Home › Updates › HighLevel Updates: November 2026 Roundup

Monthly Roundup

HighLevel Updates: November 2026 Roundup

November has had two release days so far and the bigger one is the newer. On November 3 HighLevel put Field Services into private beta, a new industry module for businesses that do the work at the customer address. It is built around requests, jobs and visits rather than appointments, and it covers the request, the quote, the schedule, the dispatch, the invoice and the payment in one place, with route optimization, service areas, timesheets, a branded booking page, an AI receptionist and a mobile app for the crew. The same day Conversation AI switched its recommendation to GPT-6 Sol, deprecated GPT-5 and GPT-5 Mini and started winding down GPT-4.1. November 2 was Ask AI on mobile, with five features in version 4.27.0: Connectors for third party apps and custom MCP servers, Contacts AI with a records based summary card on every contact, a home screen widget, three iOS Action button shortcuts and a new chat shortcut on the app icon. The same day Communities replaced its single membership price with multiple pricing plans, started charging for individual private channels and gained an analytics dashboard, and Quizzes gained conditional logic with score based redirects.

Start the 30-Day Free Trial → Official HighLevel page ↗

30-day free trial through our link (the public site offers 14) · No contracts · Cancel anytime

Published 2026-11-03 · Updated 2026-11-04

The clearest signal in November so far is that HighLevel keeps moving downstream, from finding the work to doing it and billing for it. Field Services is the furthest it has gone. A job that spans several visits, a technician route that shows the drive time saved before you accept it, a timer on each visit and an invoice raised after the second of three visits are not marketing features, and none of them have an equivalent in Calendars. Communities moved the same direction on a smaller scale, replacing one membership price with a plan ladder and letting a single private channel be sold on its own. The Ask AI releases are a different kind of change. Four of the five shorten the distance to a question rather than improving the answer: a widget, an app icon shortcut, three Action button options and Siri. The fifth, Contacts AI, is the one that changes what Ask AI knows, and it does it carefully. The summary card is assembled from your own records rather than written by a model, so the last message and the next appointment are facts, and only the briefing on top of them is generated. The quiet release with the most immediate use is the Communities analytics dashboard, because the inactive member export is a list you can work through this week. The model switch is the one item this month with a deadline on it, because GPT-4.1 agents have to be moved before HighLevel retires the model.

November is still in progress and this page is updated as HighLevel ships. The October 2026 roundup is the closed record of last month.

Field Services turns HighLevel into a job management system

November 3 put Field Services into private beta, a new industry module for businesses that do their work at the customer address. HVAC, plumbing, electrical, cleaning and landscaping do not run on appointments. They run on a request, a quote, a visit or three, an invoice and a payment, and until now none of that had a home in HighLevel.

The unit of work is a job, not a booking. One customer, one address and as many visits as the work needs, holding services, products, pricing, checklists, notes, invoices and payments. Status moves itself from Unscheduled to Scheduled, In progress and Closed based on the visits and the billing
Requests and estimates come first. Customers send a request from the booking page, or the team creates one with contact, address, description, photos, services and products. Schedule a site visit before the work is agreed, then convert the request to an estimate or a job in one click
Find availability does the hard part. It suggests times from staff working hours, assigned services, service areas, existing bookings and drive time. Assign several technicians to one visit, and choose whether the customer sees an exact time or a window
Route optimization asks before it acts. You see the proposed route and the drive time saved, and apply it only if you agree. Lock any visit that has to keep its slot and let the rest be reordered, from each technician start location to their end location
Coverage is a rule, not a hope. Service areas by radius or postcode, with an optional travel fee and extra time before visits, plus a decision on customers outside the area: block the booking or allow it with review
Billing happens mid job. Raise an invoice straight from a job and bill as you go, so a three visit job can be invoiced after the first. A billing card tracks invoiced, paid and outstanding, and online payments work once a provider is connected
The schedule has three views. A list, a calendar in day, timeline, week or month, or a map, filtered by staff, service area and status, saved as your own smart lists and exported to PDF or CSV
The crew gets a real app. Requests, jobs, rescheduling, status updates, checklists, notes and timers from the phone, synced with the web, plus custom visit statuses such as On my way and Checked in, and timesheets with a timer on every visit
The front door is covered too. A branded booking page checks the customer address against your service areas, takes payment online, in person or on a card on file, and offers instant booking or booking requests. An AI receptionist answers questions on services, prices and coverage, and Google organic booking takes bookings from the Business Profile

Access is the catch. Accounts enrolled in the beta see Field Services with a Beta tag in the left navigation, and everyone else registers interest through the form on the HighLevel release note. Worth being clear about what this is: not a calendar with extra fields, but a second scheduling engine with its own services catalog, its own staff setup and its own booking page, which you can seed in bulk from a CSV or copy from your existing Calendars services. For an agency, the interesting part is not the module. It is that a trades client who was buying lead generation can now be sold the thing that runs the vans.

Conversation AI moved to GPT-6 Sol and started retiring GPT-4.1

November 3 made GPT-6 Sol the recommended model for prompt-based Conversation AI agents, deprecated GPT-5 and set the clock running on GPT-4.1. HighLevel tuned Conversation AI around the new model, so the accuracy it reports arrives without a change in cost per reply.

It is the default for anything new. GPT-6 Sol sits first in the Model menu with a Recommended tag, and prompt-based agents you create now start on it
Existing agents keep the model you picked. Nothing is switched under you, so an agent you tuned stays where it is until you move it
98.4% on the internal evaluation, up from 96.3%. That is HighLevel's own Conversation AI test set measured against GPT-4.1, at about the same cost per reply
Replies take 1.68 seconds typically. The slowest 5% come back in 2.19 seconds
GPT-5 and GPT-5 Mini are deprecated. Agents that used them already run on GPT-6 Sol, with reply generation up to 4.5 times faster, and nothing is needed from you
GPT-4.1 is being retired. OpenAI is deprecating it and HighLevel will follow, so every GPT-4.1 agent needs moving before that lands

The switch is in AI Agents, Conversation AI, then the Model menu on the Build tab. Test the agent in Test your agent before you save, because each model writes in its own way and a prompt tuned to GPT-4.1 phrasing can read differently on GPT-6 Sol. For an agency running AI agents across client accounts the work is a pass over every prompt-based agent to find the GPT-4.1 ones, not a one click upgrade.

Communities can sell a ladder instead of one price

On November 2 Communities replaced its single membership price with multiple pricing plans, and made individual private channels sellable on their own. Before this a community was free or paid at one price, so anyone who wanted tiers had to split members across separate groups.

Plans live in Settings then Subscriptions. Each one gets a name, description, price and benefits, so Basic, Pro and VIP sit inside one community. Activate, deactivate, reorder, flag one as Most Popular and read member counts from the same screen
Pricing is flexible. Weekly, monthly, yearly or one time, across multiple currencies, with an optional free trial per plan
The Benefits tab bundles access. Private channels, paid events and paid or private courses can be included in a plan, with a live preview of what members will see, which removes the manual access grant
Members switch themselves. They compare plans, subscribe and move between them at any time, and the existing subscription is kept until the new one is confirmed. The manage subscription page shows the active plan, the payment method and the next billing date
Private channels no longer have to hide. Visible mode puts the channel in the sidebar with a lock icon, so non members see the name and description but no posts. Hidden mode is still there for anything genuinely private
A channel can carry its own price. Free, recurring monthly or annual, or one time for permanent access, in multiple currencies, with a test mode toggle to check the payment flow before real money moves
Payment is self serve on web and mobile. A member taps a visible paid channel, sees the description, member count and managers, picks a billing cycle and pays. No manual adds and no back and forth
One gap remains. Free plan configuration is still to come, so you cannot yet define what the free tier includes alongside the paid plans

The combination is what matters. A plan ladder sets the price of belonging and a paid channel sets the price of one room, which means a community can hold both the ten dollar tier and the one off masterclass without becoming three communities. Anyone already running paid content through stores and offers should check for overlap before building the same ladder twice, because membership billing and product billing are now two routes to the same money.

Communities got the analytics dashboard it was missing

A dedicated Analytics section landed in Communities on November 2, under Settings then Analytics. Revenue, membership, traffic and engagement were previously scattered or simply not recorded.

Four headline cards. Total earnings as net community attributed revenue after refunds and chargebacks in the sub-account currency, total members with new joins in the period shown separately, inactive members with an active against inactive percentage bar, and email invites still outstanding
Comparisons are honest. Every metric shows the percentage change against the previous equal length period, and when the prior period was zero it says N/A rather than inventing a number
One date filter drives everything. Today, yesterday, 7, 30 or 90 days, 12 months or custom, defaulting to the last 30 days, applied to every card at once
Traffic sources are named. Direct, Instagram, YouTube, Facebook, LinkedIn and Other, with the top four by volume shown individually and a period delta on each row
Peak activity is a heatmap. Day by hour in the community timezone, darker for busier, with the best slot called out, which is the number to schedule posts, events and live sessions against
Engagement splits three ways. Posts, comments and reactions as a donut with counts and share of total, poll votes and event RSVPs included, next to a trend line for total interactions over the period
The inactive list exports. One CSV of active members with no tracked activity in the period, which is the only card here you can act on the same afternoon
Traffic is not backfilled. Page view capture is rolling out now and counts fill in going forward, so the traffic panel will look thin until a full period has passed. Visitors are counted by user ID, falling back to a hashed IP fingerprint, with bots excluded

HighLevel says a second version is in planning on the back of feedback from this one, so treat the current panels as the floor. The useful habit to start now is reading the inactive count as a churn warning rather than a vanity metric, since a member who stopped posting usually stops paying a month or two later. For communities feeding a wider funnel, the traffic source split is the first time the platform will tell you whether email or social is actually bringing people in.

Ask AI now reads a contact before you do

November 2 put an Ask AI icon at the top of every contact in the mobile app, with a summary of the relationship already on screen. Checking where things stand used to mean opening conversations, appointments, opportunities and tasks one tab at a time, usually while the client was already on the phone.

The card is built from your records, not written by AI. Last message with the channel and how long ago it went out, next appointment with its status including your own custom ones, biggest open deal with its pipeline, stage and days since it last moved, plus open tasks with a badge for overdue. Tap a row to open that record in its usual screen
Three buttons do the work. Prep me for next appointment appears when something is booked in the next 7 days. Summarize this contact is offered for every record. Dictate to add a note is always there
Dictation proposes, you approve. Speak the note and Ask AI turns it into a draft, plus a task when one is implied. Nothing is saved until you say so
Permissions hold. The card shows only the conversations, appointments, deals, deal values and tasks that user is allowed to see, and Ask AI receives the contact record ID rather than the record
It uses your own words. If the account calls contacts patients or clients, the app does too
The limits are worth knowing. Each visit starts a fresh chat, voice mode is not available inside a contact chat yet, and deal amounts come back as plain numbers with no currency symbol

This is the first screen to hand Ask AI one of its own records, and HighLevel says more will follow. The design choice worth copying is that the facts come from the CRM and only the briefing is generated, so a wrong summary is a wrong opinion rather than an invented appointment. For anyone running a field team, the dictated note is the line item that pays for itself, because the note that never gets typed is the one that costs you the follow up.

Connectors moved to the phone, custom MCP servers included

Ask AI can now be wired to outside apps from the mobile app, including a custom MCP server you add with a name and a URL. Connecting an app, fixing an expired login or switching a tool off used to need a computer.

Two tabs. Connected lists every connected app with its account count and how many tools are switched on. Browse searches the catalog and connects a new one in a few taps
Sign in happens on the app own page in a secure in-app browser for most apps. Some ask for credentials, some for a name and email, some need no account at all
You pick the tools. Tick the exact tools Ask AI may call per app, or use Select all. A connector needs a connected account and at least one tool ticked before it will turn on
Custom MCP servers are allowed. Add a server by name and URL, with client ID, client secret and a copyable redirect URL available. The URL has to start with https:// or http://
Web and mobile stay in sync, and the accounts you connect stay private to you. Other users in the sub-account do not see them
Two traps. Tapping Back without Save discards your changes with no warning, and deleting a custom MCP server also removes every account connected to it, permanently

Connectors are set up per sub-account, so at agency level you are asked to switch into one first. Skills, which exist on the web version, are not part of Customize on mobile yet, and some Connectors text still shows in English in other languages in 4.27.0. The reason this matters more than it reads: a phone that can authorize an AI assistant against your other systems needs the same care as a laptop, so the per-tool tick boxes are not a formality.

Three new ways to start an Ask AI chat

A home screen widget, three iOS Action button shortcuts and a New Ask AI Chat shortcut on the app icon all landed on November 2. Every one of them opens a blank chat rather than dropping you back into the last conversation.

The widget has three controls. A chat bar that opens a new chat with the keyboard up, a Voice button for hands-free mode and a Camera button that takes a photo, attaches it and waits for your question
It is off until you turn it on. The Ask AI Home Widget switch is per company in the service dashboard and defaults to off. The widget picks up your app language and brand colours in light and dark mode, white-label apps included
The iOS Action button takes New Chat, Dictation or Voice Mode. Dictation and Voice Mode open the mic as well. A voice or dictation session already running is never interrupted
Siri and Spotlight work without the hardware. The Action button needs an iPhone 15 Pro or later, while Siri and Spotlight go back to iOS 16.6
The app icon has a shortcut now. Long press the icon, use the Quick Actions card or search for New Ask AI Chat. If Ask AI is already open it switches to a new chat instead of stacking another screen
It only shows for the right people. The shortcuts and the widget appear only for accounts and users that have Ask AI

All five mobile releases need version 4.27.0 or later, and all of them are limited to users with Ask AI access. Agencies shipping their own white-label app get the widget in their own colours, which is the part clients notice. Worth remembering that the fresh chat every time is a deliberate trade: no context carried over, and no stale topic bleeding into the next answer.

Quizzes can redirect on score instead of showing a result

Conditional logic arrived in Quizzes on November 2, starting with Redirect to URL and four score based conditions. Until now a quiz had no conditional logic at all, so every respondent landed on the same results page.

Four new score fields. Highest Category, Lowest Category, Overall Score and Category Score, any of which can drive a redirect
The existing conditions come too. Everything Redirect to URL already supported in Forms and Surveys works here, so answers and scores can be mixed in one rule
The results page is now optional. Send a respondent to a recommendation page, a product page, a learning resource or a specific funnel step instead
Setup is in the builder. Open the quiz, open Conditional Logic, choose Redirect to URL, set the conditions, then set the destination URL
Three logic types are still to come. Display a custom message, disqualify a lead and show or hide fields

Score based routing is the first thing Quizzes can do that Forms and Surveys cannot, which changes what a quiz is for. An assessment that sends high scorers to an advanced offer and low scorers to a free resource is two funnels behind one asset, and the scoring does the qualifying. The gap left open is that disqualification is still on the roadmap, so a quiz cannot yet turn a respondent away, only send them somewhere else.

What rolled over from the end of October

November 1 had no releases, and the only entry dated October 31 was a template drop. The four release days that closed October are documented in full on their own page.

October 30 brought Hosted Numbers in Labs for SMS on a number a client already uses, usage limits for SaaS clients on Voice AI and Conversation AI, and call quality, every recording and Answered By in the mobile app
October 29 added the Router and Loop workflow actions, AI video generation and analysis, skills and web search in the AI Agent action, and service booking in the client portal
October 27 and 28 shipped outbound Conversation AI, Workflow Maps, workflows over MCP and a REST API, Client App Builder and Managed Agents with subagents

The detail on all of it, with the setup steps, sits in the October 2026 roundup. For the running list between monthly pages, there is the weekly roundup.

More HighLevel updates

Ready to try it yourself?

November 2026 HighLevel releases. November 3 put Field Services into private beta, a new industry module built around requests, jobs and visits for trades that work at the customer address, and made GPT-6 Sol the recommended Conversation AI model with GPT-5 deprecated and GPT-4.1 on the way out. November 2 shipped five Ask AI mobile features in version 4.27.0, multiple pricing plans and paid private channels in Communities, a community analytics dashboard and conditional logic in Quizzes. Updated through the month as HighLevel ships.

Try HighLevel Free for 30 Days →

Frequently Asked Questions

Can I use HighLevel Field Services yet?

Only if your account is in the beta. Field Services entered private beta on November 3, 2026. Accounts enrolled in the beta program see Field Services with a Beta tag in the left navigation, and everyone else registers interest through the form on the HighLevel release note. It is a separate industry module from Calendars, built around requests, jobs and visits for trades that work at the customer address, such as HVAC, plumbing, electrical, cleaning and landscaping. It covers capturing a request, quoting, scheduling, dispatch, route optimization, invoicing and collecting payment, with service areas, travel fees, checklists, timesheets, a branded booking page, an AI receptionist, Google organic booking and a mobile app for field staff. Services can be imported from a CSV or copied from existing Calendars services.

Which model do HighLevel Conversation AI agents use now?

GPT-6 Sol, since November 3, 2026. It is listed first in the Model menu on the Build tab and tagged Recommended, and prompt-based agents created after that date start on it. HighLevel reports a 98.4% pass rate on its own Conversation AI evaluation against 96.3% for GPT-4.1, at about the same cost per reply, with a typical generation time of 1.68 seconds. Agents that were on GPT-5 or GPT-5 Mini were moved across already. Agents still on GPT-4.1 keep working for now, but OpenAI is deprecating that model and HighLevel will retire it, so switch them and retest before the change is forced.

Can HighLevel Ask AI connect to other apps from the mobile app?

Yes, since November 2, 2026, in mobile app version 4.27.0 and later. Open Ask AI, tap the chat history icon at the top right, tap Customize, then tap Connectors. The Connected tab lists the apps already linked with their account count and how many tools are enabled, and the Browse tab searches the catalog. Most apps sign you in on their own page in a secure in-app browser, and you then tick the exact tools Ask AI is allowed to call. Custom MCP servers can be added by name and URL, with client ID, client secret and a copyable redirect URL available. Connectors are configured per sub-account, the accounts you connect are private to you, and apps connected on the web appear on mobile and the other way round.

What does Contacts AI show in the HighLevel mobile app?

Tapping the Ask AI icon at the top of a contact opens a summary card built from the account records rather than generated by AI. It shows the last message with its channel and age, the next appointment with its status and a plus N more count when others are booked, the biggest open deal with its pipeline, stage and days since it last moved, plus open tasks with a badge for overdue ones. Each row opens the underlying record. Below the card sit Prep me for next appointment, which appears when an appointment falls in the next 7 days, Summarize this contact and Dictate to add a note. A dictated note is proposed as a draft, with a task when one is implied, and nothing is saved until the user approves it. Released November 2, 2026 in version 4.27.0.

Can a HighLevel quiz redirect people based on their score?

Yes. Conditional logic arrived in Quizzes on November 2, 2026, with Redirect to URL as the first logic type. Four quiz specific fields can drive the rule: Highest Category, Lowest Category, Overall Score and Category Score. Every condition Redirect to URL already supported in Forms and Surveys also works, so answers and scores can be combined. Open the quiz builder, open Conditional Logic, select Redirect to URL, set the conditions, then set the destination URL. Respondents who match are sent there instead of the default quiz results page. Three further logic types are planned: display a custom message, disqualify a lead and show or hide fields.