Fractional CTO vs agency vs full-time AI hire — buyer-side math
If your board just asked about AI strategy and you have 5-50 people, the highest-EV next move is a fractional senior — assuming you do diligence. The math on what each of the three paths costs you when it doesn't work.
- positioning
- fractional-cto
- buyer-guide
- ai-strategy
- team-stage
If your board just asked about your AI strategy and you have between five and fifty people, the highest-EV next move is hiring a fractional senior, not an agency and not a full-time AI engineer — assuming you've actually done diligence on the fractional. That's the take. The argument is the math on what each path costs you when it doesn't work, not what it pays when it does.
Most buyer-side writeups of this question are written by someone selling one of the three options. I'm one of those someones — I sell fractional engagements. So write off the conclusion if you'd rather, but the math underneath stands without me.
The three paths, in dollars and months
Full-time AI engineer hire — $200-300K fully loaded, 6-9 month ramp
What you're actually buying: a senior brain pointed at one team, full-time, with the ability to compound knowledge of your codebase over years. The math: $180-220K base in the EU/US, plus 30-50% in benefits, equity, recruiter fees, and onboarding cost. Six to nine months from the day they sign to the day they're shipping at full velocity.
What it costs to be wrong: roughly $150K, eighteen months of opportunity cost, and a hole in the org chart while you re-recruit. The cost to be wrong is the highest of the three options. The cost to be right is also the highest payoff — a senior AI engineer at a healthy 5-50 person company is a 10x compound over three years.
The trade you're making by picking this path: you're betting your AI problem is well-defined enough today that you can write a job spec for the right person, and that the person will still be solving the right problem in twelve months. If your AI problem is "we have a working feature that needs to scale," that bet is fine. If your AI problem is "we don't yet know what to build," you're paying for senior brain at junior output — most of their first year is going to be spent figuring out the problem, which is not what their salary is priced for.
Agency build — $40-150K project, three to four months, then a handoff
What you're actually buying: a team of mid-level builders coordinated by a project manager, working from a Figma against a fixed scope, delivering a thing that demos brilliantly. The math: $40K for a four-week scoped MVP at the cheap end, $150K for a twelve-week full build at the high end. Internal cost on top: a product manager from your side spending 30-50% of their time on coordination, plus three to four review cycles where the founders have to look at slide decks.
What it costs to be wrong: the headline number plus the ninety-day post-handoff window where your team can't maintain what they delivered. The classic agency failure mode — and the one that brings buyers to my door — looks like this: the agency shipped a working demo, your team can't read the codebase, the AI part wraps a different model than the one your engineer knows, and three months in, you're paying the agency a retainer to keep the thing running. Now you've spent $40K plus a $5K/mo extension, and you're still dependent on someone who doesn't know your business.
The trade: you're betting that "shipped on time" matters more than "owned by your team." For a one-off marketing site or a feature that lives at the edge of your stack, that bet is fine. For an AI feature that touches your core product, the bet is bad. AI features mutate every two months — model versions change, prompts need re-tuning, evals need re-running. An agency-built AI feature is a feature you'll need to keep paying someone to maintain. The agency is the natural answer to that — but it's the most expensive answer over a 24-month horizon.
Fractional senior — $4-10K/mo retainer, four-week ramp, your stack from day one
What you're actually buying: a senior engineer working two or three days a week directly in your codebase, your stack, your repo, your CI. The math: $5-7K/mo at the median, two to three days/week, no benefits, no equity, no recruiter, four-week notice to terminate. If they're worth what they charge, they ship architecture decisions plus production code in your stack. If they're not, you've lost a few thousand instead of a few hundred thousand.
What it costs to be wrong: roughly one to two months of retainer, recoverable. The blast radius is the smallest of the three options. The blast radius is the whole point.
The trade: you're betting that part-time senior beats either full-time-junior or full-time-mid-via-agency for your specific problem shape. That bet is right when your problem is "we need to ship something real, fast, in our stack, and we don't have the in-house seniority to make the architectural calls." It's wrong when your problem is "we already have a senior in-house and we just need hands to execute" — in which case you should hire a contractor or an agency, not a fractional senior whose value is the architectural judgment, not the typing.
The dimensions that flip the answer
The argument so far makes fractional senior look like the obvious pick, and at the 5-50 person stage it usually is. But the answer flips on four specific dimensions. Find your team in the table.
| Dimension | Fractional wins when… | Full-time hire wins when… | Agency wins when… |
|---|---|---|---|
| Speed to production | You need a working thing in 4-8 weeks | You can wait 6-9 months for full ramp | You need a fixed-scope deliverable in 3-4 months |
| Team's AI maturity | You have engineers but no AI seniority | You have AI seniority and need scale | You have neither and won't for 12 months |
| Problem clarity | Problem is real but solution is fuzzy | Problem AND solution are well-defined | Problem is a marketing site or one-off feature |
| Regulatory friction | Standard SaaS, light compliance | HIPAA / fully-regulated finance / defense | Low — agency won't carry compliance debt |
The dimension that wins for your specific case depends on which one is the binding constraint. For the median 5-50 person B2B SaaS team I see, the binding constraint is problem clarity — they know they need AI somewhere, they've tried prompts in a Notion doc, they don't yet know which feature to ship first or how to architect it. Fractional senior wins on that dimension because the first month of the engagement is exactly the architecture-clarification work, and you don't have to commit a $300K-a-year hire to it.
If your binding constraint is speed and you can write the spec yourself, agency is fine. If your binding constraint is regulatory friction and you can wait, full-time hire is the only correct answer because regulatory work needs context that doesn't transfer cleanly across vendors. Pick the dimension that matters most to you, then read the table.
Diligence questions for any of the three
Before you sign anything, ask the candidate or vendor these questions. They sort the real ones from the resume engineers and the demo-ware shops.
The point of these questions is not to gotcha anyone. It's to filter for people who think about their own work the way you'd want them to think about yours.
The case where I'd tell you to pick agency anyway
Three scenarios where I think agency beats fractional, and I'd refer the work out:
First, scoped marketing site or microsite. If what you need is a content site, a launch landing page, a one-off interactive demo for a conference — pick an agency. The work is fixed-scope, the AI part is shallow, the handoff to your team doesn't matter because the artifact is going to live or die in a six-month window. Agencies are good at this. Fractional seniors are overpriced for it.
Second, internal-only tool with a known design. If your CFO already mocked up the dashboard, the data pipeline is well-understood, and the AI part is one summarization call wrapped around your existing data — a contractor or small agency on a fixed bid will ship faster and cheaper than a senior engineer who'll spend a week asking "but why this way?" Sometimes the answer is "because the CFO drew it."
Third, you've already tried fractional and the engineer didn't fit. Sometimes the chemistry is wrong. Sometimes the problem really is "we need bodies, not architects." Cut your losses and pick a different shape.
What this means for your next 30 days
If you're at a 5-50 person team and your board asked about AI in the last quarter, here's the order I'd run the next month:
The trap most teams fall into is committing to a vendor type before clarifying the problem. Agencies are built to sell you a solution; full-time recruiting is built to fill a specific job spec; fractional engagements are built to compress the figure-it-out phase. If you don't yet know what you need, the right hire shape is the one that helps you find out, not the one that asks you to define it upfront.
If your team is in the median case I described — board asking, problem fuzzy, runway tight enough that being wrong about a $300K hire is fatal — the Audit Sprint is designed for exactly the diligence step in week one. Fixed price, one week, written brief, working prototype of one decision. Whether you hire me after or hire someone else, you'll have written down your own problem in plain language. That's the work most teams skip and pay for later.
// next move
Want a written architecture brief on your AI stack?
1 week, $1,500, fixed scope. Working prototype of one change in your stack — yours to keep regardless.
// related essays
- 82% → 96%field-level accuracy · 3 months
Bridge Sourcing: how I moved scrape accuracy from 82% to 96%
The 8 specific changes that moved a production B2B sourcing extraction pipeline from 82% to 96% field-level accuracy over three months — and the 2 changes that made it worse. The economic moat under the pipeline is Egypt's 0% EU import-tariff lane; the engineering moat is the calibration set. Calibration sets, schema forcing, deterministic validation, drift detection.
~10 min · 2026-04-30 - 2arbitrages stacked · time zone + senior pricing
Egypt-to-EU senior AI engineering — the 2026 thesis, not the 2024 cheap-outsourcing pitch
Two arbitrages stacked on one geography — time zone and senior-tier pricing. The 2024 framing of "outsource cheap engineering to Egypt" destroys both. The 2026 framing is single-engineer coverage of EU mornings AND US East afternoons at 60-70% of London rates, IF you filter correctly.
~8 min · 2026-04-30