Welcome to your weekly roundup of FG Funnels updates through the week of September 8th, 2026.
From AI Image Generation and AI Analyze Image landing in Workflows to Ask AI now living inside WhatsApp, a redesigned My Profile & Users experience, and Session Variables for Voice AI — this week’s updates bring some seriously powerful tools to your stack. We’re also rolling out upgrades to Conversation AI, Courses, Calls, E-Commerce Stores, 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.
Ahead of a carrier-mandated deadline on September 15, 2026, toll-free verification now asks for a Privacy Policy URL and a Terms & Conditions URL as part of your messaging use case step.
Carriers will start rejecting any new toll-free verification submitted without both links after that date, so we’ve added the fields now to make sure you’re never caught in a non-compliant window.
Good to know: if you’re already verified, you don’t need to do anything — this only applies to new submissions.
Configuring a Voice AI agent just got less overwhelming. Advanced settings — Call Settings, Agent Behavior, Transcription & Speech, Voice Settings, and System Prompts — are now hidden by default, with a simple Show advanced settings toggle when you need them.
Timezone now has its own icon front and center in the toolbar, right alongside Voice, Model, and Language, so it’s easier to find and set correctly.
Bottom line: a cleaner builder for everyday use, with advanced options still there the moment you need them — and less risk of accidentally changing something you didn’t mean to.
S2S Voice AI Agents can now make outbound calls too, bringing lower latency and more natural, responsive conversations to your outbound campaigns — at a more cost-efficient price point for scaling up.
You can select S2S agents directly in the Voice AI Outbound Call workflow action, configure AI disclosure and opt-out messaging right in the Agent Builder, and use the very same agent for both inbound and outbound calling.
Bottom line: smoother conversations, lower cost at scale, and one agent to manage instead of two.
Voice AI agents can now use Session Variables — values that get captured during a call and flow across custom actions, prompts, and SMS as they’re gathered, instead of being typed as free text every time.
Create up to 20 variables per agent, each with a friendly label, optional description, and default value. Once a custom action captures a value mid-call, that captured value automatically overrides the default everywhere it’s used — prompts, SMS, transfers, and more.
Why it matters: your agent can reference something a caller told it 30 seconds ago — an order number, a preferred callback time — consistently, everywhere it needs to, without you writing custom logic for it.
Setting up a new store now comes with a guided assist from Ask AI. If you don’t have a store yet, clicking Complete Setup opens Ask AI to walk you through creation and setup, using context about where you are in the process to guide you to the next step.
Connecting a payment provider and setting up a new domain remain manual steps for security, but if you already have a domain in your account, Ask AI can help connect it for you.
How to Use It: 1. Navigate to Ecommerce Stores and click Complete Setup. 2. If you’re setting up your first store, Ask AI will open and guide you through the setup. 3. Follow the recommended steps and complete any required manual actions such as payment setup. 4. Return to Complete Setup anytime to continue your Ask AI-assisted onboarding.
Good to know: payment credentials are never collected or configured through Ask AI.
You can now block checkout entirely when no shipping method matches a customer’s delivery address — so you’re never stuck fulfilling an order at an unintended rate, or one that can’t ship at all.
Turn on Block Shipping under Shipping & Delivery, and customers see a clear “Shipping not available” message with the option to remove the affected item or clear their cart. Shipping is automatically re-checked whenever they update their address, and you can customize the error message colors to match your store’s branding.
How to Use It: 1. Go to Settings → Shipping & Delivery. 2. Enable the Block Shipping option. 3. Configure your shipping zones, profiles, and conditions as required. 4. Publish your store to apply the latest setting.
This setting is off by default, so nothing changes for your store unless you turn it on. Digital-only carts aren’t affected.
A batch of stability improvements landed for AI Studio this week: sandbox HMR issues fixed for Vite 8 compatibility, clearer build error diagnostics, and a code editor that now loads your real project files with proper search across routes and generated files.
We also fixed rolling deploys killing in-progress agent runs, large project uploads hitting size limits, and a few code explorer bugs around file search and project switching — all in service of a more dependable building experience.
AI Studio can now run commands inside your project to check what’s actually happening — not just generate code and hope it works.
This means the agent can inspect your running app, dig into the real cause of an error instead of guessing, test that an integration actually works, recover from a stuck or outdated preview, and verify a fix before telling you it’s done.
For example, if you report “the menu still doesn’t open,” the agent can check the live page, discover the preview was just showing an outdated version, refresh it, and confirm the menu now works — all in the same request.
Why it matters: fewer guess-based fixes, less back-and-forth explaining the same bug twice, and a better shot at getting a working result the first time.
Dynamic value chips in Workflows now have three states instead of two, so you can tell at a glance whether something is broken or just not connected yet.
Why it matters: a configuration you built on purpose — like referencing a value that gets passed in from another workflow — no longer looks like a mistake you made. Red is now reserved for actual typos and errors.
A new AI Image Generation action is live in Workflows. Give it a prompt, and it creates an image, stores it automatically, and hands you back a URL you can drop into emails, SMS, social posts, or external systems.
Choose from four image models, use the Enhance Prompt button to expand a short idea into a detailed one, start from templates (Product Mockups, Social Post, Promo Flyer, Lifestyle), and apply your Design Kit and Brand Voice so the output stays on-brand. Prompts can come from a contact field, a webhook payload, or a previous step — so you can generate a unique, on-brand image for every contact or every row in a spreadsheet.
How to Use: 1. Go to Automation, then Workflows. Open a workflow. 2. Click the plus icon. Select AI Image Generation under AI Actions. 3. Select a model. Set the quality. Auto is the default. 4. Give the action a prompt. Type it in the prompt field, or insert a custom value so the prompt comes from your data at run time. 5. Click Enhance Prompt to add detail, or start from a template. These steps are optional. 6. Open Additional Settings to set the size, the background, the output format, the Design Kit, and the Brand Voice. This step is optional. 7. Save the action. Use the Image URL variable in the next actions.
The new AI Analyze Image action does the opposite of AI Image Generation — instead of creating an image, it reads one and returns the answer as plain text.
Use it to extract text from receipts, invoices, and screenshots, describe what’s in an image, or ask any free-form question about it. The output can go straight into a custom field, an email, an SMS, a webhook, or even branch your workflow based on what the image contains.
How to Use: 1. Go to Automation > Workflows. Open a workflow. 2. Click the plus icon. Select AI Analyze Image in AI Actions. 3. Select a model. Keep GPT-5.6 Luna if you are not sure. 4. Enter the image URL. Use the output of an earlier step or a custom value. The URL must be public. 5. Write the analysis prompt. Be specific. Example: “Read the total amount in this receipt. Give only the number.” 6. Keep Detail Level at Auto. Select High for small text and dense documents. 7. Save the action. The output is available in all steps after this action.
Bottom line: your customers send you images every day — this action reads them so your team doesn’t have to.
If the number a customer called can’t send SMS — like a call tracking number or an unverified toll-free number — your Missed Call Text Back follow-up used to just fail. Not anymore.
The system now automatically falls back to another eligible number in this order: the number originally called (if eligible), a number that’s messaged this contact successfully before, your location’s default outbound number, or another eligible number at the location. If none are available, the failed attempt still shows clearly in Conversations so you know what happened.
No setup required — this happens automatically for every account.
My Profile, Teams, and My Staff have all been rebuilt on our newer design system — faster, cleaner, and no longer stuck in an old iframe-based experience. As part of this, Teams (agency view) and My Staff (sub-account view) are now both simply called Users, one consistent name across the platform.
My Profile is now organized into clear sections — General, Security, Email and conversations, Calendar, and Notifications — instead of one long scrolling page. The Users experience keeps full functionality you already rely on: search, roles and permissions, ownership transfer, and call-forwarding defaults, all rebuilt with a cleaner interface.
The email signature editor also got a refresh, with a modern rich-text toolbar, source view, and better rendering so signatures actually display correctly once sent.
Bottom line: same functionality you’re used to, in a faster, cleaner, easier-to-navigate interface — including full support for your account’s Custom CSS and JavaScript, which wasn’t possible in the old iframe-based pages.
Conversation AI agents can now run on eight new OpenAI models, ranging from top-tier accuracy options to lighter, faster models at a fraction of the cost.
In testing on real production conversations, every new model matched or beat the previous baseline — with the top-tier options reaching over 90% accuracy, and some lower-cost models landing 25–89% cheaper while still improving on accuracy.
How to enable: 1. Go to Conversation AI → Edit your agent 2. Select your preferred model from the list. 3. Upgrade a secondary model if needed. 4. Click Save, then test your agent
Bottom line: pick a premium model where accuracy matters most, or a lighter one for high-volume agents — the choice is yours.
Not every business wants their AI responding at 3am. You can now set specific working days and hours for each Conversation AI agent, choose which channels follow that schedule, and decide exactly what happens when someone messages after hours.
You can run the schedule on contact timezone or business timezone, let an already-active conversation continue past closing time, send a custom off-hours auto-reply, notify your team about after-hours messages, tag after-hours contacts automatically, and send a follow-up when you reopen if no one’s responded yet.
How to Enable It: 1. Open Settings -> Labs -> Search for “Working Hours” -> Enable the flag 2. Open AI Agents → Conversation AI 3. Select your agent and go to Deploy → Working Hours 4. Enable Active Hours 5. Select the channels that should follow the schedule 6. Choose Contact Timezone or Business Timezone 7. Configure your working days and hours 8. Optionally configure: Continue ongoing conversations, Off-Hours Auto-Reply, Team notifications, Contact tags, Follow up when business hours resume 9. Click Save
Good to know: this is optional — if you don’t enable it, your agent keeps responding 24/7 exactly as it does today.
Your agents just got five new connectors to work with:
Managed Agents now have a proper off switch. Unpublish takes an agent out of production in one click — its triggers stop firing, but every setting, instruction, and configuration you built stays exactly as you left it.
You can keep editing and testing the agent in the builder the whole time it’s unpublished, then republish whenever it’s ready to go live again — triggers and all, restored exactly as configured. Anything already in progress when you unpublish gets marked “skipped” with a clear reason, so your activity history stays accurate.
Why it matters: pause a seasonal or campaign agent without losing your setup, or quietly refine a production agent offline before bringing it back live.
How It Works: 1. To unpublish: Open a published agent and choose Unpublish. Its triggers rest while its production chat stays active (you can chat with a managed agent even in draft mode). 2. While unpublished: Keep editing in the builder and run the test panel as usual — the draft is fully live for you. 3. To bring it back: Publish again. Triggers reactivate with the configuration they had, with similar behaviour as you left them.
If a contact only has a WhatsApp username on file (no phone number), you can now message them right from the mobile app — something that wasn’t possible before.
A phone number and its matching username now merge into a single row instead of showing up twice, and the 24-hour messaging window now applies correctly to whichever destination you’re actually using. Mobile now matches how the web app already handles usernames.
How to Use It: 1. Open a conversation or start a new message in the mobile app. 2. Tap the To selector — WhatsApp usernames appear as @username rows. 3. Pick the username and send. Templates and messaging windows work just like they do on web.
Mobile Conversations now matches the web experience for AI-suggested replies. Instead of picking between several short suggestions, you get one clear, ready-to-send reply for each eligible incoming message.
Tap Send to use it as-is, tap Edit to tweak it in the composer first, or dismiss it and bring it back later with Show AI suggestion.
How to Get Started: 1. On web, go to Settings → Conversation AI and set your AI Employee to Suggestive mode. 2. Ensure Conversation AI is enabled for the channels you support. 3. Open Conversations on mobile: when an eligible inbound message arrives, the AI suggested reply card appears above the composer. 4. Tap Send to deliver the suggestion as-is, or tap Edit to load it into the composer, make changes, and send.
Good to know: this works when your AI Employee is set to Suggestive Mode, and isn’t yet available for email suggestions or group/internal chats.
Connect your WhatsApp number once, and from then on any message you send it reaches Ask AI — the exact same copilot you use in the app, just a text away.
You can do everything you’d normally do with Ask AI: look up or update a contact, check your calendar, draft a blog post, generate an email template, create images, videos, or audio, set up recurring tasks like a weekly lead summary, run a custom skill you’ve taught it, or ask a support question and get routed to the right answer — all from a WhatsApp thread.
Every exchange also shows up in a dedicated Dispatch tab in the app, fully searchable, and your connection is private to you — no one else can see or manage it.
How to Connect Your WhatsApp Number: 1. Open the Dispatch tab (or Customize → Dispatch) in the app and tap Connect WhatsApp. 2. Scan the QR code with your phone and send the first message to the number shown — you’ll get a verification code back on WhatsApp. 3. Enter the code in the app to complete verification — you’re connected. 4. Message that number anytime to chat with your copilot. Past conversations are always available in the Dispatch tab.
What’s next: Slack and Instagram are coming soon as additional Dispatch channels.
Starting a new question used to cancel whatever Ask AI was already working on. Not anymore — every chat now runs independently in the background, so you can fire off a big question, start a new one immediately, and come back to both once they’re ready.
Chat history shows a spinner for chats still working and a dot for ones that finished while you were away, and each chat keeps its own draft, attached files, and model choice.
How to Get Started: 1. Open Ask AI from your Apps launcher. 2. Send a question, then tap Chat history → New Chat and send another. Both keep running. 3. Open Chat history any time. Spinners show live chats; dots show finished ones waiting for you.
Courses can now be priced and sold directly, without setting up a separate offer and linking it back. Every course now has a Pricing tab where you choose Free, One-time, or Recurring pricing and get a ready-to-share checkout link the moment you save.
You can also set when access begins, limit access to a set number of days, and enable coupons for course-specific promotions. New Course Sign-up and Course Access emails, revenue analytics, and workflow automations to grant or revoke access round out the update.
How to Use: 1. Create a course and open the Pricing tab. 2. Choose Free, One-time, or Recurring. 3. Enter the price and billing details, if applicable. 4. Save to generate the checkout link. 5. Optionally configure access controls and coupons under Advanced. 6. Share the checkout link or use course access actions in Automations.
Good to know: this is rolling out gradually to a portion of newly created sub-accounts while we measure impact, so it may not be visible in every account yet.
You can now transfer calls from the mobile app to staff members even if they don’t have a personal phone number on file — routing the call to whichever device they actually have set up (Web App, Mobile App, Phone, or SIP desk phone).
Works for both warm and blind transfers, and for both outbound and inbound calls. The staff member receiving the call can see it was transferred, and who transferred it.
How to use this: 1. Make a regular call from the mobile app (outbound or inbound) 2. In the active call, tap the transfer call icon at the top right corner 3. Go to the Staff Members tab and choose the staff member you want to transfer the call to. Device types they can receive the transferred call on are shown next to their name 4. Tap Call and Hold for a warm transfer, or Transfer and End for a blind transfer 5. They tap Answer 6. The warm transfer can be executed as usual
Good to know: re-transferring an already-transferred call isn’t supported yet, and staff with no device configured will appear disabled in the transfer list.
FGF Notifications
