Serious AI with Chris Duff
September 22, 2026, updated October 23, 2026
A fractional Head of AI is a part-time senior operator who owns two decisions inside a company: where AI gets pointed, and whether what it produces can be trusted before anybody acts on it. My rule of thumb: a company needs one once AI is touching work that carries money, and nobody’s job description says they own that call.
That is the working definition we use at Serious AI Partners, an enterprise AI implementation firm run by Chris Duff, M.D., who holds the seat for operators usually somewhere between $5M and $500M in revenue. Fractional means part of a week rather than part of a job. The seat is senior because the decisions it owns are the expensive ones, and a company that hands them to whoever has capacity that month is not really assigning them at all.
In early 2025, Chris put a web app behind Land Pricer, a previous AI underwriting tool built to price land to the standards of Serious Land Capital (the land investment firm where this method was built, with its own capital at title on every deal). Land Pricer is now in the hands of a new operating team targeting appraisal firms. The web app piece alone ran roughly $25K to $30K out of pocket, took 4 to 5 different engineers in rotation plus a senior product manager, and spent several weeks in iteration cycles where bugs piled up about as fast as they got cleared. Across the whole Land Pricer effort, before the tooling was ready for it, about $100K of Chris’s own money went in.
In May 2026 the same category of web app build, this time for SLC Chat (the firm’s underwriting chatbot, built as a web app with a Stripe paywall), came in at about $160 in subscriptions and API tokens over 3 to 4 days, with one human in the loop. Email authentication, a Stripe paywall, the knowledge base, conversation history, pricing tiers, a landing page. Land Pricer was more complex in some ways, though not by orders of magnitude. Chris wrote up the comparison in the May 18, 2026 issue of Serious News, his weekly newsletter (renamed Serious AI News on September 17, 2026), and put it at roughly 200x compression on cost and about 10x on speed, adding that the curve was “still bending toward more cost and speed savings on a weekly basis.”
Land Pricer came two years before the tooling was ready, and that tuition paid: it forced the pricing methodology into codified form, and it is part of the basis of the business Chris runs now. My read: the build is the part that got cheaper, and deciding which build is worth doing (and when) is the part that hasn’t, which is the work the seat exists for.
That call landed on the person who owned it, and the timing came out of his own pocket. Where nobody owns it, I’d expect the call to land on a CTO whose week is already full, or on nobody (and the bill arrives either way).
Four things, and they are all decisions rather than deliverables.
1. The constraint. Which decision is gating the business right now, and whether AI reaches it at all. The most expensive mistake in this work is automating something that was never the bottleneck. You get a faster version of something that wasn’t holding you back, plus a larger bill.
2. Sequence and timing. What gets built now, what waits for the tooling to arrive, and what should not be built by you at any price. This is the column Land Pricer paid tuition on.
3. The Check. That is the step where somebody validates that what got built actually works, judged against your experience and your data, and sends it back when it does not. The seat owns where that verification sits, what wrong looks like in this specific operation, and who signs. A generic check, the kind that ships inside a tool, was written by somebody who has never seen your business. Verification worth having gets built out of repetitions your team has already done and the things that have cost the business money before.
4. What it costs to run. Model spend, software, and the hours it takes from your people, measured against a baseline taken before anything changed. Skip the baseline and you end up with a working system and no way to prove it earned anything.
A fractional CTO seat overlaps this and is not the same shape. A CTO owns engineering: the stack, the team, the roadmap, uptime. A Head of AI owns whether AI is pointed at the constraint that actually gates the business, and whether anyone can trust the output. Chris holds both titles, and in a smaller company we’d happily see one person carry both.
In our view, the titles overlap enough that the label alone tells a buyer little. Iternal, an AI consultancy that sells fractional Chief AI Officer engagements, frames the fractional CAIO as the board-facing role, reporting to the CEO or board, and a fractional Head of AI as delivery-weighted, reporting to a CTO, CIO or COO.
Our Head of AI seat straddles that line: it makes the strategy call on where AI gets pointed, and it stays close enough to the build to own whether the output can be trusted. It is measured on an operating number that moved against a baseline taken before the work started. My bet is that most companies between $5M and $500M in revenue are better served by one part-time person carrying both halves.
The table puts the common titles side by side and names the source in each row it draws on: Iternal’s guide, or the fractional CAIO and fractional CTO rate benchmarks published by Go Fractional, which matches companies with fractional executives.
| Title | What it owns | Typical time |
|---|---|---|
| Fractional Head of AI (our definition) | Where AI gets pointed, and whether its output can be trusted before anyone acts on it | Part of a week |
| Fractional Chief AI Officer (CAIO) | Enterprise-wide AI strategy, governance, build-vs-buy calls and board reporting, reporting to the CEO or board (Iternal) | Two days a month to three days a week (Iternal), or about 13 hours a week on Go Fractional’s benchmark |
| Fractional AI officer | Used interchangeably with fractional CAIO, most often where one person is the whole AI function (Iternal) | Iternal’s Advisory or Embedded tier, about two days a month to one day a week |
| Fractional CTO | Engineering broadly, with AI as one feature among many (Iternal) | About 15 hours a week on Go Fractional’s benchmark |
| AI consultant | A one-off project, delivered before the consultant departs (Iternal) | Until the project is delivered (Iternal) |
| Full-time Chief AI Officer | The same accountability as the fractional CAIO seat, carried as permanent headcount (Iternal) | Five days a week, at $400K to $1.2M+ all-in (Iternal) |
The Head of AI and CTO seats get a fuller comparison in Fractional Head of AI or Fractional CTO: Which One Does Your Company Actually Need?
The rate benchmarks cited here are published for the fractional Chief AI Officer (CAIO), the neighboring and more board-facing role, so read them as a range for the family of roles rather than a quote for the Head of AI seat. Iternal’s tiers run from about $4K to $30K a month, and its guide says pricing scales with time committed:
Go Fractional’s live benchmark puts the middle 50% of fractional Chief AI Officer postings at $150 to $250 an hour, or $7,800 to $13,000 a month at a typical 13 hours a week. It draws on a small sample of job posts and candidate profiles from the prior 90 days, so treat it as a snapshot of one platform (checked October 23, 2026).
For a full-time hire, Iternal puts the all-in cost of a Chief AI Officer at $400K to $1.2M+. Iternal sells fractional engagements and Go Fractional matches fractional executives with companies, so we think their figures are best read as the market’s own asking prices.
Serious AI does not publish a rate, and the structure gets set after the diagnosis. The reasoning behind that is in What Should a Fractional Head of AI Cost in 2026?
Five conditions.
The AI bill is climbing and nobody can point to an operational number that moved. Hours, cycle time, error rate, throughput, cost per unit. If none of them shifted while spend went up, output capacity rose without throughput rising, which means the verification step never moved.
Something got built that demoed well and quietly stopped being used. Judge it by whether people are still using it a few weeks later, since a good demo is weak evidence on its own.
Nothing has started, because nobody can say where it would help. Nothing has broken yet, which is exactly why it pays to put a number on what waiting costs.
The verification lives in one person’s head, and that person is you. The way I see it, the business grows to the edge of your calendar and then stalls, because judgment doesn’t delegate through a process doc.
AI questions route to your CTO, who already has a full job. What we’d expect: nothing gets refused, everything gets deferred, and the decisions sit in the queue. I think a part-time seat clears that more cleanly than a full-time hire, because the work is a handful of high-consequence calls plus the architecture behind them, rather than 40 hours of anything.
Three cases, and naming them is part of the job.
The constraint is not a decision problem. When it sits somewhere an AI build won’t reach (demand, or a supplier), I’d want the diagnosis to say so, and that’s a perfectly useful answer.
Nobody will give access to how the work actually runs. The judgment being encoded has to come from the people doing the work, which usually takes a few sessions with them and access to the systems involved. There is no substitute for it.
Nobody will commit to a baseline. Measure both sides before anything changes: what the process produces, and what it costs to run all in. Once that measurement is skipped it cannot be recovered.
Start by writing down the constraint, the one thing slowing the business, and how well AI is working on it today (if at all).
Then take the baseline on the process around that decision, before you talk to anybody about building. Hours, cycle time, error rate, and the all-in cost to run it. Those numbers are the only way anyone later proves the work paid.
Then decide who owns the call by name, and put the shape of the engagement after the diagnosis rather than before it. A seat and a scoped build are different work and should not be priced the same way, which is covered in the Serious AI FAQ. Chris wrote up the decision to make enterprise AI implementation the primary business, including the Land Pricer tuition behind it, in the September 14, 2026 issue of the same newsletter: 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.
They overlap, and they sit at different altitudes. In Iternal’s guide the fractional Chief AI Officer (CAIO) reports to the CEO or board and owns AI strategy, governance, build-vs-buy calls and board reporting, while a fractional Head of AI is delivery-weighted and reports to a CTO, CIO or COO. The way we define the seat, a fractional Head of AI owns where AI gets pointed and whether its output can be trusted, so it takes one strategy call and stays close to the build. Our view is that the title matters less than whether those two decisions have an owner.
The rate benchmarks cited here are for the neighboring fractional Chief AI Officer (CAIO) role, so treat them as a range for the family rather than a quote for the Head of AI seat. Iternal publishes fractional CAIO tiers from $4K to $30K a month depending on time committed, and Go Fractional’s live benchmark puts typical fractional CAIO cost at $7,800 to $13,000 a month for about 13 hours a week. Serious AI does not publish its own rate, and it sets the structure after diagnosing the constraint.
We think it depends on how many decisions the seat owns and how much of the build it stays close to. For the neighboring fractional CAIO role, Iternal’s guide puts the range at two days a month to three days a week, and Go Fractional’s benchmark puts typical scope at about 13 hours a week.
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. Chris Duff holds both titles.
No. Iternal’s guide draws the line this way for the CAIO: an AI consultant delivers a one-off project and departs, while a fractional CAIO is embedded, accountable AI leadership. The same holds for the Head of AI seat, which also owns the Check, the step where somebody validates that what got built actually works, judged against your experience and your data.
Iternal’s full-time reference point is a CAIO, five days a week at $400K to $1.2M+ all-in, which its table pairs with companies where AI is the core product. In my view, the same test fits the Head of AI seat: a company whose AI work is a handful of high-consequence calls plus the architecture behind them is better served by a part-time seat.
We would put it at the point where AI is touching work that carries money and nobody’s job description says they own the call. One signal is an AI bill climbing while nobody can point to an operational number that moved.
=====
If you run a business somewhere past $5M in revenue, bring two things to the first conversation: your constraint, and where AI is falling short on it today. AI may already be working on it, just not as effectively or cost-efficiently as it could be, and we have yet to meet a company that is fully optimized with AI. 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.