Fractional CTO, CISO and Head of AI

AI has changed what senior technology roles spend their week on

A senior technology role used to be mostly throughput. The CTO wrote the architecture documents, the security lead read the alerts, the head of data wrote the queries behind every report. Most of the week went on producing things, and a smaller part on deciding what they meant.

AI tools now produce much of that first draft. They summarise an incident log, draft a policy against a standard, write the migration script and flag the alert that looks different from yesterday’s. The output still needs someone who knows when it is wrong.

For drafting work the gain is measurable. In a preregistered experiment published in Science in 2023, economists at MIT gave professionals writing tasks taken from their own occupations. Those given an AI assistant took 40% less time, and graders rated their work higher.

The gain is not automatic. In a 2025 randomised trial by METR, a non-profit research group, experienced open-source developers working in code they knew well took 19% longer when allowed AI tools, and believed they had been faster. In our view, speed comes from knowing which work to hand to the tools and checking what comes back. That is the senior part of the job.

Our view: the routine work in these roles shrinks with the right tools, and the judgement does not. Much of what filled the week is now done by tools, so a senior person who uses them well can often hold the role without being there full-time.

Four parts of the job stay human

Decisions. Which architecture, which vendor, which risk to accept. A model can set out the options. Someone has to choose and live with the choice.
Accountability. A named person who answers for the system to your board, your customers and, where they apply, your auditors and regulators.
The board conversation. Explaining risk and trade-offs to people who are not engineers, and being trusted when you do.
The 11 pm call. When something breaks, someone who knows the system picks up the phone and decides what happens next.

These are the reasons you hire the role, and they rarely fill a full working week. A fractional role gives you that judgement from someone who has carried the role before, with the routine work done alongside.

Six roles, sized to the work

Six senior roles we fill part-time, each with a named senior person and a written plan for the first 90 days.

Fractional CTO

Owns: Technical direction: the architecture, the engineering roadmap, hiring plans, and vendor and build-or-buy decisions.
Signs you need one: You are raising money or selling to larger customers, and nobody can answer the technical questions with authority. Architecture is being decided by default. A CTO search is under way and the decisions will not wait for it.
What AI now takes on: First drafts of architecture documents and decision records, code-review summaries, vendor comparisons and answers to technical questionnaires.
What stays human: Which architecture to commit to, who to hire, which trade-off to accept, and standing behind those choices with investors and the team.
First 90 days: A written view of the current system and its risks; an architecture and roadmap the team agrees to; a hiring plan; decision records for the calls already made.

Fractional head of AI

Owns: Your AI systems once they are in use: which models run, the evaluations that show they work, their cost, the data they may see, and what happens when an answer is wrong.
Signs you need one: Pilots that worked in a demo and stalled after it. Staff using AI tools with no shared rule on company data. A model-based feature in production that nobody is named to run. A board asking for an AI plan.
What AI now takes on: Prototypes, evaluation runs, side-by-side tests of models on your own tasks, first drafts of usage policies, and weekly cost and quality reports.
What stays human: Whether a feature is ready to ship, how much risk is acceptable, what data a model may see, and the call when it misbehaves.
First 90 days: An inventory of the AI already in use; a usage and data policy; an evaluation for each model-based feature; one pilot taken to production, or stopped with a written reason.

Fractional CISO

Owns: The security programme: risk assessment, policies, incident response, supplier security, and reporting risk to leadership and the board.
Signs you need one: Enterprise customers send security questionnaires your team cannot answer with confidence. A customer or investor asks for SOC 2 or ISO 27001. Nobody owns incident response. You hold personal, financial or health data, and no one is named as responsible for protecting it.
What AI now takes on: Alert triage and log review, first drafts of policies mapped to a standard, questionnaire answers drawn from existing evidence, and summaries of vulnerability scans.
What stays human: Which risks to accept, how to respond during an incident, what to disclose and when, and the conversation with the board, customers and auditors.
First 90 days: A risk assessment; an incident response plan, rehearsed once; a gap assessment against the standard your customers ask for; a one-page risk summary for the board.

Fractional VP of engineering

Owns: Delivery: how the team plans, ships and reviews its work, how it hires and onboards, and whether commitments are met.
Signs you need one: The team has grown past what one lead can manage. Releases slip and nobody can say why. Hiring is under way with no agreed bar. AI tools have raised the volume of code faster than the team can review it.
What AI now takes on: Status reports, release notes, first-pass code review, onboarding material and delivery metrics.
What stays human: Team structure, performance conversations, the hiring bar, and the decision to cut scope or move a date.
First 90 days: A delivery process the team agrees to; a review standard for AI-written code; a hiring plan; a short set of delivery measures reported each month.

Fractional chief architect and technical due diligence

Owns: The shape of the system across teams: platform decisions, technical-debt priorities and migrations. Or an independent view of a codebase before you buy it, invest in it or rebuild it.
Signs you need one: A large migration or replatforming is planned. Teams are making design decisions that conflict. You are acquiring or investing in a company and need to know what its technology will cost to own.
What AI now takes on: Code and dependency scanning, first-pass diagrams of an unfamiliar system, documentation drafts and summaries of large codebases.
What stays human: Which migration path to take and in what order, which risks matter to the deal, and putting a name to the assessment.
First 90 days: A map of what exists; a target architecture and the order of the migration; decision records. For due diligence: a written assessment of the technology, the team and the cost of remediation, inside the agreed window.

Fractional head of data

Owns: How data is collected, governed and used: pipelines, the warehouse, reporting, data quality, and which numbers the business treats as true.
Signs you need one: Two dashboards give different answers to the same question. Analysts spend more time fixing data than using it. An AI project is waiting on data nobody trusts.
What AI now takes on: Writing and tuning queries, documenting tables, first drafts of pipelines and data-quality checks.
What stays human: Which metrics the business runs on, who may see which data, and what to build first.
First 90 days: A map of where the key numbers come from; tests and freshness checks on the ones that matter; agreed metric definitions; a data access policy.

Every engagement starts with one piece of work

Each engagement is sized to what your company needs, from a few days a month to several days a week, and agreed up front, out-of-hours cover included. Whichever role it is, the first step is bounded: one piece of work, one named result, and a clear decision at the end of it.

A technical call. A senior engineer, not a salesperson, talks through where you are and which role, if any, fits.
A first piece of work. Usually the role’s first month: an assessment, a plan, or one system taken to production, with a fixed scope.
A decision. Keep the role, resize it, hand it to someone you hire, or stop.

You work with one named senior person. Behind them is a bench of around 150 engineers, so the work does not rest on one person’s experience. When the right answer is a permanent hire, we say so.

How we work with AI

We use AI to move faster, and hold what it produces to the same standard as our own work.

A senior engineer reviews everything AI produces before it reaches you.
You always know who is accountable: a named person, never a tool.

Tell us which role you are missing.

Agnizar builds custom software and AI, and fills the senior roles around it part-time. Tell us where you are; a senior engineer replies within one business day, not a salesperson.

Start with a technical call