Welcome to your weekly roundup of FG Funnels updates through the week of October 22nd, 2026.
From Claude and Gemini models landing in Conversation AI to AI Studio gaining Git Sync, chat widgets, and checkout upgrades, bulk feature updates for sub-accounts, and Limited Time Offer templates for WhatsApp — this week’s updates bring some seriously powerful tools to your stack. We’re also rolling out upgrades to Events, Gift Cards, Voice calling, External Tracking, and more — all designed to save you time and give your clients a better experience.
If you have any questions about these updates, email us at [email protected] or reach out through live chat inside the app — we’re always happy to help.
Events can now be used in your users’ own language. All the user-facing text across Events has been translated using FG Funnels’ AI translation engine, so admins and attendees see Events in their selected language.
On the admin side, event creation, management, and settings are translated. On the attendee side, registration, checkout, tickets, and the Attendee App are translated too, down to labels, tooltips, validation messages, and empty states. Supported languages include English, Spanish, Canadian French, French, Brazilian Portuguese, Portuguese, German, Dutch, Italian, Danish, Norwegian, Finnish, and Swedish. Anything not yet translated simply shows in English, so nothing breaks.
AI Studio Payments just got more flexible, with customizable checkout fields, two-step lead capture, and detailed order confirmations.
You can hide the phone field and any billing address field, or make them optional or required (name and email are always required). A new two-step checkout works like the Two Step Order form in Funnels: step one saves the buyer as a CRM contact, so people who don’t pay are still leads, and step two takes the payment. After payment, the confirmation now shows the order ID, items, taxes, and total paid, plus the next charge for subscriptions, and you can send buyers to your own thank-you page that shows the real order.
Just ask AI Studio in plain language, like “Make phone required and hide the billing address,” “Make it a two-step checkout,” or “Send buyers to my own thank-you page after payment.” Projects already using Payments get all of this on their next prompt, with no need to reconnect.
You can now add native chat widgets to your AI Studio apps, built and styled by AI Studio to match your site instead of a generic bubble. Just ask AI Studio to “add a live chat,” and it builds a real FG Funnels chat widget right into your app. It runs on the same chat widget you already use, so your conversations, contacts, and widget settings are the ones already in your account, with nothing extra to set up.
AI Studio offers live chat, or email chat for visitors who leave a message and get a reply by email, and it asks a few quick design questions (greeting, placement, tone) before building. Since the chat is built into your app’s own code, it uses your site’s colors, fonts, and layout, and a project can have more than one widget. Security and spam protection are built in, and visitors can share attachments, pick up where they left off, and see delivery status on their messages. Business hours, out-of-hours messages, chat ratings, and every other widget setting are respected, and every message lands in Conversations.
Under More → Analytics → Chat Widgets, you’ll see how often the chat loaded and opened, the open rate, visitor satisfaction, and a table of every lead the widgets brought in. Under More → Chat Widgets, you can manage each widget without leaving AI Studio: rename it, edit messaging, set up the contact form, add compliance notices and consent, turn on spam protection, and set live chat business hours.
Live chat and email chat are supported in both TanStack Start and React projects. Voice chat, WhatsApp, Facebook Messenger, and Instagram chat widgets aren’t supported in AI Studio yet.
Connect a GitHub account and AI Studio creates a private repository for your project, pushes every version you save, and pulls back every commit made outside it. Open the repo in any editor or with any coding agent (Claude Code, Codex, Cursor, Copilot, Gemini CLI), and the changes land back in your project with a preview build.
To connect, open More → Git in a project and connect your own GitHub account or an organization you own. You can share an account with the whole sub-account or keep it Only me. Pushes and pulls happen automatically (both can be turned off, and Push now and Pull now are always available), the branch is set to main but you can switch it anytime, and when you pull, a fresh preview builds in the background with an Update preview banner so the page never reloads under you. If GitHub and AI Studio both have new commits, the Git screen flags the branch as diverged and gives you two ways out: Force push or New branch.
Sync history shows every commit that moved, with links to GitHub and plain-language explanations for any failures. Pushes are also scanned for secrets, so known token formats block the push, and likely hardcoded credentials trigger a warning.
There’s a new permission too, Manage Git & download code, found under Settings → Team → Permissions → AI Studio. It’s on by default for every role, so nobody loses the ability to download code, and agency admins can turn it off per user. Without it, the Git tab and every Download entry are hidden. Only the connected branch is pulled, and projects that never connect GitHub see no change.
Gift Card communications now support translations, so customers receive messages in the language of their selected locale. That covers purchase confirmations, recipient emails, and SMS notifications.
Previews in the Gift Card send flow reflect locale changes across every delivery method too, including email previews that show translated content with both sample and actual gift card data. Translations are handled automatically based on the selected locale.
Managed agents can now answer questions from real data across three new areas. For ad reporting, they can pull Facebook and Google Ads reporting rows. For Agent Studio, they can list managed agents, get an agent’s details, and execute a managed agent. For Courses, they can enroll or revoke a contact, list enrollments, and check progress and completions.
So you can ask things like “how many contacts are enrolled in my course, and how far along are they?” or “show me last week’s Facebook ad reporting” and get an answer based on your actual data.
There’s nothing new to grant. Each area keeps its existing permission (Agent Studio needs Settings, and Facebook and Google ad reporting need their matching Ads Reporting permissions), and everything stays bound to your account.
Number pools help you track where inbound calls come from, but until now you couldn’t take those calls in the web or mobile app, or ring several team members at once. Now you can.
You can send number pool calls to multiple team members at the same time, and each person can answer on the device they prefer: web, mobile, desk phone, or their own phone number.
Call tracking now labels Facebook ad calls correctly. Before, many calls from Facebook ads showed up as “Facebook” or “Direct,” so you couldn’t tell which ads were driving calls. They now show as Facebook Paid, and other ad platforms and AI referrals are covered too.
Found a trial chat that could have gone better? You can now turn it straight into an optimization run. Once your agent has replied in Test your agent, click Optimize Prompt, and the optimizer uses that trial conversation and the prompt currently in the builder (including unsaved edits) to generate fresh scenarios, test the agent, and try improved variations when needed.
The flow is simple: start a trial chat, run the optimization (testing uses an evaluation copy of your agent, and configured actions may run), then review the strongest completed variation with its accuracy, scenarios, transcripts, evaluations, reasoning, and prompt differences. Apply the improved prompt to the builder, then save the agent when you’re ready to make it live. You can follow progress in the builder, open the full Prompt Optimizer, or stop the remaining work without losing completed results.
The trial chat may clear after a reset or reload, but the optimization run stays available in Execution History. This is available in the Conversation AI builder when the Conversations AI Prompt Optimizer Labs feature is enabled.
External Tracking is now available inside AskAI, so you can connect, understand, troubleshoot, and analyze an external website just by chatting. Keep your existing website, connect it to the platform, and track supported website activity and form submissions without rebuilding or migrating anything.
AskAI can guide you through setup, check whether tracking is configured and receiving activity, list tracked domains, pages, and forms, and flag likely form-capture issues. It can also review visitors and the contacts they became, and pull page views, sessions, unique visitors, form submissions, contacts created, conversion rate, traffic sources, devices, and geography, along with attribution data like UTMs, campaigns, referrers, and landing pages. If something isn’t working, it can trace the journey from script installation through page views, form submissions, and contact creation to find the likely point of failure.
Read-only requests run directly, while any configuration change needs your explicit confirmation first, and existing permissions and access controls still apply.
How to Use
Open AskAI. Ask a question about connecting, tracking, troubleshooting, or analyzing your external website. Provide additional details when requested. Review the result or proposed action. Confirm any configuration-changing action.
Example requests: “Can I connect my existing website to my CRM?” “Is my tracking script working?” “Which pages generate the most leads?” “Why aren’t my website leads showing up?” “Which forms are being tracked?” “Where is my traffic coming from?” “How many leads came from Google?” “Is form tracking enabled?”
You can now automatically add Internal Comments to conversations using Workflows. Until now, teams had to add internal comments by hand to share context or keep teammates in the loop. With the new Add Internal Comment workflow action, you can automate it based on specific events, and even @mention team members directly in the comment.
For example, when a high-priority lead enters a workflow, you can automatically add a comment mentioning the assigned team member with next-step instructions. Or when a customer misses an appointment or a conversation needs escalation, you can add the relevant context and mention the right people, all without manual effort. These comments show up in the conversation history and nothing is ever sent to the customer.
How to Use
Go to Automation → Workflows. Create a new workflow or open an existing one. Add the Internal Comment action. Enter your comment and @mention team members as needed. Save and publish the workflow.
Creating and editing WhatsApp templates just got a modern upgrade. The new template builder (Settings → WhatsApp → Templates → Create / Edit) puts category, type, naming, body, media, buttons, and variables on one clean layout, with a live phone preview that updates as you edit, so what you see in the builder matches what the contact sees on WhatsApp. The AI template writer now uses the revamped builder end to end too.
Media headers also no longer need a hard-coded public URL. You can use custom fields or dynamic values, which resolve per contact at send time, so you can send things like invoices, IDs, or product images without cloning templates. It works across Workflows, conversations, bulk actions, calendar appointments, and review requests. If a dynamic value resolves to an invalid or empty URL, the send fails clearly, so you never get silently wrong media.
Let WhatsApp run the countdown for you. A limited time offer puts a live countdown clock inside the message itself, drawn by WhatsApp, ticking down in the customer’s chat and turning red in the final hour. The discount code sits right beneath it, ready to copy with a single tap.
To create one, go to Settings → WhatsApp → Templates → Create template, choose the Marketing category, then the Limited time offer type. You’ll set an offer heading, an image header, and body text, and two buttons are required by Meta: Copy code and a URL button. The expiry gets set when you send, so one approved template works for campaign after campaign. You can send from Conversations, Bulk actions, Workflows, and Calendar, and the AI template writer can draft one for you.
A salon can fill Friday’s empty chairs with a code that expires at noon, an online store can recover abandoned carts with a discount that visibly runs out, a gym can push a joining offer before the month closes, or a travel agency can move unsold seats on a deadline customers can watch shrinking.
Until now, every Conversation AI agent ran on an OpenAI model. Now you choose. Pick Claude from Anthropic, Gemini from Google, or any OpenAI model, agent by agent, from the same dropdown you already use. Same setup, same workflows, three times the choice.
Seven new models are available in Agent Builder settings: Claude Opus 5.5, Sonnet 5.5, Haiku 5.5, and Haiku 4.5, plus Gemini 3.8 Flash, 3.7 Flash, and 3.5 Flash Lite. Every existing OpenAI model stays right where it is. To use them, turn on Claude & Gemini Model Support in Labs, then go to Conversation AI → Edit your agent, pick a model from the dropdown, and optionally set a secondary model as a fallback. Save, then test the agent.
Every model was tested on everyday customer scenarios (booking, cancelling, rescheduling, triggering workflows, handing over to a human, and answering business questions) and graded by an independent judge, and every model clears roughly 96% on appointment conversations. Gemini 3.7 Flash (98.8%) and Gemini 3.8 Flash (98.5%) edge out GPT-4.1 (98.2%) on appointments. Claude Sonnet 5.5 hit a perfect 100% on picking the right action. And Claude Haiku 4.5 and GPT-6 Luna recognized every business question (100%), making them, along with Gemini, great picks for knowledge-base-heavy agents.
The new custom field types now work in the mobile app too, with stronger validation rules for your contacts and opportunities. Teams used to store links, meeting times, owners, and formatted notes in plain text fields, which can’t check a value, so wrong links, wrong times, and misspelled names got saved. Not anymore.
Five new field types are available. URL lets you store a link and open it with one tap (http, https, mailto, tel, and sms are supported). Date & Time and Time let you pick values that show in the sub-account time zone. User lets you pick one or more team members from a searchable list, shown as avatars. And Rich Text lets you write notes with bold text, lists, and links, with a preview on the record and a full editor when you need it.
You set validation rules on the web, and the app applies them when you edit a field: allow or block specific URL domains, limit dates to the past, future, or a date range, block specific weekdays, and limit times to a range like 09:00 to 17:00. The app shows a clear error right on the field when a value breaks a rule. On web, you can also add an image or icon to each option of a radio or checkbox field, and the app shows it next to the option. You can also filter opportunities by URL, Rich Text, Date & Time, and User fields.
To get started, create a field on web in your sub-account’s custom fields settings, set its validation rules, and save. Then open a contact or opportunity in the app, where the new fields show up in their folders, and tap any field to edit it.
FGF Notifications
