briefs are emailed as links.
• Settings → Insights gains Today's rule switches, per account and per community.
• Part of Conversation Intelligence, which is being rolled out to selected accounts.

25 September 2026 — Insights: Ask#

• Ask questions about your conversations in plain English from any Insights view, and get numbers, tables, charts and a link to the view that shows the same thing. Every number is checked against the data before you see it, a rate is called better only when the evidence supports it, and every answer states its dates, time zone and scope. See Ask.
• Follow-up questions build on the last answer ("Now just chat"), and each answer's context shows as chips you can remove to answer again.
• Saved questions and a weekly digest of them by email.
• Settings → Insights gains the reporting time zone, the day weeks start and a daily limit on questions.
• The Leaderboard can rank one objective only, such as booked meetings.
• MCP: a new insights:read scope, with the tools get_scorecard, compare_insights and query_insights. See the MCP tools reference.
• Ask comes with Conversation Intelligence, which is being rolled out to selected accounts.

25 September 2026 — Insights: topics, objections, knowledge gaps and chat#

• Your own taxonomy. Settings → Insights → Taxonomy: edit the topics, objections and intents the analyser uses, add your competitors and the names they go by, and publish them as versions. Each community can add its own topics. Conversations analysed from then on use the new version, and you can re-analyse up to 90 days after seeing an estimate. See The taxonomy.
• Topics and demand, and Objections and competitors: what people ask about and what stops them, how that changed, and how those conversations ended, with the best-handled ones to learn from (Conversation Intelligence).
• Knowledge gaps. Questions the assistant could not answer are grouped, counted and shown with the communities they came from. Draft an answer (it becomes a knowledge document) or link a fix, then see whether the question stopped coming. See Knowledge gaps.
• Chat, on every plan: the website chat funnel from the launcher to the target, where visitors stopped, answer buttons, return visitors and follow-up page chats.
• Two new flows: through each objection, and through the website chat funnel.
• Trends mark each taxonomy version as it is published.

25 September 2026 — Insights: leads and readiness#

• Leads (Conversation Intelligence). Every person is followed across their calls, website chats and follow-up page, one lead per community: known by their CRM record, a number or email they confirmed, or, for a website visitor who has not said who they are, their browser. See Leads.
• Readiness replaces warmth. Cold, Warm or Hot now counts what the lead did in every conversation as well as on their page: a visit that stands, a timeline they gave, coming to the visit, a cancellation with no rebook. Every point is shown with its reason, and silence halves the points every two weeks. The follow-up panel on a call and the daily CRM note show it; without Conversation Intelligence they use that call's own facts. See Readiness.
• Insights → Journey shows what happens after the conversation: target reached, page opened and came to the visit, by readiness and by how far leads went with their page, with what leads are interested in, how often family open a shared page, and whether readiness really predicts who comes. Flows gains the lead journey.
• Insights → Leads lists leads by readiness, momentum or last activity, and each lead has its own page with its journey. Staff can close a lead (moved in, chose elsewhere, not a fit…) and reopen it.
• WelcomeHome prospects' status (open, moved in, closed) is read once a day for active leads and shown on the lead and at the end of the journey.

25 September 2026 — Insights: experiments#

• Experiments. Try a new version of a playbook on part of a number's calls or a website key's chats, alongside the current one, with each conversation placed at random. Choose what you are testing before it starts, and how many conversations each version needs. See Experiments.
• Tested results. An experiment's result on the measure it was planned for is marked Tested once every version has enough conversations: the only comparison in Insights that can say one version did better, rather than did better during its period.
• An experiment ends by itself at its end date or once it has enough conversations, and stops if a new version of its playbook is published or where it runs changes. Conversations go straight back to the usual version.

25 September 2026 — Insights: how the AI did, and Conversation Intelligence#

• How the AI did. Every call and chat now gets craft checks (quick answers, short turns, one question at a time, no talking over the caller, no asking twice) and a craft score, with compliance counted beside them. The scorecard shows the craft score and each check's pass rate. See Insights.
• Conversation Intelligence (a paid add-on, being rolled out to selected accounts): each conversation is read and labelled with who called and why, topics, sentiment, objections, buying signals and a summary. Every label shows the words it came from, and can be corrected in one click. See Conversation Intelligence.
• Settings → Insights gains the list of topics the analyser uses and a daily analysis budget.

24 September 2026 — Insights: comparisons and flows#

• Insights in the main menu: a scorecard for the account, a community, a playbook or a version, and views to compare them side by side, as a matrix, as a ranking and over time. See Insights.
• Honest comparisons. Every rate shows how many conversations it rests on and its likely range, and is called better or worse only when the evidence supports it. Compare adjusts for differences in traffic.
• Flows draw where conversations go: through the stages, the lines of questioning, and from the chance to the result. Any step lists the conversations behind it.
• Charts in Analytics and Insights have a new look, and every chart can still be shown as a table.

24 September 2026 — Insights: objectives#

• Objectives. Each playbook says what it is for on a new Objectives tab: a main objective, fallbacks and other objectives, and its own for website and follow-up page chat. Every conversation is judged against them, and the booking rules choose each conversation's target, so a caller the rules sent to a callback is not a missed tour. See Insights.
• Whose miss was it. The call page shows when the assistant had the chance, when it tried, and when a conversation fell short, whether that was the AI, the playbook, operations or the caller.
• Simulate on past conversations tries a draft's objectives on the last 30 days first.
• Settings → Insights sets what each objective is worth.
• Analytics and Calls show one channel at a time. Phone calls, website chats and follow-up page chats are no longer mixed in one rate, and a chat's length no longer counts the half hour after its last message.
• A playbook whose objective is a booking now needs a calendar connected before it can be published.

24 September 2026 — Fewer "is that right?" questions#

• The assistant confirms like a person. When a caller says something plainly ("memory care for my mom"), the assistant says it back in passing as part of the conversation instead of asking whether it heard right. If the caller carries on without correcting it, it counts as confirmed. It still asks outright when it had to read between the lines, or when the caller pushed back.
• A clear answer to a question is an answer. However the assistant worded the question, a plain reply is taken as the answer straight away and never checked back.
• Always confirm the answer, a new setting on each question, for answers that would take the call the wrong way if misheard. The assistant reads those back together as its notes ("just so I have my notes right: …") and nothing acts on them until the caller says yes. See Questions and flows.
• Changed answers are kept up to date. When a caller takes something back ("sorry, I'm confused, it's for my aunt"), the answer is changed straight away, and any rule that depends on it is checked again.
• Yes-or-no answers are heard in passing. "My aunt is a veteran, do you have any deals?" now answers your veteran question as well as asking one. Add the phrases callers use under Callers say it with to help it hear more.
• The call debugger shows answers confirmed these ways as Confirmed by carrying on and Confirmed when read back.

23 September 2026 — Reps' own calendars, and Microsoft 365#

• Each rep's own calendar. Every rep connects their own Google or Microsoft 365 calendar from My calendar. Tours are booked on the calendar of the rep giving them, only when they are free. See Each rep's own calendar.
• A Sales rep role for people who only give tours: they connect their calendar and see nothing else of your account.
• Staff emails can now be edited on a community's staff list, which shows whether each rep has connected a calendar.

23 September 2026 — Google Calendar#

• Book on Google Calendar. Connect a Google Workspace calendar by signing in with Google, and choose it as the playbook's calendar. The assistant offers times when the right staff member is free, and books the meeting on your booking calendar with them invited. Your CRM stays the CRM. See Google Calendar.
• Reconnect on any connection that signs in with its provider, when access lapses.
• Callers get a Google invitation only if you turn it on, and then the booking notes stay off the event.

22 September 2026#

On the call#

• Choices can carry the words your callers use. Each choice on a question now has Callers also say: "mom, mum, mother" on A parent, "dementia, memory loss" on Memory care. The assistant hears those as that choice, whether it has asked the question yet or not, so "it's for my mother" in the first sentence is checked back later rather than asked for. Choices you leave empty behave exactly as before.
• What a caller says before you ask is kept, and checked back. A caller who opens with "memory care for my mother" is no longer asked for it again. The assistant holds what they volunteered as mentioned, not confirmed, checks it back in its own words ("you mentioned memory care, have I got that right?"), and only then counts it as an answer. A caller who says the same thing three times is not asked a fourth.
• Nothing unconfirmed decides anything. Until the caller confirms it, no booking or not-a-fit rule reads it, it does not satisfy Required before booking, and your reports do not count it.
• It is still on the record. The call page shows it under What the caller told us, labelled mentioned, not confirmed, and the CRM activity says it was mentioned rather than writing it into a CRM field.

Testing a call#

• The map follows the call. While you are on a test call, the question map zooms to whatever the assistant has just reached: the question it saved, the line it switched to, or the ending it landed on. A call you open afterwards stays where you put it.

Security and data#

• Connections use HTTPS. A new webhook URL or WelcomeHome API address must start with https://. A connection saved over http:// before keeps working, with a warning on its card until you reconnect it with an https:// address.
• Tool-server contacts update on the same call. When a tool server is your CRM, a second save of a caller's details during one call (a new email, say) now reaches it instead of being taken for a retry.
• Queue history is kept 30 days. Copies of delivered CRM writes, tool calls and messages are deleted after 30 days.
• Discarding a tested draft keeps its test calls readable. A draft you made test calls on is hidden instead of deleted, so those calls still show what they ran. The editor starts again from the published version, and the next draft gets a new version number.
• Team members join by invite only. People are added only with Invite a team member on the Team page, which also checks that a login belongs to one account.
• Only a publish changes what runs. What each community runs is saved only by AnswerStack after the publish checks; members keep seeing it.
• App access covers partner software. Turning connected apps off, or a suspension of app access, now also stops service identities' OAuth clients on the MCP server. Their API tokens keep working on the REST API.
• Fixes. Two jobs for the same knowledge document no longer collide; a tool-server connection is no longer left half-saved when storing its credential fails.

Billing#

• Issued invoices never change. A correction for an earlier period, or a cancellation inside your grace window that comes after that period's invoice, goes on your next invoice instead, labelled with the period it belongs to.
• Moving a tour is not a cancellation. A tour moved to another time earns no cancellation credit, because the caller still has a tour. A call earns the credit once, however many tours it booked.
• Invoices appear once issued. An invoice we are still preparing is no longer shown on your Usage page.
• One billing time zone. Every billing date (when a period starts and ends, and when promotions, adjustments and credits apply) is read in your account's billing time zone, which is UTC.

Texting and the follow-up page#

Two texts change wording: the re-subscribe confirmation is new, and meeting times now name the time zone.

• Only a bare STOP opts out. STOP, CANCEL, END, QUIT, UNSUBSCRIBE and OPT OUT opt a number out when they are the whole reply, however they are typed. "Cancel my tour" is a request: your team is told as for a callback, and the caller hears that someone will follow up, with the STOP line kept.
• START re-subscribes. START, UNSTOP or YES from a number that replied STOP restores its texts, clears the CRM's do-not-text flag the STOP set, and sends one confirmation. The call record shows when.
• Times say their zone. "9:00 AM Central" on the page, in {{meeting.time}} and in the times offered to move a tour.
• Buttons, one half at a time. A card or section button saves with only its words or only its link. The checklist says what is missing, and the page leaves it off until it is complete.
• Page conditions are checked. A "who gets a page" or "show when" rule on something nothing provides now stops publishing, like a booking rule would.
• Fixes. The test text uses the playbook you have open. A tour that moves after its reminder gets a reminder for the new time. Two quick presses of Send the text again send one text. The email copy of the link can no longer be lost to a database error, and the CRM's do-not-contact check on an email looks the contact up by email address. Staff are no longer emailed twice about a callback when recording the first email fails.

Calls and the builder#

• Outcomes show how far a call got. A call that was booked, transferred, took a callback, was declined, was not a fit or was redirected keeps that outcome even when it ended on a hang-up or the 20-minute limit. Caller hung up now means nothing decisive happened.
• A refused transfer counts as failed. A backup line that rejects the call (SIP 607 or 608) is recorded as a failed transfer, not an unanswered one, and the next number is still tried.
• A transfer rule's reason is optional. Left empty, the person taking the call hears "Booking rule:" and the rule's name. The reason box no longer snaps back to "Hot lead" while you type.
• Customizations cannot strand a rule. Turning off a question that a booking rule, condition or tool still reads is refused on save, with the same message the publish checklist shows.
• Call lists keep up. The past calls list and an open call's page reload as calls change, at most once every 5 seconds and only while on screen.
• Revoking a catalog tool asks first, and says how many accounts use it.
• Checklist links land on the question. A publish checklist message about a question opens its line of questioning and puts the cursor in it; inline errors on the question map point at the right question.
• WelcomeHome. Large accounts' calendars now load over a few calls instead of timing out on every one, and a returning prospect's first call is no longer overwritten when WelcomeHome's search is unavailable.
• Phone numbers in logs. Numbers written in national format, such as "020 7946 0958", are now masked in our logs as well.
• Fixes. Most characters the agent ever sees and Questions before giving something useful can be cleared while typing; Try it sends a tool's number choices as numbers; spend alerts check the budget before saving; a widget key's websites box starts from what the key allows now; the Settings page shows the saved words after saving; a community's custom values keep the name their placeholders use when the profile is saved; turning a follow-up link off no longer promises a new link can be sent.

Connectors we offer#

The Add a connection list on the Connections page now shows only the connectors we offer for new connections. Nothing changes on a call.

• Your connections keep working. A connection you already have keeps working in calls, checks and playbooks even if we stop offering its connector, and it stays on the Connections page.
• Your sandbox CRM stays. The practice CRM made with your account is always available.

21 September 2026 — A more natural manner#

The assistant now talks like a person, not a form. This changes how the assistant sounds on a call.

• Your questions, in its own words. The questions in your lines of questioning are a guide to what to find out. The assistant no longer reads them out word for word; it keeps their meaning and any choices they list.
• It listens first. It responds to what the caller said before asking the next thing, bridges from one topic to the next, and stays with the moment when a caller shares something hard.
• Gentler asks. Personal questions come with a reason, and contact details are asked for one at a time, casually.

21 September 2026 — Steadier calls#

Calls no longer end or derail on a misheard word. This changes how the assistant behaves on a call.

• One sentence, one decision. The assistant starts thinking before the caller finishes. What it decides now counts only once the caller's sentence is final, so a half-heard word can no longer end the call, move it to goodbye, or count as a second warning.
• Context, not keywords. "I hope I didn't dial the wrong number", "she had a stroke last year" and "what a pathetic situation at her old place" are read as the ordinary sentences they are.
• Wrong numbers are confirmed. After the redirect line the assistant asks whether there is anything else, and only ends the call when the caller says no.
• "Someone about memory care." A caller who asks for someone about a topic is offered help right away or a person, their choice. "A real person", "a human" or "representative" still transfers at once.
• No dead ends. A caller who corrects the answer that ruled them out goes back to the questions. After a transfer nobody answered, a caller can keep asking questions. A "none of" not-a-fit rule waits until its question is answered.
• A slow calendar gets a second try. The first calendar failure lets the assistant try again; the second offers a callback. The caller can still ask questions either way.
• Voice service failures. If speech recognition, the model or the voice fails, the caller hears your fallback line, then your backup number, instead of a dead line. The call record says why the call ended.
• One voice engine. The voice engine setting is gone: every call listens, thinks in text and speaks, so every guardrail checks each sentence before it is said. Playbooks that used speech-to-speech now run this way. OpenAI voices speak with OpenAI's text-to-speech model, and Google voices are no longer offered (they need speech-to-speech); pick another voice before you publish such a playbook again.

21 September 2026 — The call debugger#

See what the assistant decided, and why, on every call.

• Live while you test. Call the draft now opens a debugger beside the call. The question map outlines the line of questioning the assistant is on and ticks each answer as it lands. A timeline shows each decision as it happens: rules checked and why, answers saved, tools used with what went in and what came back, and guardrails that stepped in.
• Afterwards, on any call. Every call's page has a Call debugger with the same view, for real calls and test calls alike. Problems only jumps to what went wrong, and Timing shows where a caller waited.
• Private by default. Sensitive answers stay hidden until you press Show sensitive answers. Opening the debugger and showing sensitive answers are both written to the audit log. The debugger follows the transcript's retention.

21 September 2026 — Tools and connected apps (MCP)#

AnswerStack now speaks MCP (Model Context Protocol) both ways. The assistant can use tools from your other systems on a call, and your own AI apps and scripts can work with your account.

On the call#

• Tools from your other systems. The assistant can look things up while it talks, such as open units or a price list, and say the answer. It states facts only from your brief, the community's profile, your knowledge, and the answer of a tool you approved, on that call.
• Code decides when. A tool runs when the assistant decides to look something up, in the parts of the call you allow, or automatically at a fixed moment: when the call starts, once the caller confirms who they are, when an answer is recorded, before offering times, after a booking, or at the end of the call.
• You choose what may be said. Each value a tool returns is either sayable or for your rules only. Rules-only values, such as a lead score, feed flows and booking rules and never reach the conversation.
• Tools that change something ask the caller before they act, run at most twice on a call, and never act twice on a retry.
• A tool that needs one more detail has the assistant ask the caller in its own words, never for payment, ID or passwords.
• A slow or failing tool never stops a call: you choose what the assistant says instead.

Building it#

• Tools in the main menu: add an MCP server with no sign-in, an API key or the provider's own sign-in, review each tool, write the description the assistant reads, and Try it.
• Pinned tools. If a server changes a tool after you approved it, the tool stops running until you approve the new version. A server that fails often is paused and resumes on its own.
• The tool catalog: servers AnswerStack offers with their tools already approved. Turn one on, or connect your own account with the vendor.
• A Tools tab in the playbook editor, written as sentences like the booking rules, with group customization to keep or turn off each tool per community.
• A tool server as your CRM or calendar. Map each step, such as finding a caller or booking a meeting, to one of a vendor's tools. Saving runs read-only checks, and a guided test proves the writes.

For your team and your software#

• Connect AI apps. Claude, ChatGPT or any MCP client can connect to your account after a member approves it on a consent screen: which permissions, which communities, and for how long. Writes are never ticked for you.
• AnswerStack MCP tools: calls, transcripts, analytics, bookings, callbacks, playbooks, knowledge, community profiles and usage. An app can edit a playbook draft but never publish it.
• Connected apps in the new Developer area, with rules for owners and admins: turn apps off, choose who may give write access, or allow only approved apps.
• API tokens and service identities. Personal tokens up to your own role, service identities for software, and OAuth clients for partners, on the MCP server and on REST routes that accept them.
• A callback marked done from an app or the API can carry a note.

Known limits#

• Tools run on remote servers over HTTPS only.
• A community can keep or turn off a tool, but not change what it sends.
• Compliance mode blocks apps and tokens from reading transcripts, recordings and callers' details.

21 September 2026 — Website chat#

• The website chat answers now. Visitors type, and the same assistant from your phone line answers: the same knowledge, booking rules and guardrails, and it says it is an AI first. See Website chat.
• The Web widget tab has a live preview (desktop and phone, open and closed, light and dark) and the code to put it on your website: a bubble, your own button, inside a page, a tag manager, or React.
• The follow-up page carries the chat when you turn it on under Chat, and it can pick up where the call left off. When it does, it says so: a dated note at the top and its own greeting, which you can change on the Web widget tab.
• A website chat appears under Calls, marked as a chat. It is billed per visitor message and per booking, never by the minute.

21 September 2026 — A roomier page editor#

• Edit the page on the Follow-up page tab now opens the editor full screen: sections on the left, the preview in the middle, the selected section on the right.
• The preview shows the page at phone, tablet and desktop size, and Preview only gives it the whole screen.
• A section's settings are headed with its name and split into Content and Who and when. Colours, typeface and logo moved to Page style, so nothing in the section panel changes the whole page.
• Card text, Before you come and your own sections can use bold, italic, a subheading, lists and links. Nothing else — see Formatting.
• Your own section can now be added from Add something, and can carry a picture and a video — the place for a welcome. Who you will meet says where each community adds its staff videos.
• Opening words and Chat sit above the sections in the editor. Chat has a switch for whether it knows about the call; the chat itself arrives with the website chat.

Nothing about calls changes.

20 September 2026 — Follow-up pages#

After a call, the assistant can text the caller a link to a page that is theirs alone. Nothing is texted without permission captured on that call.

For the caller#

• A page of their own with their tour details, a recap of what they told you, the questions you could not answer, and an introduction to the counselor they will meet.
• Cards that respond to what they asked about — the pet policy for somebody who mentioned a dog — conditional on what was said, never on who the caller appears to be.
• Move or cancel the tour from the page, under the same booking rules the assistant used on the call. A caller who was not a fit on the phone does not become one on a web page.
• A link for family with its own shorter expiry, showing the place, the person and the directions and none of the caller's own details.
• Reminders 24 hours and 2 hours before, and an optional note the morning after. Nothing sends between 9pm and 8am in the community's time zone.
• Email instead of a text, for a caller who declines a text and agrees to email.
• Built for a phone: WCAG AA, 18px text, and nothing loads from a video host until they tap play.

For your team#

• A builder with a live phone preview of a real caller's page, a switch for every page state, and a Send me a test text button.
• Per community: their own staff and videos, photos, directions and local cards, within whatever the account leaves unlocked.
• On the call record: whether the text went out and where it got to, what the caller did on the page, and a Cold / Warm / Hot indicator with its reasons written next to it.
• In your CRM: the first open, a share, a booking change and form answers as they happen, and everything else as one tidy activity per contact per day.
• Show rate in Analytics, split by whether the caller opened their page — read back from your CRM tour by tour, and reported as not known rather than guessed when nobody has marked a tour up.
• Two more meters: text messages and chat messages.

Known limits#

• A community can customize cards and content but cannot reorder or reword the sections locally.
• Recording is not offered on the page in this release.
• Chat on the page is not joined up yet.
• Link previews show your community's name and logo, not the caller's tour.

19 September 2026 — First release#

The first release of AnswerStack: an AI assistant that answers your inbound sales calls, consults with the caller, and books a tour in your CRM.

On the call#

• Answers inbound calls in your community's name, in a stock voice you choose.
• Says it is an AI in the first sentence of every call, followed by the recording notice and your opening question. Neither the disclosure nor the notice can be switched off.
• Recognises returning callers from your CRM and greets them by name once they confirm who they are.
• Asks your questions one at a time, in your words, and never re-asks something it already knows.
• Answers from your brief, the community's profile and your knowledge — and says so plainly when the answer is not there.
• Offers two specific times and books through your CRM, reading the confirmation back.
• Hands a caller to a person whenever they ask, and takes a callback when nobody is available.
• Tells a caller in an emergency to call 911, refuses medical, legal and financial advice, and refuses card numbers, passwords and ID numbers.
• Stops recording and storing the transcript immediately if a caller objects.
• Calls are capped at 20 minutes, with a check-in after eight seconds of silence.

Building it#

• Playbooks with a six-step guided setup: company and offer, voice, questions, booking rules, guardrails and CRM. A tabbed editor afterwards, with autosave, undo and redo.
• Booking rules as sentences you build from dropdowns, with a template gallery, a plain-words summary on every rule, a Try it panel, replay against past calls, and warnings that explain themselves.
• Questions as cards: the words the assistant says, an answer type, a required toggle, a sensitive toggle, skip rules, not-a-fit rules, and a sample conversation preview.
• Guardrails with sensible defaults already on, safety lines, never-say phrases and redirect lines.
• Knowledge: uploaded text, Markdown, HTML and CSV files, web pages and pasted text, at account or community level, with an Ask it a question box that shows what the assistant would answer from.
• Communities: a profile for each location with its own hours, booking windows, staff, price points and talking points, filled into the account playbook's placeholders.
• Customizations per community, with locks so the account keeps what must stay the same, and an overview grid showing where each community differs.
• Your words: choose what a location, a booking and a caller record are called, everywhere in the app and on the call.

Going live#

• Publish checklist that blocks a publish until the disclosure, the questions, the catch-all rule, the handoff lines and the connections are all in place.
• Version history with plain-language differences and one-click restore.
• Kill switch per playbook: route every call to a person or to voicemail immediately.
• Phone numbers that follow the latest published version or stay pinned to one, plus a test number that can run a draft.
• Call the draft from the browser, with a live transcript and the booking rule that fired.

Watching it work#

• Calls with filters, live calls, transcripts, the rule that fired, guardrail hits, recordings and a callbacks queue.
• Analytics: booking rate, qualified booking rate, transfer rate, drop-off by stage, response times, cost per call and per tour, top unanswered questions, and handoffs by community.
• Usage on the meters — minutes, calls answered and calls that booked at this release — by community and by number, with optional spend alerts and past invoices.

Integrations#

• WelcomeHome as CRM and calendar: caller lookup by phone and email, history, prospect creation with resident and influencer and their relationship, call activities with a summary, tour activities with a counselor and a calendar entry, and do-not-call honoured.
• Sandbox CRM for every account, with sample data, an activity feed, an API log and demo tools.
• Generic webhook for anything else, with signed payloads and retries.
• API with the same roles and limits as the app, and a reference generated from the API itself.

Known limits in this release#

• Inbound calls only. The assistant never dials out.
• English only.
• Introducing the caller before a transfer needs an outbound calling line on our side. Without one, the setting falls back to passing the call straight through.
• Notifications are emailed, and texted as well once your account has a registered sending number.
• Knowledge accepts text, Markdown, HTML and CSV files. PDF and Word are not supported yet.
• Only stock voices. No custom or cloned voices.
• Not offered as a HIPAA-covered service. See HIPAA status.
• Publishing is immediate. There is no scheduled publish or staged rollout.