The management consulting market is worth roughly a trillion dollars, and in the first half of 2026 the AI labs stopped pretending they were not interested in it. Both OpenAI and Anthropic now position advisory work as a category they intend to serve directly, not as a byproduct of selling model access.
The consulting response has been predictable and mostly wrong. Firms are publishing pieces about the irreplaceable nature of human judgement, the importance of relationships, and the limits of automation in complex change programmes. Most of that is true. None of it addresses what is actually happening.
What is actually being competed for
Repricing toward zero
Holding or rising in value
Synthesis of public information
Knowing which question is the right one
First-draft frameworks and decks
Having seen the failure mode before
Benchmarking against published data
Accountability for the outcome
Document production and assembly
Access to a specific room
AI labs are not competing for the partner conversation in the boardroom. They are competing for everything underneath it: the market scan, the benchmarking deck, the process map, the maturity assessment, the options analysis, the implementation roadmap. That work is the majority of billable hours on most engagements and it has always been the part clients were least happy paying for.
Productivity gains of 30 to 60% across knowledge functions are now the baseline expectation for AI-enabled firms rather than a competitive differentiator. When the baseline moves, the price of the work moves with it. A deliverable that took three analysts two weeks and got billed accordingly does not survive the client asking why.
The repricing, not the replacement
The useful frame is not that AI replaces consultants. It is that AI reprices the components of consulting. Some components collapse in price to near zero. Others hold or increase in value, because they become scarcer relative to the abundant parts.
“AI does not replace consultants. It reprices the components of consulting, and the components that hold their value are the ones you cannot buy from a model.”
What collapses: synthesis of publicly available information, first-draft frameworks, benchmarking against published data, document production, and anything where the value was mostly assembly. What holds: knowing which question is the right one, having seen the failure mode before, carrying accountability for the outcome, and access to a specific room.
Where this leaves an independent
AI does not replace consultants.
An independent consultant is better positioned for this than a large firm, and most independents do not realise it. A firm has a cost structure built around leveraged junior time. When the junior work reprices toward zero, the pyramid stops working. An independent has no pyramid to defend.
Turn what you know into what you own.
Vibepreneur builds structured ventures from professional expertise, with positioning, launch assets, and growth systems included.
Join the WaitlistWhat an independent does have is a repeatable diagnostic they run on every engagement, a set of judgements they make faster than anyone else in their niche, and a client base that trusts them. Those three things are a product. They have simply never been packaged as one.
From engagement to system
The move is not to compete with AI on volume of analysis. It is to encode the judgement layer into something that runs without you, and to sell that. If you have run the same operational diagnostic forty times, the diagnostic is an asset. If you know which three signals predict whether a transformation programme fails, that knowledge has a shape that fits into software.
This is what most of our showcase ventures have in common. ChartTrace started as a practice director who knew which documentation gaps caused denied claims. ScopeGuard started as an agency director who knew where margin leaked. Neither built a consulting practice into a bigger consulting practice. Both encoded the judgement and sold the encoding.
The timing question
The reason to move now rather than in two years is that the encoding is cheap for the first time and the market is not yet crowded. Building a narrow vertical tool no longer requires an engineering team. What it requires is the domain knowledge to know what the tool should decide, which is the part the AI labs do not have and cannot easily acquire.
Nobody at a frontier lab knows which three signals predict a lease non-renewal in a mid-market commercial portfolio. Somebody who managed that portfolio for nine years does. That asymmetry is the whole opportunity, and it has a shelf life.
If you are a consultant working out what to do with that, the system walks through how expertise becomes a structured venture rather than a bigger practice.