Every week another AI and automation services firm appears. Most describe themselves in almost identical terms. That raises a question worth asking out loud: once everyone has access to the same models and the same tooling, what is left to compete on?
Not the technology. The best models are available to anyone willing to pay for them. Integration patterns are becoming standard. A convincing demo takes days rather than months. Anything you can buy, your competitor can buy as well.
What has not commoditised is access to the parts of a business that were never written down.
In most organisations the process documentation describes an idealised version of the work. The real version sits with the people doing it: the exceptions they handle without escalating, the judgement calls they make, the reason a step exists that nobody has questioned in years. AI is good at reading systems. It is far weaker at reading people, and it cannot read what nobody has recorded.
So the constraint is not technical. It is whether people will tell you the truth, and people tell the truth to those they trust.
There is often a good reason they will not. With AI, the person explaining the exceptions can see exactly where the explanation goes. Every edge case they hand over is a line in the thing that will handle it next time. Holding one back is not obstruction, it is a rational read of their own position, and it is invisible, because an exception nobody mentioned looks identical to an exception that never existed.
That puts a fairly ordinary set of people skills at the centre of the work. Turning up. Being straight, including when it costs you something. Doing what you said you would do. Treating the person who has done the job for fifteen years as the expert, because they are. Being someone a nervous team is willing to be honest with about how the work really happens.
None of which is the whole job. Trust gets you an accurate picture, not a good decision about what to do with it. Plenty of people are trusted and still build the wrong thing. The point is narrower than it sounds: without the accurate picture, the judgement that follows has nothing reliable to work on.
It is why two AI consultancies founded in the same month can read identically on paper and behave nothing alike. A developer who starts one arrives with the build already solved, and has to learn how to sit in a room with people who have every reason not to help. A business architect who starts one arrives with that conversation already familiar, and has to learn what the technology can and cannot do. Both gaps are real. In my experience, the second tends to close faster.
Whichever gap they start with, though, both then hit the same wall, and it is a bigger one: will larger organisations ever seriously use smaller players?
Trust of this kind attaches to individuals, not to logos. Large firms know that, which is why they sell something else entirely: transferred risk, indemnity, a name a board can point to if it goes wrong. Procurement is built to buy exactly that, and head to head a small firm will lose to it almost every time. It would be dishonest to pretend otherwise.
The awkward part is that the two purchases work against each other in practice. Risk is transferred by adding formality: scope fixed up front, layers of contract, a supplier large enough to absorb the consequences. Knowledge does not arrive that way. It comes out in unguarded conversation, from people who cannot be obliged to mention an exception they have never mentioned before, and the scope cannot be fixed up front because working out what it is happens to be the job. Almost everything that makes the arrangement safer on paper puts more distance between the supplier and the people it needs to hear from.
Large organisations end up buying both anyway, frequently on the same programme. Which is why the realistic ways in are shaped by relationships rather than credentials:
None of those routes look impressive. I suspect they account for most of the work that gets done.
Which leaves me with questions rather than answers.
Interested in how others see it.
