
Interaction data capture tool - your bot can now capture specific data points from a conversation, so key details are recorded automatically as the call happens.
Business hours tool - expose your business hours to your bot, then pair it with the transfer tool so calls only transfer to a live voice conversation during your opening hours.
You can now test your voice bots with an automated test suite. Build testing scenarios for a bot to catch regressions before they reach customers, then re-run those scenarios each time you make a change to confirm everything still behaves as expected.

Voice Bot Tool Names - Voice bots now read the tools’ names as well as their prompts. This should give your voice bot greater accuracy in selecting the correct tool.
Auto Assign Tag Accuracy - Automated tag assignment for interactions has been upgraded for better accuracy. For best results, remember to give your tags meaningful descriptions.
LLM Scenario Tool - A Scenario tool has been added to LLM Chatbot. This tool allows you to define more specific prompts that will only be added at the appropriate time during a conversation.
Outbound SMS Canned Message Labels - When sending outbound SMS messages, the canned message selector now contains the message short-code label for quicker identification of long canned messages. You can also search for canned messages from inside the selector.
Parking Lot URLs - Interaction parking lot configuration pages now contain sample URLs. These URLs can be used by system integrators to more easily debug connections between Talkative and Mitel.
Average Messages Report - A new statistic for the average number of messages sent per interaction has been added to the analytics suite.
Voice AI Cut Off - Fixed a transcription cut-off issue on calls where the customer hangs up mid-conversation.
Analytics Large Chart Scrolling - Fixed a scrolling issue on analytics charts that use long labels.
URL Rendering - Fixed an issue where URLs rendered incorrectly when followed by a dot character.
AI Knowledge Base UI - Fixed a text overflow issue in the AI Knowledge Base editor overview section.

The Talkative system should now be much faster and snappier to use thanks to some optimisations made. You can expect around an 80% improvement to the loading speeds across the platform.
You can now enable a setting to make Voice AI consider calls from the same number to be coming from the same customer. With this enabled, agents will be able to view previous conversations from the same phone number in your interaction logs.
💡 Currently the previous call history is not given to Voice AI bots, this means the Voice bots will not be able to reference previous conversations.
To enable this, search for “Phone Call Attribution” within the settings

⚠️ PSTN numbers can be spoofed, which may lead to false attribution of interactions. You should not rely on this as the only strategy for identifying a caller.
Speech to Speech Interrupts - The interrupts toggle on Voice AI Speech to Speech mode will be more reliable.
Speech to Speech Reliability - Improved overall Speech to Speech reliability with interrupts and noise cancelling.
Google Speech to Text to Speech Depricated - Google Speech Text to Speech has now been deprecated for Voice AI. Any bots using this should be moved to ElevenLabs.
Voice AI Default Settings - New Voice AI configs will now default to “Speech to Text to Speech” instead of “Speech to Speech”.
Rerun Old Knowledge Base Test Runs - When running test runs within Knowledge Bases, you will now be asked if you would like to run a specific previous test run.
Canned Messages Indexing - Canned messages are sorted alphabetically on the configuration page.
New Company Setup - Newly created companies now have Tags and Wrap up forms available in their Voice AI queues.
Analytics CSV Downloads - Analytics CSV downloads has now been reformatted and improved.
Billing Page Formatting - The billing page has improved number formatting.
Live Dashboard Timestamps - Live Dashboard now shows the full timestamp in charts.
Licencing Notifications - If a licensing limit is reached, then increased, the customer will be notified if they are approaching their new limit via email.
Knowledge Base Back Button - The back button in AI Knowledge Bases now works as expected.
Side Menu UI - The monitoring sidebar no longer resizes if users have a particularly long first or last name.
AI Knowledge Base UI - Fixed AI Knowledge Base test runs UI on smaller screens overflowing.
Log In Issue - Fixed issue where some users would be unable to log in if they had not logged in in over 2 years.
Depricated Integrations Hidden - Hide not-working Google Business Manager and Twitter integrations.

Store sensitive values - API tokens, login credentials, and other secrets - securely inside Talkative, then use them in your voice bot's HTTP Request tools.
Why it matters
Your secrets are never exposed in Interaction Logs or tool recordings.
Talkative staff cannot view your stored credentials.
A safer way for security-conscious teams to connect Voice AI to their systems.
Search for Company Credentials to get started. 💡 Currently usable in HTTP Request tools within Voice AI only.

Voice AI (in Speech-to-Text-to-Speech mode) now understands digits entered on a customer's dial pad. Prompt your bot to listen for keypad input to build hybrid voice/IVR flows.
Why it matters
Capture account numbers, reference codes, and phone numbers reliably.
Build familiar IVR-style journeys inside a natural voice conversation.
Test easily - the Voice AI browser testing tool now includes a dial pad.
💡 DTMF is not currently supported in Speech-to-Speech mode.

Escalate to Email for LLM Chatbot - chatbots can now escalate conversations to email, with different email paths configurable per tool.
Tag Context field - add a short description to each tag so AI Auto Suggest and AI Auto Assign tag more accurately.
Clearer Knowledge Base messaging - the Knowledge Base now makes clear that Voice users can keep using it even with no AI Interactions.
OpenAI US region - requests in the US region now always use OpenAI's US servers.
Custom email report now shows the total number of usages rather than total tokens used.
Fixed formatting on the AI Knowledge Base prompt preview.
Improved interaction end behaviour when a customer and agent end at the same time.
Fixed an issue where video could keep making noise on the customer side after the agent hung up.

Pronunciations
You can now control how your Voice AI pronounces specific words and phrases in Speech to Speech mode. Navigate to the Pronunciation Dictionaries page to configure a dictionary using IPA pronunciations or word substitutions, then point your Voice AI bot to it.

This does not work in Speech to Speech mode. If using Google Text to Speech, we recommend using substitutions instead of pronunciations.
Call Logs Now Visible
The Twilio Call Log page now displays recent calls, making it easier to identify the exact SIP URL format needed when configuring Voice AI.

EU Localisation
LLM usage in Voice AI is now handled via EU-based servers for London and Frankfurt instances. This means Speech to Speech is now fully EU-processed. Speech to Text to Speech and other instances are unaffected.
Improved Interrupts
Voice AI in Speech to Text to Speech mode will now handle short, sharp interrupts (No/Yes) better.
More Accurate Transcripts
In Speech to Text to Speech mode, Voice AI now tracks how much of a message it has delivered when interrupted. This leads to more accurate responses and cleaner Interaction Log transcripts.
LLM Chatbots now show tool uses in Interaction Logs
Tool used by your LLM Chatbot are now visible in Interaction Logs, making it easier to review your chatbot configuration.

Historic Billing and Usage View
Billing history is now available on the Billing page, allowing you to review usage from previous months.
Improved loading speeds in the Supervisor Dashboard and Chat Monitoring.
Improved Interaction Log performance for companies with large interaction volumes.
Analytics charts with too many X-axis items are now hidden to improve readability and performance.
The SharePoint Integration can now be restricted to specific sites.
AI Knowledge Base emails now respect whitelabelling settings.
Improved Agent Console performance.
Improved Custom Reports speed.
Better error handling when a token limit is hit while scraping a website in the AI Knowledge Base.
Updated the Talkative Chatbot AI template to be more helpful when creating new bots.
Fixed an issue where Deepgram transcription would intermittently fail.
Fixed a display issue with the message input in the Agent Console in certain edge cases.

Improved Voice AI Interrupts
Voice AI now handles interruptions much more smoothly when operating in speech mode. Conversations feel more natural and responsive, with fewer awkward pauses or missed cues.
Voice AI and chatbots can now automatically tag interactions. To enable this, please see this article.
The "Time saved with chatbots" metric now correctly reflects that voice bots are included in that figure, the measurement itself hasn't changed, just the label.
A new custom report is available for tracking Email Assistant usage. Reach out to your Customer Success Manager to get this set up.
You can now add existing tags when creating or editing queues, making it easier to keep things consistently labelled.
Various interface polish improvements, including cleaner-looking search fields.
The Interaction Parking API now supports a flat format option. By appending ?format=flat to your unpark URLs, interaction data is returned as a flattened one-dimensional JSON object.
Interaction Parking Lots now support prefixes, meaning they can be used regardless of your SIP URL structure.
Fixed an issue where building a new knowledge base could fail when scanning certain URLs.
Fixed a problem where deleting a queue would sometimes make your company settings page inaccessible.
Fixed a rare case where leaving a page with unsaved changes would not prompt you to save.

Regionalised EU OpenAI Models
OpenAI models used in Talkative (currently not Voice AI) can now process data via EU-based servers. This means data for features like the LLM Chatbot and Agent Assist can be handled within EU regions (eu/euc), helping support regional data requirements.
Internal Widget Deployments
It’s now easier to deploy your chat widget for internal use.
If you don’t have a website to host your widget, you can now generate a shareable link that always points to your widget. This allows anyone in your company to access and use it.
To set this up, click “Add widget to your site” in the config editor, then expand the “Deploying internally?” section.

Improved Voice AI Transfer Records
Voice AI transfers are now more clearly recorded in Interaction Logs. Transfers will show a structured record, making it easier to understand what happened during the interaction.

Billing alerts now include overall interaction usage.
You can now paste in sitemaps to add sources to your AI Knowledge Base (useful if automatic detection fails).
The Authorization header for Voice AI tool usage is now hidden from Interaction Logs.
Outbound SMS (with Mitel) now supports canned messages.
Non-Open Media users can now enable “Agents can see all interactions” in Company Security Config (useful for Voice AI setup within Mitel).
The full message input area in web chat is now clickable, not just the centre.
Added default values to advanced Deepgram options when creating Voice Assist.
Errors inside collapsed inputs now display correctly when submitting a form.
Interactions that time out after 24 hours now correctly send data to Salesforce and other CRMs.
Fixed an issue where Knowledge Bases could refresh to empty if multiple refreshes were triggered simultaneously.
Fixed an issue where inbound SMS messages could be dropped after enabling Voice AI on legacy companies.

I hope your first quarter has gone well, it's been a busy one here at Talkative. Total interactions up 26% across all channels, but the real success this quarter is Voice AI, which grew 8x.
The US market is moving fast on Voice AI adoption for 2026, with auto-attendants becoming the default way businesses handle inbound calls. March was when it really showed up for us, more customers brought their auto-attendants live and our usage jumped sharply!

Some of the key things that landed in Q1:
Digital AI - a prompt-based chatbot builder that ditches the flow chart entirely. Write how you want it to behave and it works the rest out.
AI Knowledge Base redesign - everything in one place, becoming the brain of your AI deployments. It’s now easier to see whats working and what needs improving, character limit was also upped to 20,000.
Voice AI upgrades across the board - longer prompts (25,000 characters), new endpointing controls, more languages, the ability to see incoming phone numbers, and advanced configuration reducing the number of support assistance needed.
We also shipped simpler screen sharing for video escalations, two new reporting exports for your team's performance data, weekly billing summaries, and a few smaller quality-of-life improvements.

Q2 is taking shape with a few clear priorities:
Agentic AI and autonomous workflows - moving beyond "answer this question" to "handle this whole task."
Platform stability and performance - making sure everything scales smoothly as usage grows.
Deeper integrations - more connection points so you're not jumping between tools.
CS & Partnerships expansion - we're bringing more people onto the team so support's faster and more proactive.
I've also opened up Talkative Tuesdays - drop-in time every Tuesday afternoon. Bring questions, ideas, something you're stuck on, or just a chat. No agenda needed.
If you want to dig into any of this, whether you're thinking about Voice AI, want a walkthrough of Digital AI, or just want to chat through what you're working on… please reach out!
Book a Talkative Tuesday slot, drop us a line at support, or message your CSM directly.

A new agentic AI chatbot that replaces the traditional decision tree approach. Instead of building out branches and nodes, you write a single prompt that tells the AI how to behave, and it handles the conversation naturally, deciding how to respond, when to search your Knowledge Base, and when to transfer to a human.
You can run it alongside your existing chatbot, so there's no pressure to switch everything over at once. To get started, search for "LLM Chatbot" in the settings bar. For a full guide on how to set it up, see this guide here.

This feature is in beta, if you have any feedback, let your CSM know.
When you escalate a chat to a video call, you can now offer the customer the option to share just their current browser tab, rather than their whole screen. This is quicker for the customer and better for their privacy, since they're only sharing the tab they're on rather than everything on their desktop.

This also works during video calls, you can offer tab sharing from there too.
If you don't see this option, you may need to save and publish your chat widget for it to appear.
Two new downloadable data exports are now available:
Reservation Ended Reasons - an export showing how many interaction reservations were rejected or timed out for each agent over a selected date range.
Agent Time by Status - an export showing how much time each agent spent in each status, and how much time they spent handling interactions, over a selected date range.

This feature is not switched on by default - if you'd like access, reach out to your CSM.
The screen where you manage your Voice AI tools has been tidied up. It now has clearer error messages when something goes wrong, and it's easier to move between tools.

You can now create Interaction Data via an HTTP API, making it easier to push data into Talkative from external systems. See the guide here.
Knowledge Base and Voice AI prompts will now warn you if you go over the prompt length limit.
URL rules in the widget editor now show the full URL where possible, making them easier to read.
New chat widgets now come with an email form included, which makes it easier to set up email escalation from a chatbot. Existing widgets are not affected.
Website scraping has been improved, it now picks up more pages each time it runs, and shows a loading indicator while it's working.
Interaction data has been moved to the top of the information section in the console, so it's easier to find.
Outbound SMS now has an option to not continue into a live interaction. Instead, any future messages from the customer will queue as a normal interaction. You can configure this in the Outbound SMS Interactions panel in settings.
New endpointing options for Voice AI, which let you adjust how quickly the AI jumps in when the customer pauses mid-sentence.
Added support for new languages in Voice AI (via Deepgram).
The input box for agent messages can now stretch further.
Added a Swedish language preset for chat widget text.
Fix page title sometimes showing stale page titles.
Add voice routes to settings menu