GoHighLevel AI Employee: what each AI agent does, what it costs, and where it stops

What are GoHighLevel's AI agents, and what are they called today?

GoHighLevel's AI agents are a set of built-in tools inside HighLevel, mainly Conversation AI for text channels and Voice AI for phone calls, sold together under the name AI Employee. Feature names change often, so this guide states what HighLevel's own help articles said when we checked them on September 24, 2026.

The naming is the first thing that trips people up. HighLevel's help center describes "AI Employee" as the plan and access structure that covers several products: Conversation AI, Voice AI, Reviews AI, Content AI, Ask AI and AI Studio. Older tutorials, and some third-party guides, still use names or prices from earlier releases, which is one reason two articles on this topic can disagree.

Here is the current map, taken from HighLevel's "AI Tools in HighLevel" article. We quote HighLevel's own one-line descriptions where they are short enough to be useful.

This guide concentrates on the two that answer your leads: Conversation AI and Voice AI. Those are the ones people search for, and the ones where a wrong setup costs you a real customer.

Source, at the time of writing: HighLevel Support Portal, AI Tools in HighLevel, checked September 24, 2026.

What does Conversation AI do?

Conversation AI is HighLevel's text-based agent. It replies to inbound messages, answers questions from a knowledge base you supply, collects contact details, books appointments, and can trigger workflows, across SMS, email, Facebook, Instagram, WhatsApp, the chat widget and live chat.

The channel list above comes from HighLevel's bot setup article. In practice the same bot logic can sit behind several of those channels, which is the useful part: a person who texts you at 9 p.m. and a person who uses the website widget at 9 a.m. get the same answers and land in the same contact record.

A bot runs in one of three modes, per HighLevel's documentation:

• Off: the bot is inactive.
• Suggestive: the bot drafts a reply and a team member reviews and sends it.
• Auto-Pilot: the bot sends replies on its own.

Auto-Pilot has a message ceiling. HighLevel's article says a bot can send between 1 and 100 messages per conversation, with a default of 100. Only bot-generated messages count toward it. Once the limit is reached, the bot goes dormant in that thread until someone reactivates it. Human replies do not count, and a human can step in at any point.

We start almost every bot in Suggestive mode. Reading twenty or thirty drafted replies tells you more about the gaps in your knowledge base than any test script does, and nothing wrong reaches a customer while you learn. Switching to Auto-Pilot is a one-setting change once the drafts stop needing edits.

What does Voice AI do?

Voice AI answers and places phone calls. It can hold a spoken conversation, look things up in a knowledge base, book appointments, send a text, update a contact field and trigger a workflow. It can transfer the call to a person, and it can also run as a voice widget on a website.

HighLevel's Voice AI overview says agents can answer calls during business hours, after hours, weekends and holidays, and lists multi-language support without naming the languages. The available actions in the create-agent article are Call Transfer, Trigger Workflow, Send SMS, Update Contact Field, Appointment Booking, Custom Action (a webhook) and Agent Transfer.

Outbound calling runs from a workflow: you add the Voice AI outbound call action, pick an agent and a From number, and the workflow schedules the call. The same agent can handle inbound and outbound if its prompt covers both.

Two documented details matter more than they look. First, Call Transfer does not run during a Web Call test, so you have to test a transfer with a real phone call. Second, agents need an eligible number to take live calls. HighLevel's create-agent article lists an LC Phone number, a Twilio number or a Number Pool, and the overview notes a number already tied to another function, such as an IVR, cannot also host the agent. Check your own number's eligibility in the account before you plan a launch date around it.

Sources, at the time of writing: HighLevel Support Portal, Overview of Voice AI Agents and Create, Test, and Deploy a Voice AI Agent, checked September 24, 2026.

What are Agent Studio, Managed Agents, and the other AI tools?

Agent Studio is HighLevel's visual builder for custom, flow-based agents, and Managed Agents lets you describe an autonomous CRM agent in plain language. They are separate from Conversation AI and Voice AI, and Agent Studio is billed pay-per-use even when you pay for an AI Employee plan.

Conversation AI is a bot that waits for a message and replies. Agent Studio and Managed Agents are for tasks that involve several steps and several tools, such as reading CRM data, deciding what to do and updating records. HighLevel's article lists Managed Agents with 100 runs per month on Growth and 1,000 on Unlimited.

Most small businesses do not need either on day one. If the job is "answer the phone and book the appointment" or "reply to web chat and qualify the lead", Conversation AI and Voice AI cover it. Agent Studio earns its cost when the logic is unusual enough that a standard bot prompt cannot express it, and that is a design conversation, not a checkbox.

Reviews AI, Content AI, Workflow AI and Ask AI are useful but sit outside the lead-response question this guide is about. We mention them in the pricing section because they share the same plans.

How do you set up Conversation AI?

Go to AI Agents, then Conversation AI, then Create Bot, choose a setup method, configure the bot's goals and prompts, test it, and set its status to Suggestive or Auto-Pilot. HighLevel offers a guided form, a prompt-based bot and a flow-based builder.

The three methods, from HighLevel's setup article, suit different jobs. The guided form suits straightforward lead capture and booking. The prompt-based bot gives you more control if you are comfortable writing detailed instructions. The flow-based builder is for advanced logic with multi-step paths. Picking a bot template only changes the starting prompts, per the same article.

The order we follow, in build practice:

1. Write the knowledge base first. Services, hours, service area, policies, what may be quoted and what must go to a person. The prompt is easier to write once the facts exist.
2. Pick one job. Answer web chat and book, or reply to missed-call texts. Not both, not yet.
3. Create the bot with the method that fits, and set its goals: collect name, phone and email, then offer a booking.
4. Connect the calendar and confirm the bot books into the right one, in the right time zone.
5. Set the mode to Suggestive and review drafts for a week or two.
6. Add stop rules. Decide what makes the bot go quiet: a reply from a human, a booking, a keyword. The Auto-Pilot article notes a workflow action can reactivate a sleeping bot on a time, keyword or tag trigger.
7. Move to Auto-Pilot for the channels you trust, and keep reading transcripts weekly.

Source, at the time of writing: HighLevel Support Portal, How to Create and Set Up a Conversation AI Bot (modified September 10, 2026) and Auto-Pilot Mode in Conversation AI (modified July 28, 2026).

How do you set up Voice AI?

Go to AI Agents, then Voice AI, then Agent List, and click Create Agent. Configure the voice, model, language, prompt and welcome message, add a knowledge base and actions, and test by web call and by phone call. Then deploy the agent to a phone number and set call routing and working hours.

HighLevel's create-agent article gives the sequence below. The creation options are Browse Marketplace, Create Custom Agent and Flow Based Builder.

1. Open the agent list and choose Create Agent.
2. Choose the creation method: a marketplace agent, a custom agent, or the flow builder.
3. Configure the core agent: name, voice, model, language, prompt and welcome message.
4. Connect the knowledge base (optional, but we treat it as required for anything that answers questions).
5. Add actions: transfer, booking, SMS, contact field updates, workflows.
6. Review call and post-call settings.
7. Test with a Web Call, which needs no phone number, then with a Phone Call to exercise the real telephony path.
8. Deploy to an eligible number or Number Pool.
9. Set call routing and working hours.

Two habits save the most time. Test the transfer to a human with a real call, because the web test skips it. And soft-launch: route only after-hours calls to the agent for the first week, so a person still answers during the day while you listen to real transcripts. Callers say things no test script anticipates.

For outbound, add the Voice AI outbound call action inside a workflow. Do that after inbound is behaving, and only once the consent question in the compliance section below is answered.

Source, at the time of writing: HighLevel Support Portal, Create, Test, and Deploy a Voice AI Agent (modified September 18, 2026) and Voice AI Outbound Calling.

How does HighLevel charge for AI agents?

HighLevel bills AI usage mainly through a usage wallet, per unit consumed. Two optional per-location subscriptions sit on top, at the time of writing: AI Employee Growth at $50 a month and AI Employee Unlimited at $97 a month. Phone-system charges are separate on every plan.

That is the short version of a model that has changed more than once. Many articles still quote a flat per-message price for Conversation AI. HighLevel's current pricing article bills Conversation AI on model token usage instead, so treat any flat per-message figure you find in an older tutorial as out of date until you confirm it against HighLevel's page.

The figures below come from HighLevel's "AI Products Pricing" and "AI Employee" articles, checked September 24, 2026. They are HighLevel's numbers, not ours, and they will change.

A worked example using only those published rates, for illustration and not a quote: a four-minute Voice AI call using the Voice Engine and a $0.015 per minute text-to-speech voice comes to $0.24 in engine and voice charges ($0.06 per minute), before language-model tokens and before the phone-system charge for the call. Premium voices cost more per minute, and HighLevel lists a range of speech and model options with different rates, so your real number depends on the choices in your agent.

The practical question is where the subscription breaks even. If you expect a few hundred agent replies and under 100 call minutes a month in one location, Growth's allowances may cover you. Heavier voice use is where the Unlimited plan is worth pricing against pay-per-use, remembering that "unlimited" is bound by fair use. For the platform subscription itself, see our GoHighLevel pricing breakdown.

Sources, at the time of writing: HighLevel Support Portal, HighLevel AI Products Pricing and AI Employee (modified September 14, 2026). Confirm current rates there before you budget.

What are the limits of HighLevel's native AI agents?

The documented limits are message and call ceilings, number and country restrictions, and fair-use terms on "unlimited" plans. The undocumented limit is quality: an agent only knows what you give it, and no plan setting fixes a thin knowledge base.

Sources, at the time of writing: HighLevel's Auto-Pilot, Voice AI overview, create-agent and outbound compliance checks articles (compliance article modified September 13, 2026), checked September 24, 2026.

Limits a help article cannot list are the ones we see in practice. A stale knowledge base produces confident wrong answers. A voice agent with a pause before each reply feels unnatural. A caller with a noisy line or a heavy accent gets misheard, so key details such as phone numbers and dates should be read back. An agent with no clean handoff to a person becomes a frustration machine, which is why the handoff gets tested first.

What do the compliance rules mean for AI calling and texting?

In the US, calls using AI-generated voices count as "artificial" voices under the TCPA, so they generally need the called party's prior express consent. HighLevel checks calling hours, country and volume, but says contact consent and opt-outs remain your responsibility. This is not legal advice.

The FCC confirmed in a February 2024 declaratory ruling that calls made with AI technologies that generate human voices fall under the TCPA's rules on artificial or prerecorded voices, which means prior express consent is required (FCC announcement).

HighLevel's own compliance article is direct about the split. The platform enforces calling hours, same-country calling, rate limits and caps, and eligibility checks. It does not validate that a given contact agreed to be called. Its article says plainly that accepting the platform's outbound calling terms does not establish consent for any individual contact. Honoring opt-outs and Do Not Call preferences, presenting any AI disclosure the law requires, and keeping proof of consent are on you.

The safe pattern is the one we use in builds: inbound calls first, and outbound only to people who asked to hear from you, such as someone who submitted a form with consent wording or a customer being reminded of an appointment. Texting has its own registration duties in the US (A2P 10DLC), covered in our GoHighLevel setup guide. Rules differ by state, industry and country and they change, so have your own counsel confirm what applies to you.

When are HighLevel's native AI agents the wrong choice?

They are the wrong choice when your account is not ready for them, when the conversations are high-stakes or heavily regulated, when the call flow needs logic beyond a prompt, or when nobody will own the knowledge base. An AI agent on top of a messy CRM automates the mess.

• Your account is disorganized. If duplicate contacts, dead pipelines and overlapping workflows are the daily reality, an agent will book into the wrong stage and text the wrong person. Fix that first, using our cleanup checklist.
• The conversation is emotional or high-stakes. Complaints, medical emergencies and bereavement belong with a person. A bot can route them, not handle them.
• Regulated data is involved. Health, financial and children's data carry extra obligations, and no tool is compliant by default. Get counsel involved before you put an agent in front of it.
• You want mass cold outbound. Consent rules and HighLevel's own caps make this the highest-risk use.
• Nobody owns the content. If no one will update hours, prices and policies, the agent will confidently quote last year's.
• The volume is tiny. If you get a handful of enquiries a week, a good contact form and a fast human reply may cost less than the setup and tuning time.
• The logic is unusual. Custom pricing rules, multi-system lookups and unusual routing may need Agent Studio or a workflow tool such as n8n, not a prompt.

None of this makes the native tools weak. It means they are one part of a system, and the parts around them decide the result.

Where do Autoesta's AI calling and chat agent services fit?

HighLevel supplies the tools; Autoesta designs, connects, tests and maintains the agent around your pipeline. Our AI calling agent service covers voice, and our AI chat agent service covers text and web chat, both built inside GoHighLevel.

You can build a working agent yourself from the steps above. What a service adds is the work HighLevel's articles leave to you. That means the call flow and guardrails, the knowledge base, contact matching so callers do not create duplicates, outcomes written back to the right pipeline stage, consent capture and opt-out handling, and a tested human handoff. Where data has to move outside GoHighLevel, we use n8n or Zapier.

Our two service pages go into the design detail, so we will not repeat it here. What is worth stating is how it is scoped. At the time of writing, a typical small automation build of two to three automations starts at about $1,000, a full GoHighLevel setup runs about $3,000 to $5,000, the audit is free, cleanup starts from $200, and monthly support starts at $200. AI agents are usually scoped as part of a larger build or added to an existing system, and builds include six months of maintenance. HighLevel's usage charges from the pricing section are separate and paid to HighLevel. Final scope and price are set on a call.

If you are not sure whether you need us at all, the honest test is the wrong-choice list above. A single-location business with a clean account and a modest call volume can often run the native tools alone. A multi-location business, a regulated practice, or anyone who has already had an agent give wrong answers will usually save time by having it designed. The related AI automation services page shows where agents sit next to other automation. For the calendar side, see appointment booking automation, and for how agents fit into a full account, our GoHighLevel setup service and GoHighLevel expert page.

Alpit Patel, a HighLevel Certified Admin, reviews the content Autoesta publishes, and the team has completed 320+ projects in 15+ countries. The dental case study shows a patient-intake system we built inside GoHighLevel, as reported by Autoesta.

Questions people ask about GoHighLevel AI agents

Is Conversation AI free in GoHighLevel?

No. At the time of writing, HighLevel bills Conversation AI by token usage on pay-per-use, or through the Growth plan (1,000 agent responses a month) or the Unlimited plan, per HighLevel's AI Employee article. Confirm current terms on HighLevel's pricing pages.

What is the difference between Conversation AI and Voice AI?

Conversation AI handles text channels such as SMS, chat and social messaging. Voice AI handles phone calls and the voice widget. Both use a knowledge base and can book appointments, and they share the AI Employee plans.

Can I use Voice AI with my existing phone number?

Possibly. HighLevel says the agent needs an eligible LC Phone number, Twilio number or Number Pool, and a number already used for something like an IVR cannot host it. Check eligibility in your account first.

Can the AI transfer a call to a person?

Yes, Call Transfer is a documented Voice AI action. It does not run in a Web Call test, so test it by phone.

Do I need consent before an AI agent calls someone?

In the US, generally yes for outbound calls with AI-generated voices, per the FCC's 2024 ruling, and HighLevel leaves consent to you. This is not legal advice; ask your counsel.

Which plan should I pick?

Start on pay-per-use or Growth in one location, measure your real responses and minutes for a month, then compare with Unlimited. Do not buy for a volume you have not seen yet.

Should I build it myself or hire someone?

Build it yourself if your account is clean, the job is simple and a mistake costs you an afternoon. Hire when other people depend on the account daily. You can contact us or book the free strategy call below.