Insights

You and your competitor use the same AI. What will decide the winner?

AI knows the whole world and almost nothing about your company. The difference is what you give it to read.

If you have started adopting AI, it already helps your employees. It writes their emails faster, summarizes their documents, translates their correspondence. The catch is that it does exactly the same for your competitor’s employees, because you are both renting the same models, at the same price, and most likely from the same providers. And there are two problems hiding inside this comfortable picture. The first is that this help is heavily subsidized: the AI companies are buying market share, which means you are getting hooked at prices that are not real and will not last. The second is more important: help that everyone receives equally is not a competitive advantage.

In my last twelve years in analytics and in building decision support systems, one thing became clear to me: a company is as good as its leadership, its people and its technology. For decades, technology was a place where an edge could be bought. If the technology, which today means the intelligence, is now identical on both sides, then the difference between companies has to come from the other two.

And that allows us to raise the question that matters. The mass media keep feeding us the idea that AI will replace people. The simplest question we can pose back is this: how can it replace us if we are still analog? Many owners consider their companies digital. But digitalization so far had two chapters: first we moved from paper into programs, and then we moved the data out of the programs into governed foundations where it can be analyzed, which is the work I described last week. The next chapter is a question almost nobody has asked yet: do you document your decisions, and how?

Here is why that question decides everything. With a governed foundation, AI can analyze your numbers autonomously and tell you what happened. But when a KPI drops in March, the explanation is rarely in the data alone. It is the pricing decision from February, the campaign that was paused, the approval rule that was tightened in a meeting nobody recorded. AI has read the entire published world, but the reasons behind your decisions were never published anywhere, so the most intelligent tool ever built walks into your company as the most ignorant new hire you have ever employed. Owners fear it knows too much. It knows too little to replace anyone, and too little to help. So before we ask whether AI will replace us in decision making, we should ask what it would need to know to even take part.

The answer is not a new invention but one of the oldest ideas in management. Peter Drucker saw knowledge work coming back in 1959 and gave the knowledge worker its name in the sixties, and the quality movement has repeated the same rule for decades: if you cannot describe your work as a process, you do not know what you are doing. What the agent era changes is the economics of that old truth. Documentation used to be hygiene, something you did for the auditor. Now it is horsepower. Anthropic proved it on themselves: the same AI that answered correctly 21 percent of the time jumped above 95 the moment they wrote their internal know-how down and gave it to the machine, the untold half of the story I described in my previous article. Agent skills are management skills, written down.

Now run the comparison forward. Two companies, same industry, same size, same software, same rented model. Let us go one step further: say both of you have even built the governed foundation, so both AIs read the numbers correctly. Even then, the race is not a tie. The first company also writes down its exceptions and the reasons behind its decisions. The second keeps them in the heads of experienced people, who will eventually resign, retire, or simply forget. The first company’s agents get better every month, because every written page trains them further. The second company’s agents stay generic forever, as generic as the competitor’s. The gap between the two shows up on no balance sheet until the day it is unbeatable, and it may grow bigger than today’s gap between the companies that use AI and those that do not.

There is another perspective that reflects the same change equally well. In companies with proper knowledge management, speed increases because knowledge stops queueing. The business person prepares a proper request together with the IT agent instead of opening a ticket and waiting a week for the review. Sales automates the routine decisions inside clearly written rules. Knowledge finally flows between departments, which is exactly where companies leak the most value, and the quiet status cost of asking, the fear of looking uninformed in front of another department, disappears. A document can look perfect and be wrong, so the people who review and validate matter more in this world, not less.

There is one more way to read this change, and it is maybe the most important of all. This is not replacement. Human intelligence is still the only one that sets its own questions and answers for the consequences, the one that monitors and evaluates the outputs, and nothing in this era retires that. What happens is a trade: the standardized decisions and actions move to the machine, and what stays with people is everything that is hard, the exception, the new product, the reason to change course. AI is automating our comfort zone, and pushing our intellect toward the harder questions, where innovation remains our responsibility.

So what should you do? Start with a flip: companies have spent two decades learning to monitor the business, the reports and the KPIs, and the next step is to monitor the decisions. Write down the ones your company makes every week, the approval, the discount, the exception, who makes them, on what grounds, and what happened afterwards, and keep that record where your own agents can read it, not in a public chat window. Then automate step by step: hand the machine the standardized decisions first, and let every automated decision fire instantly, with the reason attached. The hard ones stay with people. And test every documented decision with one question: could it be executed tomorrow without calling a meeting? If not, it is not documented yet.

The winner between you and your competitor will not be decided by the AI you both rent. It will be decided by what your company knows without its people, and whether that knowledge is written where your agents can read it. And who will manage those new colleagues? That is a question of its own, and I will address it separately.

Value Discovery

Bring the decision that is late.

A free Value Discovery, sixty to ninety minutes, built around your industry and your Monday morning.