AI Chatbot for Coaches and Consultants: Answer, Qualify and Book

A prospect reads your services page at 10 p.m. and has one question: do you work with people in their situation? Your contact form promises a reply within two business days, and you spend tomorrow in back-to-back sessions. By the time you answer, they have booked a call with someone who replied first, or simply lost the impulse.

This guide is for independent coaches, consultants and small advisory firms: career, executive, business and leadership coaches, fractional executives, and specialist consultancies where the first sale is a discovery call. It covers the questions prospects ask, what to put in the knowledge base, how to qualify without a form wall, how booking on a real calendar works, how the assistant can remember returning clients, when it should hand off to you, and the ethical limits that matter more in this work than in most.

The setup steps use PepoChat. The principles apply to any AI assistant you put on a practice website.

What can an AI chatbot do for a coaching or consulting business?

An AI support agent is an assistant on your website that answers from your own content, takes actions such as booking a meeting, and hands the conversation to a person when it should. For a practice, its work is narrower than in ecommerce or software support, and that is a good thing.

It does four jobs well. It answers the logistics questions that stall a decision: who you work with, how sessions run, what it costs, what happens first. It qualifies, learning enough in a couple of questions to know whether a discovery call makes sense. It books that call on your calendar while the prospect is still interested. And it hands off anything that needs you.

It does not coach. It cannot run a session, read a person's situation the way you do, or decide whether you are the right fit for someone in difficulty. Treat it as the front desk of your practice, not a junior coach.

What questions do coaching and consulting prospects ask?

Prospects ask about fit, price, format, availability and outcomes, roughly in that order. Most of those can be answered from pages you already have; a few should always reach you. The table maps the common ones to what the assistant should do.

Two patterns stand out. Most questions are answers, and the quality of those answers depends entirely on what you publish. And the two most sensitive rows, advice and distress, are refusals followed by a route to a human.

What should go in a coach's or consultant's knowledge base?

Everything a prospect would need to decide whether to book a call, written as plain statements. The knowledge base is the set of pages, documents and Q&A pairs the assistant searches before every reply; it answers from what it finds there and says so when it finds nothing.

• Service pages, one per offer, saying who it is for, what happens, how long it takes and what it costs or how the fee is set.
• A frequently asked questions page in plain sentences: "Sessions are 60 minutes on Zoom." "Most clients work with me for three to six months." PepoChat's best-practice guide makes the same point: clear statements produce better answers than marketing copy.
• Who you work with, and who you don't. This single section does more qualification work than any form.
• Your process: discovery call, proposal, first session, check-ins.
• Case studies and testimonials, with the context that makes them honest.
• Policies: cancellation, rescheduling, refunds, confidentiality, and how you handle notes.
• An about page with your training and credentials.

Program outlines and brochures can go in as PDFs, and a question your pages do not cover can be added as a Q&A pair in a minute. The guide to training an AI chatbot on your website and PDFs walks through the imports. On the Free plan a whole-site crawl counts as one of your five knowledge sources, and Q&A pairs count as none.

Keep client material out. The knowledge base is what the assistant may quote to anyone, so session notes, worksheets you made for a specific client, and private correspondence do not belong in it.

Be careful with case studies. The FTC's Endorsement Guides FAQ says that statements like "Results not typical" or "Individual results may vary" do not change how people read a specific result; an advertiser needs proof the result is typical, or must "clearly and conspicuously disclose the generally expected performance in the circumstances shown in the ad." An assistant that quotes your most dramatic success story to every prospect repeats that claim at scale, so publish case studies with their context and add a rule against promising outcomes.

How do you qualify coaching leads without a form wall?

Answer first, then ask one question at a time, inside the conversation. Lead qualification is learning enough about a prospect to know whether a discovery call is a good use of both your time. A form does it before helping, which is the wrong order.

Nielsen Norman Group's form usability guidelines open with "keep it short" and recommend removing any field whose information can be "collected more conveniently at a later date". A discovery call is that later date. Before it, you need two or three facts: what the person wants to change, roughly when, and whether they or an employer would pay.

The assistant gets those in conversation. Tell it what to ask in its persona instructions:

You are the AI assistant for Maya Chen Coaching. Maya coaches first-time managers in tech.
Warm, direct and brief: two or three sentences per answer.
When a visitor describes their situation, ask one question at a time: what they want to change, and whether they or their employer would pay.
If they match the "who I work with" page, offer the free 20-minute discovery call and book it.
If they don't match, say so kindly and point to the free resources page.

PepoChat's Lead Capture setting decides when the assistant asks for contact details. For a practice, "when the agent needs a follow-up" or "before handing off to the team" keeps ordinary questions free of friction; the agent asks once, in one sentence, and never asks again if the visitor declines. A booking captures the name and email anyway. The guide to AI chatbot lead generation compares the three timings in detail.

The strongest qualification signal is the booking itself. A prospect who picked a slot after reading your "who I work with" section has done more self-selection than a form could ask of them.

How does booking a discovery call in the chat work?

The assistant checks live availability on your calendar, offers a few slots in the prospect's own timezone, and books the one they pick, with the confirmation and meeting link sent by the calendar. In PepoChat this runs through a PepoSmart scheduling calendar connected with an API key on the Appointments page.

The part you control is which meeting types the assistant may offer. You switch on each one, up to 12, and write a short "when to offer this" note that the assistant follows every time. You can add up to 2,000 characters of general scheduling instructions, such as "Only offer discovery calls to prospects, not current clients."

When a prospect asks to talk, the assistant checks the matching type over the next days, up to 14, and shows a handful of openings in the visitor's local time; the booking is made in yours, so nobody does timezone arithmetic. If the visitor has verified their email, the booking uses it; otherwise the assistant asks for a name and email. PepoSmart then re-validates the slot so two people can never take the same one, assigns a host (round-robin works for a team), creates the meeting link and sends confirmation emails and reminders. A Find a time button can also open PepoSmart's booking page inside the widget for people who prefer to browse.

Paid sessions stay out of the chat by design. If you sell an intensive, link its booking page from your site; the assistant can explain what it includes.

Can the chatbot remember returning clients?

Yes, if you switch it on and the client proves who they are. Meeting memory lets the assistant recall what a returning, verified client discussed in past meetings with you: dates, key interests, commitments and summaries, taken from your PepoSmart meeting notes.

It is off by default and has two gates. You enable it under Plugins, then PepoSmart. And the visitor verifies their email in the chat with a six-digit code sent to that inbox; without verification, nothing is shown. Raw transcripts, recordings, coaching scores and details about other attendees never cross into the chat, and every access is logged on both sides. The assistant refers to meetings by date, says when it is using the notes, and uses only what is relevant to the current question.

For a client between sessions, that turns "remind me what we agreed" into a useful answer at 11 p.m. For you, it raises a confidentiality question first. The ICF Code of Ethics asks coaches to "maintain the strictest level of confidentiality with all parties involved". Before switching meeting memory on, tell clients in your agreement that your website assistant can recall meeting summaries once they verify their email, and leave it off if your sessions regularly cover health or other sensitive personal matters. The explainer on what meeting memory is covers the design in more depth.

When should the chatbot hand off to you?

Whenever the visitor asks for you, whenever it cannot answer from your content, and whenever the conversation turns personal. PepoChat escalates on its own when a visitor asks for a person or accepts its offer of help after it finds nothing relevant, and offers a person to a frustrated visitor.

Solo practitioners rarely watch a chat inbox live, and the Team availability setting has an option for exactly that: "No one answers live chat, we follow up by email." The assistant then never promises a person. When someone asks for you, it gives your away message, asks once for their email and hands the conversation over. You get an email with the last three messages and a Reply by email button, at most once per conversation every six hours. This matters because visitors are not emailed when you reply in the inbox, so the address is how you reach them. The human handoff docs describe each mode.

New leads can also reach you the moment they appear: a Slack card or a signed webhook on any plan, or an email to the team on Starter and Growth. In a study of how companies follow up web leads, Harvard Business Review's The Short Life of Online Sales Leads concluded that "most companies are not responding nearly fast enough." The assistant replies instantly; the notification makes sure you do too.

What are the ethical limits of an AI assistant for coaches and consultants?

It must say it is an AI, it must not advise, it must not promise outcomes, and it must route anyone in distress to real help. These are hard lines in a profession built on trust.

Disclose the AI, and do not impersonate yourself. Name it "Maya's AI assistant", not "Maya", and say so in the greeting. The guide to AI chatbot disclosure laws covers the EU and US rules.

No advice beyond your published content. PepoChat's assistant answers only from your knowledge base and cannot be instructed out of that. Put the rest in the "Never do this" field, which it follows even when the knowledge base holds the information. The ICF code asks coaches to "work within the boundaries of my level of coaching competency" and to "make verbal and written statements that are true and accurate about what I offer". Your assistant's replies are written statements made on your behalf.

Never give coaching, career, legal, financial, medical or mental-health advice. Offer a discovery call or a person instead.
Never promise results, promotions, income or timelines.
Never quote a fee that is not on the pricing page.
Never say or imply that you are Maya. You are her AI assistant.
If someone mentions self-harm or a crisis, give the crisis line from the knowledge base and offer to connect them with Maya.

Plan for distress. Visitors to a coaching site are sometimes in a hard place. The assistant can only quote what is in your knowledge base, so add a short page or Q&A pair with crisis resources for where your clients live. In the US, the 988 Suicide and Crisis Lifeline takes calls, texts and chats around the clock.

How to set up an AI chatbot for a coaching practice

Plan an afternoon, most of it on content rather than settings.

1. Create a workspace. No credit card is needed.
2. Add your website and PDFs. Crawl the site, upload program outlines, and add Q&A pairs for the fit and price questions your pages miss.
3. Write the greeting and suggested questions. Name the jobs: "I'm Maya's AI assistant. I can explain how coaching works, what it costs and book a free discovery call."
4. Add agent instructions: the persona and "Never do this" rules above.
5. Set Team availability. Choose "No one answers live chat" if you follow up by email, or business hours if you check the inbox at set times.
6. Turn on lead capture and a notification channel, such as Slack or a webhook.
7. Connect PepoSmart on the Appointments page, switch on the discovery call with its note, and leave paid types to your booking page.
8. Decide on meeting memory after updating your client agreement.
9. Install the widget with one script tag or the WordPress plugin, and add your domain to the allowed list.
10. Test in the drawer. Ask the questions from the table above. Bookings made in a test chat are real, so cancel any you create.

Every feature here, including booking and meeting memory, is on the Free plan, which includes 200 AI replies a month and one team member. New-lead emails to the team are the one paid extra, from the $49-a-month Starter plan; the pricing page has the details.

What to do next

Write or tidy the three pages that matter most, who you work with, how it works and what it costs, then create a free workspace and point the assistant at them. Connect your calendar and switch on one discovery-call type with a careful "when to offer this" note. Before going live, ask it the hard questions from the table, especially the requests for advice, and read every answer as if a prospect had sent it.