Someone whose only job is to answer what AI changes about your business, your economics, and who you hire. Sized to the need, standing, and not permanent.
AI has become a board-level question at your firm. Nobody’s job description contains the answer.
Not because nobody is working on it. Someone in technology is running a pilot. Someone in operations sat through a vendor demo last week. Someone on the executive team has read enough to be worried and not enough to be useful. There is activity.
What is missing is a single person who owns the answer, whose read is current rather than eighteen months old, and who is accountable for what gets decided. Most firms discover this the first time a decision has to be made with money attached: a contract to sign, a build to fund, a role to post. The question goes around the table and comes back unanswered.
Many firms are too small to justify the role and too exposed to go without it. The work does not scale down to a hire. It scales down to a seat.
What was worth building six months ago is now a feature inside something you already pay for. What looked like a durable advantage turned out to be a wrapper. What your vendor called AI in a sales deck is not in the service agreement, and the productivity they claim may never reach your price.
Meanwhile the standard for what counts as acceptable output keeps rising, because your competitors are getting faster and your customers can tell. None of this settles into a stable answer you can decide once and file.
You will not get there by adding people. The volume and the compression scale against each other, and the accuracy the work now requires is above what stacked layers of human review deliver at speed.
It is what to build and what to buy, what to invest in now and what to wait on, where to experiment, how work gets validated when a machine produced it, and what it does to hiring and training.
Those questions have new answers every quarter, and at most firms nobody has answering them as their actual job. That is the seat.
What the things you intend to launch, sell or deliver actually require, and how much of it you can do today.
What is a real product, what is a wrapper on a model you already pay for, and the call before you sign.
How accuracy beyond human review gets engineered, and who signs before machine-produced work reaches a customer.
A few real workflows, instrumented before they change, so the answer is evidence rather than opinion.
What this does to hiring, and how your team becomes the team that stands behind machine-produced work.
What changed, what it means for your operation, and the two or three things worth doing.
More than a monthly check in, less than a hire. Monthly rather than hourly, and enough presence to be in the room when decisions are made rather than told about them afterward. Some firms keep the seat. Others use it as a bridge to the full time hire, and making that case is part of the job.
The first ninety days instrument current state, decompose a workflow, and map the capability gap against where you are going.
It runs without me. If a quarter passes with nothing owned or decided, hold me to it.
I sell no software, take nothing from vendors, and have no bench to staff. Where a tool is needed I will pick the best one, and where you already own it I will say so.
I helped found Booz Allen’s Value Engineering practice, led transformation at JPMorgan and at Citi, and built the transformation function from the ground up at two more firms.
Most of the business at a fund administrator, the largest institutional relationships at a global provider, and a P&L as a general manager. I know what it costs to change something that is already running.
Including the one I run my own week on. I design and build my own agents rather than commissioning them, which is why my read on what is real is current rather than briefed.
The operating problem repeats across industries. The domain details do not, which is why I will say early which parts of your problem I know cold. The rest of the value is the other direction: the blind spots in any industry are shared by everyone in it at once, and the answer that is ordinary somewhere else has usually not been tried here.
The question is not static, which is why it cannot be answered once and filed. It wants someone whose only job is to track it, who watches how firms in several industries are approaching the same problem rather than only yours, and who stays at the edge of the tooling, which being outside makes possible.
The framework behind the validation work is set out in Trust Is the Bottleneck and The Validator Operating Model. The complete version is available as a briefing, by email.
The first conversation is thirty minutes and free. Bring one decision you are not sure how to make. You will leave with an honest read on whether there is anything worth doing.