I'm Training AI to Replace My Own Business

Introduction

We put it to Dan Boyles that most consultancies take a decade to go global, while he did it in under two years with a small team.

In this episode, we asked him what he knows that the big firms do not. His answer was refreshingly simple: his company does not sell anything. No product, no platform, no agenda.

From there, the conversation covered how ready businesses really are for AI, why it is not an IT project, the biggest mistake Dan made after launching his company, and the uncomfortable fact that he is using AI to do parts of his own job. We also talked about the Hello AI Summit at the Emirates Stadium, the never-ending renaming of Microsoft Copilot, and why hiring people for their personality still beats hiring them for their prompts.

Meet the Guest

Dan Boyles is the founder of Hello AI Collective, an AI consultancy that is only about eighteen months old. Before starting his own company he worked at Microsoft, where he was a Global Master Trainer, and at TikTok. He describes himself as someone who always struggled to work for other people, and says the idea was to build the company he himself would want to work for.

According to Dan, he and his team have trained around 17,000 people, completed 57 deployments and built more than 400 agents. The clients he mentioned include Monster and Formula One teams, and the company works with partners such as Microsoft, Lenovo, Adobe and Sage. Dan also speaks publicly a lot: 37 or 38 talks this year alone.

This was his fourth appearance on the podcast. You can find him on LinkedIn.

Setting the Stage

Almost every company is being told it needs AI. Almost every software platform now has an AI add-on. And much of the advice companies get comes from people who have something to sell.

That makes an independent view valuable, and it raises practical questions that most leaders have not answered yet:

• How ready is my business for AI, really?
• Who should own the rollout?
• Should we start now or wait until things settle down?
• And if AI can advise, what is left for consultants to do?

Episode Highlights

“We don’t sell anything”

Dan’s explanation for the fast growth is that Hello AI is agnostic. It does not care whether a client runs on Microsoft, AWS, Google or on-premises.

Every engagement starts with two weeks inside the business. At the end the client receives a report of well over a hundred pages, covering the business opportunities that were found (revenue-generating or cost-saving), the current tech stack and how hard it is to fit AI into it, the strengths and weaknesses of the data and the culture, an adoption plan and recommended tools. In other words, an eighteen-month AI rollout plan. What the client does with it is up to them.

Dan also credits visibility. The company hired a football stadium in its first six months, and a team of former directors of AI at football clubs, global brands and government who have done the work themselves.

Productivity is the floor, client experience is the ceiling

Dan’s newest keynote carries this message: do not chase productivity. If everyone becomes faster, nobody is faster; you have simply set a new baseline. What a business cannot buy is the client experience it can build from ten, twenty or thirty years of its own data and relationships.

Training AI to replace my own business

Asked how consultancy will change, Dan was blunt. He expects that paying someone to explain what AI could do for you will be redundant within about eighteen months, because AI can already do much of that. He is even using AI to advise him on how to advise people on tools he does not know yet. That is why, in his own words, the company is trying to exit quite quickly.

What he does not expect AI to replace is the knowledge of what goes wrong: the mistakes already made, the nuances, and the judgment that comes from doing it by hand.

Deep Dive: Is Your Business Really Ready for AI?

Nobody knows how ready anyone is

Cloud migrations had a tangible outcome and a readiness checklist. AI does not. Dan says every client has different needs, and the real variable is people rather than technology.

He described a large company with nobody who can handle the data and a junior IT team that has never run a transformation, next to a fifteen-person accounting firm that already has its licenses, labels and compliance in order. Leadership often thinks the company is ready when there is no training team and no culture shift prepared.

His encouragement: most businesses are more ready than they fear. Buy the licenses and use them. According to Dan, only about one percent of businesses in the UK are really using AI to gain an advantage, so almost nobody is as far behind as they think.

Fix the process before you buy the tool

Both Dan and Gerjon see the same pattern: a large share of AI requests turn out to be broken processes. Dan is happy to tell a client that this is a Power Platform problem, or that they should use their existing Microsoft stack better, and not spend money on AI.

One example: a company spending six figures a month on Claude for forty people. After digging into what it was used for, Dan estimated they could save around sixty thousand dollars a month by doing the same work with what was already included in their Microsoft licenses. It was “not as cool”, which, as James said, should not come into the equation when you run a business.

Why Microsoft is still where most clients live

About ninety percent of Dan’s clients sit in the Microsoft ecosystem. His reasoning: the guardrails, regulation and enterprise security are already there, and Azure lets you run other models, including Claude, inside the same environment. When someone files a subject access request, you answer it from one place.

He also sees a growing move towards private AI: buying hardware and running local models for regulated industries or for predictable costs, turning tokens from an operating expense into a capital expense. And he notes that more large enterprises are moving to Dynamics, with an Azure SQL database as the data foundation.

Real-Life Stories & Examples

The Hello AI Summit at the Emirates

Dan is an Arsenal supporter. Over lunch with partners, he asked whether they should hire a football stadium. Two weeks later he had a contract with Arsenal. About two hundred people came to the first Hello AI Summit, when the company was only seven months old, to hear practical messaging that was explicitly not a sales day. The contract for next year has been signed, and Dan has asked for his logo on the front of the stadium this time.

Microsoft keeps renaming everything

Copilot Chat, Copilot Cowork, Copilot for GitHub, Copilot for Sales, Copilot for Finance: the conversation turned into a roll call of Microsoft product names. Dan told how Copilot Cowork was free until the second of September, after which consumption billing started and some users were paying fifteen to seventeen dollars to build a single PowerPoint deck, without anyone having warned them.

His colleague Ethan now spends half a day a week tracking Microsoft changes, because buttons keep disappearing from the admin center and training material goes out of date. Dan’s message to Microsoft: stop renaming your products. Gerjon pointed out the upside for consultants, since if the consultants do not understand it, the customers will not either.

A second brain built on Azure SQL

Dan’s company keeps every proposal, technical design and adoption plan in one Azure SQL database, so anyone on the team can use all of it. When he had to prepare a training for Monster in Australia on short notice, he asked the system for the earlier material, and within ten minutes it produced the trainer deck, the trainer document and the end-user deck.

The same approach helps with relationships and finance. It remembers details Dan would forget, and it helps him with reports (Excel is not his strength) as the company prepares for an acquisition. It does not generate revenue directly, he says, but it lets him deliver the best work he can.

AI agents for sales and proposals

Dan has built an agent that tracks his top ten competitors: who they hire and fire, which sectors they are strong in. Every Monday it tells him which sector to target. When competitors dropped their sports divisions, he went after specific sports teams. For one prospect, a proposal arrived within five minutes of the first call, styled like that prospect’s own menu (Wingstop). The long assessment reports are made the same way: Dan talks to his AI tool for a few hours and the first draft is done, instead of three or four weeks of writing.

Hiring for personality

Dan asks applicants to give a ten-minute presentation on anything except AI. One candidate, a former linesperson, taught him the offside rule in the car park and got the job on the spot. Another showed him how to make an Eton mess. He wants to see what makes people happy, because he can teach them AI but not interaction. Gerjon’s own hiring question is whether you would want to sit next to this person at Christmas dinner.

Dan also described an interview where a candidate was obviously using AI to answer live, and fell apart on a random follow-up question. His new company in the US is called Say Hi, where “HI” stands for human intelligence.

Key Takeaways

• Staying technology-agnostic is a competitive advantage when everyone else is selling a product.
• A short, deep assessment of opportunities, data, culture and tooling is more useful to a client than another demo.
• Before you spend money on AI, look at the process. Many “AI problems” are process problems.
• AI is not an IT project. If you hand it to IT alone, adoption tends to stall at 10 to 20 percent. It needs a change team and clear ownership, but not the IT team.
• Do not wait for things to settle. They will not, and waiting is a decision that gets more expensive every month.
• Do not give AI to leadership only. The people doing the work every day need it most.
• AI is more likely to help you hit a target (for example not hiring more people) than to replace your sales team with agents.
• A demo is not adoption. Start from the work, give people the prompt, the workflow and the tested agent, and build it with them.
• Productivity is the floor. Client experience, built on your own data and relationships, is the ceiling.
• Judgment, experience and knowing what fails are what consultants will be paid for once tools explain themselves.
• In a world of AI-generated sameness, human connection is what wins work.

Closing Thoughts

Dan’s advice for leaders boils down to a few things: do not wait, do not hand it to IT alone, do not give it only to the top, and start with the work instead of the tool.

The most striking part of the conversation is that someone who builds his business on AI expects the tool-selling and AI-explaining part of consulting to disappear soon. What remains is the human part: the experience of having made the mistakes, the judgment, and the connection with the people on the other side of the table.

Thank you to Dan for joining us a fourth time, and for the invitation to do a live episode at the next Hello AI Summit (possibly from a private jet, once the tax bill is paid).

What about you? How ready is your business for AI, and who owns it? Let us know.