Serious AI with Chris Duff
October 6, 2026, updated October 12, 2026
A fractional CTO owns engineering, meaning the stack, the team building on it, and whether the software ships and stays up. A fractional Head of AI owns where AI gets pointed and whether its output can be trusted before anyone acts on it, so a company whose software is the bottleneck needs the CTO, and a company whose AI bill is climbing while no operating number moves needs the Head of AI.
That is the line Serious AI Partners, an enterprise AI implementation firm run by Chris Duff, M.D., draws between the two seats. Chris carries both titles for operators usually somewhere between $5M and $500M in revenue, and spent 10+ years directing engineering teams, since 2015, before he ever prompted a model. The example below comes from his own land investment firm, where the method was tested before he offered it to anyone else. If you want the definition of the Head of AI seat on its own, it is written up in What Is a Fractional Head of AI, and When Does a Company Need One?
| Fractional CTO | Fractional Head of AI | |
|---|---|---|
| Owns | The stack, the engineering team, the roadmap and uptime | Which decision AI gets pointed at, and the check on what it produces |
| Decides | How to build it, who builds it, and what to buy versus write | Whether to build it at all, what goes first, and what waits for the tooling |
| Measured on | Shipping on schedule, stability, and engineering cost and velocity | An operating number that moved against a baseline taken before the work started |
| The failure it prevents | Software that breaks, a team that stalls, a stack no one can maintain | A working system pointed at work that was never the constraint, plus a bigger model bill |
| First question it asks | What are we running, and will it hold up as we grow? | What is the constraint, and where is AI falling short on it? |
The seats overlap, and Chris holds both titles. The split is in the question each seat is paid to answer. A CTO is paid to build it well. The Head of AI question comes both before and after the build: whether it should exist at all, whether it went at the right decision, and how anyone would know it worked.
Serious Land Capital is the land investment firm where this method was built, and it had its own capital at title on every deal, so a mispriced parcel was a loss the firm absorbed itself. Underwriting, which means working out what a parcel is actually worth and whether the deal clears at that price, took 45 to 60 minutes per parcel by hand. It was the first thing Chris pointed AI at, and where AI gets pointed is the first of the two decisions the Head of AI seat owns. It was the highest-value hour in the business and the one where being wrong cost the most.
The first fully measured automated run, on August 18, 2026, took 3.5 hours and used 10 to 15% of a week’s model allowance. It produced the correct answer, so judged on whether the software worked, the run passed. Judged on cost, the economics were unsustainable, and judging what it costs to run is one of the four things the Head of AI seat owns.
We read the fix as mostly engineering work, the kind a CTO seat exists for. The procedure became software, 40,190 lines of it with 10,036 of those being tests, so the deterministic steps stopped being model calls at all. What was left got matched to the model each task needs, with cheaper models on extraction and formatting and the top model held back for judgment and the written read. Underwriting now runs in under 30 minutes unattended at about 1% of a week’s allowance.
The check on that output is the Head of AI seat again. On a test deal the system disagreed with the manual read by roughly $19K. The cause surfaced the same day, and the re-run landed at $87,992 against a hand-built $88,000. In our view, deciding what “wrong” looks like on a land deal, and where the check that catches it sits, is Head of AI work.
In our read, that is one project answering two different questions. The full before and after is on the Serious AI Proof page.
Match the seat to the situation you are in today.
In our view, yes, and Chris lists both titles on his About page. In our read, the risk is that the engineering half is louder: a failing build tends to announce itself the same day, while a system pointed at the wrong decision can run smoothly for months, and the bill arrives with no line item explaining it. Our read is that in a shared seat, the CTO questions crowd the week and the Head of AI questions get pushed to next quarter.
Judged only on “does it work,” the August 18 run passed, with a correct answer and unsustainable economics. If one person holds both seats, put the Head of AI question on the calendar first and write the baseline down before the build starts.
Our view is no, because the two seats are not the same shape of work. A CTO seat owns engineering, meaning the stack, the team, the roadmap and uptime. A Head of AI seat is a handful of high-consequence calls plus the architecture that follows them. The Serious AI FAQ covers how a seat and a scoped build get structured, and the short version is that the shape follows the diagnosis. Be wary of anyone who quotes either seat before understanding which question you need answered.
Write down the one thing slowing the business right now. If it is software (releases, stability, a team without a senior lead), start the CTO search. If it is a constraint AI is not helping with as much as it could, or AI work with nobody checking what it produces, start with the Head of AI.
Either way, take the baseline before you talk to anyone. Hours, cycle time, error rate and the all-in cost to run the process you care about. Then ask every candidate for either seat the same two questions: what would you measure in week one, and what result would tell you to stop? A CTO candidate should answer in engineering terms. A Head of AI candidate should name a business number and a check.
Chris wrote up the decision to make enterprise AI the primary business, and why he now calls himself an AI workflow engineer and fractional CTO, in the September 14, 2026 issue of his weekly newsletter, since renamed Serious AI News and pointed at enterprise AI operators: I’d Already Made This Decision (Two People Noticed First).
AI can produce the analysis. It cannot be wrong, because being wrong requires having something at stake, and that part stays with a person. Serious means underwritten.
=====
If you run a ~$5M+ business and cannot tell whether your next AI hire should be an engineer or someone who decides where AI goes, the work starts with diagnosing the constraint that is gating you, then building with verification sitting where a person still owns the outcome. You leave the first call knowing what your constraint is whether or not you ever hire us.
Fill out a short form, and we read every answer ourselves. If there is a fit we will reach out and find a time that works for you, and if there is not, we will say so and point you toward someone better suited. Either way, the moat question is worth answering now, in a market AI has already reset.