Why You Build Specialists, Not Buy AI
By Jeff Chung - Chief Strategy Officer
07/22/2026
Ask a leadership team what they should do about AI right now and the question usually comes back the same way. Which one should we buy?
That question comes from procurement, and procurement is the wrong verb for what actually has to happen next.
You are not buying AI. You are standing up a workforce of specialists. The distinction sounds cosmetic. It is not.
What "buying AI" gets you
Walk through the "buying AI" motion end to end and you see what you actually end up with.
A general-purpose chatbot license, which one person on the team will explore for a week and then set aside because it does not know the business. An agent framework, which requires engineering to plumb into anything specific enough to matter. A copilot subscription, which sits inside a tool people already use and makes certain steps faster without changing what the work is. A vector database or two. Some prompt libraries. Committee meetings about governance.
None of that is a workforce. It is infrastructure looking for a role to fill.
The failure mode is predictable. Six months in, the AI budget has been spent. The tools are technically deployed. And someone starts asking why the work does not look different yet.
The tools were never the thing that was going to change the work. A specialist standing up inside a defined role was.
What "building specialists" actually means
A specialist is not a chatbot with a wrapper on it. It is a scoped, roled, memory-carrying entity that holds a slice of the work end to end.
Building one is closer to onboarding than to installation. You define the role. You give it identity: who it is, how it introduces itself, what tone it holds. You give it memory: what it knows about the business, the customers, the history. You give it skills: the actions it can take, the tools it can call, the code it can run. You give it operating rules: where its authority ends, when it pauses, when it escalates.
None of that is bought off a shelf. All of it is designed.
That is the verb shift. Not procurement but composition. Not a license but a workforce, standing up one specialist at a time.
The purchase pattern is different
Because the unit of value is a specialist, the way you buy is different too.
You do not license a platform per seat and wait for adoption. You stand up a specialist for a defined outcome and measure whether the outcome moves. You do not buy tokens by the million and hope the volume matches your value. You have a specialist and a job, and you know what job it is doing.
The buyer question is not "which AI vendor," because the vendor is not the thing that produces the outcome. The specialist is. The buyer question becomes "which slice of work first, and who does that specialist need to be?"
That is a workforce question, not a procurement question. It gets answered by looking at operations, not by comparing vendor decks.
What this changes for the room
The AI conversation has been sitting in IT and procurement for two years. Neither is the right room.
Standing up specialists is workforce design. It sits with the leaders who own how work gets structured. It requires the operating knowledge that lives inside functional leaders, not inside vendor relationships. It produces org design work, not just software rollouts.
There is no way to buy through this shift. The buying is a small step inside a much bigger design decision.
That is why the question is not which AI to buy. It is which specialists to build first, and what the workforce standing up around them will look like a year from now.
Chief Strategy Officer, Arkane Digital
Jeff advises organizations on AI transformation, focusing on connecting business strategy to practical implementation without fragmented or reactive adoption.