Help Scout AI Answers: What It Can and Can't Do (2026)

TL;DR:

• Help Scout AI Answers is an OpenAI-powered chatbot that lives in the Beacon widget and answers customer questions from your Docs.

• It is genuinely good at deflecting repetitive FAQs, and Help Scout publishes an average resolution rate around 73 percent.

• It also has four limits worth knowing before you turn it on. It cannot take actions.

• It learns only from your Docs, not your past tickets. Tuning it is a manual, reactive process.

• It bills $0.75 per resolution, with a spending cap that switches the AI off once you hit it.

What AI Answers actually is

AI Answers is the customer-facing part of Help Scout's AI. It sits inside Beacon, Help Scout's chat widget. When a customer asks a question, it reads your Help Scout Docs and any public URLs you have added as sources, then writes a conversational answer on the spot. If it cannot help, it hands the customer to a human in the same window. It runs on OpenAI, and Help Scout has stated your data is not used to train OpenAI's models.

That is the whole job. Answer common questions from your knowledge base, around the clock, and pass the rest to your team. For a lot of support volume, that is useful.

What it does well

Credit where it is due. For repetitive, FAQ-style questions, AI Answers works. Help Scout publishes an average resolution rate of about 73 percent across its customer base, with some teams reporting higher. It replies in over 50 languages, can match your brand voice, and gives customers a clean path to a human when it stalls. If your knowledge base is mature and most of your questions are simple and repeatable, it will take real load off your team.

The first hurdle: AI Answers can't take action

By design, AI Answers is built to answer questions from your knowledge base and escalate to a human when it cannot. It is read-only. It cannot check an order status, process a refund, reset a password, look up a subscription, tag a ticket, or route a conversation. Anything that requires touching a customer's account or another system still goes to a human.

This matters because a large share of real support is not "what is your return policy." It is "where is my order" and "please refund this." Those need an action, not a paragraph. AI Answers can describe how a refund works. It cannot issue one.

The second hurdle: it doesn't know your past tickets

AI Answers learns only from your Docs and the URLs you feed it. It does not learn from your past tickets or your agents' real answers. Every solution your team has already written, thousands of resolved conversations in your own voice, sits outside what the customer-facing AI can use.

The practical effect: the AI is only as good as the knowledge base you hand-build and hand-maintain. There is no auto-generation, no gap detection, and no staleness alert. When your product changes and someone forgets to update a Doc, the AI keeps confidently giving the old answer. And you pay for each of those answers. More on that next.

How the pricing works, and the cap that bites

AI Answers is a separate add-on. After a three-month unlimited trial, you pay $0.75 for every resolution, on top of your plan. Help Scout defines a resolution as a chat the AI handles without passing it to a human.

Two things to plan for. First, the cost is variable. A busy month with a few thousand deflected tickets is a bigger bill, which makes budgeting harder than a flat plan. Second, and easy to miss: if you set a spending cap to control that cost, AI Answers switches off the moment you hit the cap, and stays off until your next billing cycle. So the self-service your customers rely on can disappear in the middle of your busiest week, which is exactly when you needed it most.

Tuning happens by hand

One more thing before launch. Improving AI Answers is a manual, reactive loop. You watch real conversations, spot the answers it gets wrong, then write or edit a Doc and re-test. There is no system that flags knowledge gaps or suggests what to add. For a simple knowledge base, that is manageable. For anything nuanced, you are doing the QA yourself. Expect ongoing manual upkeep to keep answers accurate as your product changes.

So, is it worth turning on?

It depends on your support.

AI Answers is a fair fit if your knowledge base is already mature, most of your questions are simple FAQs, you are fine with read-only answers, and your volume keeps the per-resolution cost predictable. For a lot of small teams, that is the reality, and AI Answers will help.

You will hit the ceiling fast if your customers ask you to do things, not just tell them things. Or if your real support knowledge lives in past tickets rather than a polished Docs site. Or if your volume makes $0.75 per resolution and a hard spending cap a monthly gamble.

If that second description is why you are reading this, it is worth seeing how an AI-native alternative to Help Scout handles the same work. AI that can take actions, learns from your real support history, and bills on flat, predictable pricing. A model built for resolution, not just deflection.

Either way, go in with eyes open. AI Answers is a solid FAQ deflector. It is not, by Help Scout's own design, an AI Agent that resolves a ticket end to end. Know which one your team actually needs.