Why I unhired five people.

In June of 2023, I sat across from a senior engineer I had hired four months earlier and told him the work was ending. He had relocated to Bangkok for the role. He had just shipped his first feature, and it was good. The thing I had to let him know was that I had been wrong about what kind of business I was building — and that meant I had been wrong about hiring him.

That conversation, and the four others like it I had over the next ninety days, became the foundation of every engagement I run today.

The hire

In late 2022 I had been running consulting solo for a few years. Six-figure engagements with small but ambitious teams. Then a few clients asked at once for things bigger than I could ship alone — a fintech wanting a fraud detection platform, a Nordic studio wanting full multi-currency invoicing, a clinic network wanting a consent rewrite that would touch a hundred screens.

The conventional wisdom answer is what every founder hears: scale up. Hire. Build a team. So that's what I did. By summer of 2023 I had six people: two senior engineers, a designer, a project manager, and two juniors I was training up. An office in Bangkok. A Slack with channels. The whole thing.

The clients said yes to bigger contracts. The team shipped. From the outside it looked like I had built a small consultancy.

I have not been more wrong about what I was building.

The cracks

Three things started showing within months. I missed all of them at the time.

Quality went down as headcount went up. Each client had been buying me — my judgment on their problem, my hands on their code. With a team in front of me, what they actually got was a relay. The brief came to me. I translated to my designers. They shaped a spec. The PM wrote the ticket. An engineer wrote the code. A senior reviewed it. I signed it off. By the time the work reached the client, three layers of context had been lost and three layers of did I get this right? anxiety had been added. The work was technically fine. It just was no longer mine.

Margins compressed. Six salaries plus pension contributions plus tools plus an office plus the time I was spending not billing. The bigger contracts looked great in the proposal stage. After overhead they earned less per hour than my solo work had. I was doing more, growing more, and earning less.

I became a manager. Sixty percent of my time was 1:1s, hiring, performance conversations, coordination calls, and resolving disagreements between people who all wanted to be right. I had taken twenty years of building software and traded it for the operational role I had spent twenty years actively avoiding. The work I loved was now what other people did, and what I did was the work I had always disliked most.

I did not see these clearly enough, soon enough. I told myself the cracks were growing pains. They weren't. They were structural.

The unhirings

The decision didn't come at once. It came over five conversations, three months, and two trips to the office at six in the morning before anyone else arrived — to think clearly, without interruption.

The first conversation was the easiest, in the way the obvious things are easy. A senior engineer pulled me aside and said something like, "I think we both know I'm not getting the work I came here for." He was right. I owed him the conversation. We agreed to wind down his role over four weeks, and I helped him land in a senior role at a Singapore scale-up. I am still in touch with him.

The next four were harder, because they came from my decision and not from mutual recognition. Two of them I had recruited specifically. One had relocated to Bangkok for the role. One had passed up an offer at a company that paid more.

I will not write the details of those conversations. They were difficult to give and difficult to receive. I did them in person, in single sessions, on a single day each, with a notice period that was longer than I was contractually obliged to give and a settlement structure that I am told was generous. None of that made it okay. It made it less worse.

By October of 2023 the team was one person, a junior I had trained who had become very good. By February of 2024 he had taken a senior role somewhere that could pay him properly for what he had become. We are still in touch, and I helped with the negotiation that got him the senior offer.

I had hired six people in eighteen months and ended that period with none of them.

The misdiagnosis

The word that haunts me from that period is capacity. I had told myself I was solving a capacity problem. The clients want more than one person can ship. Hire more people. Solve the math.

What I was actually destroying was continuity. The unique thing my consulting offered — one mind on the problem from brief to ship — could not be replicated by adding more minds. It was an architectural feature of being one person, not a temporary constraint to be scaled past. The clients were not paying me to be a small agency. They were paying me to be the operator they could call directly.

The agency model sells capacity. The operator model sells continuity. They are different businesses.

The agency model sells capacity. It is what is easy to scale. You hire, you sell more, you hire, you sell more, and the unit you sell — a body of work-hours — is fungible.

The operator model sells continuity. It is harder to scale. You can only sell as much as one person can ship, and the unit you sell — one mind on the problem — is not fungible at all. But it is much harder for a competitor to copy.

I had spent eighteen months trying to scale the harder thing using the playbook of the easier one. The result was that I ended up with neither.

Three rules I now live by

I came out of that period with three rules. They sound simple. They cost me a lot of money and a lot of relationships I now cannot have back.

1. The choice between capacity and continuity is a one-way door.

Once you scale a solo practice into an agency, you cannot un-scale without burning the relationships that brought you the original work. You cannot tell a client who has been with you for three years that the senior engineer they have come to trust is no longer on their account because you have decided to refactor your business model. You can do it — I did, with the partial version of it — but you should not. Make the choice deliberately, once, with your eyes open. Do not stumble into the wrong one because the conventional wisdom says scale.

2. The only way to scale a solo operator is with tools, not people.

Every hour I spend on something a tool could handle is an hour I am not spending on the thing only I can do. Today my work runs alongside five named AI agents — Cassiopeia, Atlas, Cartographer, Sextant, Polaris — each handling a specific slice of toil that used to be either my problem or another human's. A platform that runs twenty-four hours a day and answers client questions when I am asleep. Code review that catches what tired eyes miss at midnight. Eval harnesses that block deploys when a model gets quietly worse. None of these replace judgment. All of them replace the work that is beneath judgment.

3. Charge for the operator, not for hours.

Hourly billing aligns the incentive in the wrong direction. The client wins when the work takes longer; you win when the work takes longer; nobody wins when the right answer is less work. Subscription pricing — included hours, response-time SLA, included AI usage — aligns incentives differently. The client is paying for access to the operator. The operator is paid to make the work small. Both sides are pulling toward the same thing.

What this means for clients today

Every engagement I take on now starts with these three rules already baked in.

You are working with one operator from brief to ship. Not a designer who hands off to an engineer who hands off to a senior. Not a PM who translates between you and the people doing the actual work. The same person who hears your first call writes the spec, ships the code, and reviews your bill.

The toil is handled by named tools that are honest about being tools. Not "we leverage AI" handwaving — five specific agents, each on a specific job, each disclosed with the model they're built on. The hours I bill are spent on the parts that need a human at all.

The pricing is predictable. Most clients sit on a monthly subscription that includes response time, included hours, included AI tokens, and a strategy call. They know what next month will cost. So do I. There is no incentive on either side to make the work bigger than it needs to be.

This is what the decision in 2023 cost me. It is what every client benefits from now. The five people I unhired are not on this page; the lessons they taught me are in every line of how I now work.

A note for founders considering the same

If you are reading this and thinking about whether to scale a consulting or product practice, the question I wish someone had asked me before I hired the first person is this: would you rather sell access to yourself, or sell access to a team?

These are different businesses. They have different unit economics, different sales motions, different defensible advantages. Most founders default into the agency model because it looks like growth. Some of them are right to. Some of them are me in 2022, about to spend eighteen months learning a lesson that could have been a conversation.

If you decide to stay solo, do it with the tools to make it possible. If you decide to scale, do it because you actively want the agency, not because you accidentally hired your way into one.

There is no wrong answer. There is only the wrong answer arrived at without having asked the question.

— Mikkel, Bangkok