Your workflows can now watch video. The AI Analyze Video action is available under AI Actions. Give it a video URL and a prompt. Google Gemini watches the video, listens to the audio, and returns a text answer for the actions that follow.

This is a premium action. Each execution has an additional charge.

Video analysis is billed per execution based on the tokens consumed.

Your actual cost per analysis depends on how you configure the action. The main factors are:

The video URL must be public. Page links from Vimeo, TikTok, and similar sites are blocked by those sites, so use a YouTube link or a direct video file link instead.

A video used to be a dead end in a workflow. You could send a link, but nothing could read it. Now a workflow can watch the video and act on what it sees. No manual review, no note taking, no rewatching.

Save the summary to a custom field, send it in an email or SMS, pass it to a webhook, or feed it to another AI action. Pair it with AI Video Generation to check a generated video before it reaches a customer.

Booking and managing services just got easier! Your clients can now — without needing separate booking links or switching between platforms.

With meetings and services available in one place, your clients get a smoother, more convenient booking experience.

Previously, clients had to rely on separate booking links to schedule services, creating unnecessary friction in their experience.

Now, the , making it easier for clients to engage with your business.

That's it! Your clients can now discover and book your enabled services directly from their Client Portal.

Service Booking is now available in the .

AI Builder now tests what it builds. Describe a workflow, let AI Builder build it, then click Test workflow in the same chat. The test runs for real, every step reports back on the canvas, and AI Builder tells you in plain words what worked.

A test runs the real steps. Tags are added, notes are written, and messages are sent. Test with your own contact when you do not want to reach a customer.

Building with AI is fast. Knowing it works used to be the slow part. You had to leave the chat, find the test, and dig through logs to see what happened.

Now build and test live in one place. Describe it, build it, test it, read the result. No guessing, no digging through Execution Logs, no hoping it works in production.

Your AI Agent can now search the internet. Turn on Web search in the new Built-in tools section, and the agent looks up the facts it needs in the middle of a run. You can also point it to exact pages: paste their URLs in the Instructions, and the agent reads them when it needs them.

AI image generation and AI video generation are already live as separate actions. Both are coming to the AI Agent Action as built-in tools very soon.

Web search is a premium tool. It is billed per search, on top of the AI Agent Action.

A heating and cooling company follows up on every quote request. Some leads ask about rebates, and rebate programs change often.

The AI Agent Action is set up like this:

A lead asks about heat pump rebates. The agent checks the rebate page you gave it before it replies. Another lead asks a question that page does not cover, so the agent searches the web for the answer. No stale facts, no guesswork, no manual lookups.

Your agent is no longer limited to what it already knows. Prices, programs, and public details change all the time. Web search lets the agent check them at the moment it needs them.

You stay in control of the sources. Name the pages you trust in the Instructions, and tell the agent when to use them.

You pay only when the agent searches. A run that needs no search costs nothing extra.

Your workflows can now make video. The AI Video Generation action is available under AI Actions. Give it a prompt. Google's Veo 3.1 creates a short video, stores it at a public URL, and returns that URL for the actions that follow.

This is a premium action. Each execution has an additional charge.

Video generation is billed per execution, based on the length of the video and the settings you choose.

Note: Treat any cost figures as a planning estimate, not a fixed rate.

Your actual cost per video depends on how you configure the action. The main factors are:

The generated video is stored at a public URL. Anyone with the link can open it without a login. Keep this in mind when a prompt pulls in customer data.

Until now, a workflow could send a video you uploaded in advance, but never make one. Now a workflow can shoot the clip itself. No camera, no editor, no studio.

Use the video URL in the actions that come after. Send it in an email or SMS, save it to a custom field, or pass it to a webhook. Because the prompt can come from your data, one workflow can produce a different video for every contact who passes through it.

The calendar view now gives you more control over how it looks and what you can do with it. A new panel brings your display preferences together in one place, lets you take the exact view you're looking at out of the product, and filters and event colors are easier to work with. These are display and workflow improvements only. Availability, booking rules, and what customers see on your booking page don't change.

Open the more options (3 dots menu ⋮) menu next to Manage view in the calendar toolbar and select Calendar view settings to open a single panel for your display preferences. Your choices are saved for the next time you open the calendar.

The "View in timezone" setting moved out of Meetings Preferences and the Services/Rentals Global Settings. Any timezone you set before carries over automatically.

Admins get a Staff colours tab to recolor how appointments for the staff members look.

Choose whether the event tile is filled with the calendar/service color or the assigned staff color. The other color shows on the tile's left strip.

The left color strip is wider, so you can tell which staff member or calendar an event belongs to at a glance.

From the same more options (⋮) menu, choose Export to PDF or Export to CSV to export the current view with the same filters and date range.

Your AI Agent can now learn skills. A skill is a reusable procedure that the agent loads only when a run needs it. Write it once, keep it in your library, and attach it to any AI Agent Action in Workflows.

Every account includes four skills tagged Built-in. Attach, edit, or delete them like any other skill.

A heating and cooling company follows up on every quote request with an email and a text. A few leads report emergencies, and those need a different answer.

The AI Agent Action is set up like this:

The custom skill fires on this line: "Use when a lead reports no heat, no cooling, a gas smell, or a water leak, or says the request is urgent." Its instructions say to offer the next same day slot, to never quote a price, and, for a gas smell, to send safety steps first and never book a visit.

A routine quote request gets a friendly email and a short text. A lead who writes "no heat since last night" gets a same day slot. A lead who reports a gas smell gets safety steps first. Same agent, same Instructions, the right behavior every time.

Your Instructions stay short. Rules that apply only some of the time move into skills, so each action describes its own job and nothing else.

Write once, use everywhere. One skill serves every AI Agent in your account. Fix it once and every agent that uses it gets the fix.

The agent stays focused. Situational rules stay out of the way until the situation shows up, and several skills can work together in a single run.

You can now send one contact down every branch it matches. The Router action sits in the same family as If/Else, with one big difference. If/Else stops at the first branch that matches. Router checks every branch and runs each one that matches, one after another, left to right.

A home services business asks one question on its form: "What service do you need?" The lead can pick more than one: HVAC, Plumbing, Electrical, or Carpentry.

The Router action has four branches:

A lead picks HVAC, Plumbing, and Electrical. With If/Else, only the HVAC branch would run. With Router, the test output shows Branch 1, Branch 2, and Branch 3 all passed, each with its reason. Every team gets the lead, and each branch runs its own next steps: a note, an internal notification, a tag, a task.

Real leads rarely want just one thing. Router lets one workflow act on every match instead of only the first one.

One contact, every matching branch, in the order you choose.

You can now repeat a set of actions for every item in a list, right inside a workflow. The Loop action runs the actions inside it once for each item, or once for each number in a range. One trigger, many records, no code.

The goal: a service request form sends one webhook per customer, with a list of services inside it. You want one opportunity, one task and one note for every service.

One test run. The loop runs three times and creates three opportunities, three tasks and three notes. The Test tab shows 3 items, 3 succeeded, 0 failed.

A single trigger often carries many records: the services in a request, the products in an order. Earlier, handling each one meant code or a long workaround. Now you build the steps once and the Loop runs them for every item.

Range mode covers the other case: run a set of actions a fixed number of times, 5 or 10, with no list at all. Parallel mode keeps loops fast when the actions call external systems. Build once, run for every item.

You can now work with your workflows from your AI assistant. Workflows in the LeadConnector MCP server, now in private beta, let Claude, ChatGPT, Cursor and other AI tools find workflows, explain what failed and for whom, count enrollments, and build and publish workflows for you.

With workflows in the LeadConnector MCP server, you can:

Ask which steps in New Lead Nurture failed this week → the assistant reads the logs → it lists each failed step with the contact and the error.

Ask how many contacts entered Webinar Follow-up each week this month → the assistant counts enrollments by week → you get one number per week.

Describe a workflow that tags new contacts webinar-lead and sends a welcome SMS → the assistant checks it for errors → it saves a draft for you to review in the builder.

Support doc Here