Running a business on AI only works if you use it yourself before you sell it to anyone else. We are a five-person company (and growing), and we run our own operations on the same AI agents and workflows we build for clients. We have also kept an honest scorecard of what worked, what stalled, and what we would do differently. Here is the real picture, including the parts that did not go to plan, because the honest version is more useful than the highlight reel. Running a business on AI is about practical experience.

We eat our own cooking

Every tool we recommend to a client, we use. Every workflow we design for someone else, we have tested on ourselves. Take a real client engagement. It starts with a meeting that is recorded and transcribed. Within minutes, we have structured minutes: decisions captured, action items extracted, risks flagged. That used to take someone 30 to 45 minutes after every call.

Running a business on AI means understanding the tools inside out. The insights we gain help shape our approach.

Those minutes feed the scope of work. The scope informs the configuration document. The configuration feeds sprint planning. That is the chain: meeting to minutes to scope to configuration to sprints to delivery. AI produces the first 60 to 80 percent at every step. Humans do the rest: the checking, the judgement, the “does this actually make sense for this client” decisions. Nothing gets lost between a conversation and the work.

The process of running a business on AI allows for continuous improvement of our methodologies.

The honest scorecard: what worked

The contracts. Our legal agent found sixteen issues we did not know we had, then redlined our own agreements from a customer’s perspective and told us we were being unfair. We fixed them. For the first time, our contracts match our values.

Running a business on AI requires us to adapt and refine our contracts to align with our values.

The people function. Our employment contracts were once cobbled together from templates found online. Now they are compliant, they cite the actual legislation, and they exist as a real framework rather than a hope.

And the surprise: the process forced us to define our own business. Every agent brief made us answer questions we had been dodging for years. What does this function actually do? What does good look like? We thought we were building AI. We were building clarity.

In running a business on AI, we learned the importance of defining our structure.

The honest scorecard: what did not

The finance function stalled. Not because the agent could not do the work, but because we could not give it what it needed. Our own financial information was not organised well enough to brief it properly. The AI was ready. We were not. An agent can only be as good as the context you feed it. If your house is not in order, it hands your mess back to you, neatly formatted.

When running a business on AI, it’s crucial to have organized financial information.

The early outputs were also too long. A four-person company does not need a 40-page compliance framework. It needs two pages someone actually reads. We had to learn that fit for purpose means the right amount, not the most.

Finding the right balance when running a business on AI can lead to more effective outputs.

What we would do differently

Start with the standards, not the agents. We jumped into building before we had defined what great looked like, then spent weeks retrofitting standards onto work already in motion. If we started again, the first month would be nothing but definition.

Our journey of running a business on AI taught us to prioritize defining standards.

This matches what the wider research shows. McKinsey’s 2025 State of AI survey found that adoption is widespread but most organisations still struggle to turn it into measurable value, with the jump from pilots to real impact being the hard part (see the McKinsey State of AI 2025 report). Clarity and discipline, not more tools, are what close that gap.

The real verdict

We are perhaps 40 percent through this journey. Nowhere near done. But that 40 percent has already changed how the business runs, and how clearly we think about it. The biggest surprise was not what the AI produced. It was discovering how much of our “people problem” and “output problem” was actually a clarity problem all along.

The clarity gained from running a business on AI has reshaped our understanding of operations.

Why this matters when you choose a partner

If you are evaluating any technology partner, ask them one question: do you use this yourself, every day, to run your own business? Not “have you done a demo”. Not “have you implemented it for clients”. Do you depend on it? The best partners are not the ones with the best pitch decks. They are the ones who have already hit the walls you are about to hit, because they went first.

In choosing a partner for running a business on AI, firsthand experience is invaluable.

Frequently asked questions

Can a small business really run on AI? Yes, in the sense that AI can fill and support functions a small team cannot otherwise afford, provided a human reviews the output and the underlying information is organised well enough to brief the AI properly.

What is the hardest part of running a business on AI? Defining your standards. The technology works; the difficulty is being clear about what good output looks like and keeping your own information in order enough to brief it.

Should I start with tools or standards? Standards first. Define what great looks like for each function before you build, or you will spend longer retrofitting standards onto work already in motion.

Work with a partner who runs on this too

We build with agentic AI and intelligent automation because we run our own business on them first. If you want a partner who has already made the mistakes so you do not have to, book a scoping conversation.

By running a business on AI, we offer insights that can help you navigate your own challenges.