BIONIC BUSINESS BY SAMUEL WOODS
A consultant or coach gets the most from AI agents by handing over the work around the client hours: preparing for calls, screening enquiries, following up on open deals and writing up sessions. The advice, the relationship and the price stay with you.
I'm Sam Woods, and I publish Bionic Business, a newsletter that thousands of online entrepreneurs read. This page is for consultants and coaches who run the practice alone or with a couple of contractors. It has eight cases from clients, the three agents I'd build first, one job that needs no agent, and three builds that cost their owners money or nearly did. Last updated November 2026.
Every figure on this page came from the business it describes. The last column of each table says what the figure leaves open.
Five that worked:
| Practice | What was built | Build time | Result | What the figure leaves open |
|---|---|---|---|---|
| Solo consultant, $290K a year | An agent writes a one-page brief before each discovery call and drafts the confirmation | Two afternoons | Show rate went from 62% to 81% across 94 calls | She changed the brief and the confirmation together |
| Consultant, $230K | An agent scores each enquiry against five criteria and drafts the reply | About four days | Average project value went from $6,200 to $11,400 | Five months and 83 enquiries |
| Executive coach, $185K | One saved prompt writes each session summary from the client's folder | Two evenings | Admin after a session went from 40 minutes to 9 | The time comes from her own log |
| Consultant, $165K | An agent assembles enrolment emails that can mention only what is already built | Three days | 38 seats sold against a target of 30 | One cohort, priced differently from her last one |
| Coaching business, $690K | An agent writes a one-page ad recommendation each morning and the founder decides | About ten days | Cost to win a customer fell from $412 to $317 | A landing page changed in the same months |
Three that went wrong, or nearly did:
| Practice | What the agent or prompt did | What it cost | What the figure leaves open |
|---|---|---|---|
| Business coach, $240K | One saved prompt turned group-call recordings into posts | Nothing, because a test came first. 5 of 20 drafts held a detail that identified a participant or that someone had asked to keep private | 20 drafts from six calls |
| Fitness coaching business, $1.7M | An agent moved ad budget each day on three-day results | An estimated $31,000 to $44,000 in lost revenue in one month | The loss is modeled, and the season may explain part of it |
| Fractional CFO firm, $1.1M | Three agents handled qualifying, call prep and proposals | Close rate fell from 44% to 23%, and two signed clients had no one assigned to them | Weaker leads in a bigger pipeline may explain part of the drop |
A solo consultant I work with, at about $290K a year, had an agent write a one-page brief before every discovery call and draft a two-line confirmation that mentions one thing it found. Across 94 booked calls, her show rate went from 62% to 81% and her close rate on the calls she held went from 29% to 38%.
She built it in two afternoons. She changed the brief and the confirmation at the same time, so she can't say which one did it.
The build is in Run This Briefing Agent (Stack, Prompts).
A consultant at $230K a year wrote five criteria for a client worth taking. An agent scores every enquiry against them and drafts an invitation or a decline for her to approve. Over five months she declined 54 of 83 enquiries, where she used to decline about one in five, and her average project value went from $6,200 to $11,400. She has overruled the agent seven times.
The build is in Run This Lead Qualification Agent (For Maximizing Sales).
Deals go cold on the days you're with clients. A Pipeline Agent keeps a watchlist of every open deal and drafts the follow-up, so it happens on those days too.
The build is in Run This Pipeline Agent (For Closing More Deals).
An executive coach I work with, at $185K a year, records every session with the client's consent. Afterwards one saved prompt reads that client's folder, which holds their goals, every earlier session note and what they agreed to do. It writes a summary for the client, a private note for her and three check-in prompts for the weeks between sessions.
Across 112 sessions, her admin after each session went from about 40 minutes to 9, by her own time log. She took on four more clients in the same working week, which is about $52,800 a year. She reads and edits every summary, and changes something in about one in four. Her clients also report finishing more of their actions between sessions, though that figure is their own score on a one-question check.
It took her two evenings, and a new client takes about fifteen minutes to set up. I'd build it as one project per client in Claude or ChatGPT.
An independent consultant at $165K a year presold a $1,490 group program when about 40% of it existed. Her sales page had two labeled lists: six items available on day one, and four to be built during the cohort, each with a target week, under a line saying anything on that list might move. The agent that assembles her enrolment emails can read only the first list.
She sold 38 seats against a target of 30, which came to $56,620, and refunds ran at 7.9% against her earlier average of 11%. Two of the four later items shipped on their target week, one shipped two weeks late and one was replaced after a cohort vote. Her previous cohort was priced at $890, so the two launches don't compare cleanly.
The build took three days, and separating the two lists took about an hour of that.
Two coaching businesses I work with put an agent on their ad budget, and they gave it different amounts of authority.
The first agent has no authority to spend. It belongs to an online coaching business at $690K a year that puts $31,000 a month into ads. Every morning the agent writes one page on what to cut, scale or leave, with 3-day, 7-day and 14-day return side by side, and the founder spends about ten minutes on it.
Over five months she approved 71% of its recommendations, and her cost to win a customer fell from $412 to $317 while spend rose 19%. A landing page also changed in July, and she doesn't claim the agent is the only cause. The build took about ten days.
The second agent moved budget every day on its own. It belonged to a fitness coaching business at $1.7M a year spending $94,000 a month, and it judged each ad set on three-day return. Most of that business's buyers book a call before they buy, and the median sale takes nine days. In one month the agent made 41 changes, and 12 of them cut an ad set that lost money at three days and made money at fourteen.
The business estimates $31,000 to $44,000 in lost revenue that month. That figure is modeled from how the ad sets performed before the cuts, and the season may explain part of it. The nine-day lag was in the founder's head and in the CRM, and nowhere the agent could read.
A business coach at $240K a year, with a part-time editor, planned to turn recorded group calls into newsletter and social posts with one saved prompt. Before publishing anything, they tested 20 drafts against six real recordings.
Three drafts held details that would identify a participant to the rest of the group, and two more used a situation a participant had asked to keep in the room. Nothing was published. Members of a group program pay partly for confidentiality, and nothing in the setup recorded who had agreed to what. The test covers 20 drafts from six calls, so another group could turn up more or fewer.
A fractional CFO firm at $1.1M a year added three agents in one month, for qualifying enquiries, preparing discovery calls and drafting proposals. Qualified opportunities went from 9 a month to 21, and proposals from 6 to 14.
Close rate fell from 44% to 23%, and the wait between a signed proposal and the first session went from 9 days to 26. Two clients who signed in September had no CFO assigned to them. The firm can serve 17 clients at a time and had 15, and none of the three agents knew that number. The firm puts the lost deals down to the delay, and weaker leads in a bigger pipeline may explain part of it.
If you work alone, your own calendar is that ceiling. I'd write down how many clients you can serve before putting an agent on lead flow.
Keep the advice, the pricing, the hard conversations and anything a client would feel. A good starting rule is that agents draft and you send, and you loosen it one job at a time. The consultant with the call brief, the consultant with the enquiry screen and the executive coach all still approve what goes out.
In the cases on this page, agents write the brief before a discovery call, score new enquiries against written criteria, assemble enrolment emails and prepare a daily ad recommendation. A saved prompt writes the session summary. In the ones that worked, everything a client or a prospect reads was written or approved by the owner.
The call brief took its owner two afternoons and the session write-up took two evenings, with no developer. The enquiry screen took about four days, after two sittings to write the five criteria. The issues above include the prompts.
The five builds on this page that worked took between two evenings and about ten days. You'll also need a Claude or ChatGPT subscription. Everything else is the calendar, email and CRM you already pay for.
Use Claude or ChatGPT for a saved prompt or a Skill. Use Cowork or Codex for an agent that runs on a schedule while you're with a client.
I'd have it recommend and keep the spending decision with you. The coaching founder above gets one page each morning and has reversed four decisions she first approved. She will consider letting the agent act alone after 60 recommendations in a row with none reversed, and she is at 19.
You can once you have a record of who has agreed to what. The business coach above couldn't publish five drafts in 20, and no prompt can know what a participant asked to keep private.
Start with the stage that limits your income. The CFO firm automated everything before the sale while delivery was close to full, and its close rate halved. The firm blames the wait, and weaker leads may explain part of it. The $230K consultant used her agent to turn enquiries down: she declined 54 of 83, and her average project value rose from $6,200 to $11,400.
Each issue above walks through one agent with its prompts, and all three are free to read. I teach you to build and run the system yourself, so you can change it when your practice changes. If you haven't built anything yet, The Pocket Guide To Your First Working Agent starts with your inbox.
Bionic Business is a newsletter by Samuel Woods. Three free issues go out each month, including The Scan, my round-up of AI news with a verdict on each item, and issues that walk through one system you can run that week.
Cortex is the paid side, at $50 a month. Members get a strategy issue and a build guide each month, with the files to install the system, plus Skill and agent packs. You can cancel at any time, but there are no refunds, for any reason. All sales are final. See Cortex.
Subscribe free: