Fractional Chief AI Officer or fractional AI CTO? Compare what each one delivers.

Hire a fractional Chief AI Officer when your main problem is control: policy, risk, budget and adoption across many teams. Hire a fractional AI CTO when your main problem is delivery: architecture, model choice, evals and systems in production. Many small companies need the second role first. They have no AI system in production to govern.

Both titles are new. Sellers use them in different ways. Some people who call themselves a fractional CAIO write code every week. Some people who call themselves a fractional CTO only give advice. Thus, do not select by title. Select by the deliverables in the contract.

I work as a fractional AI CTO. I also do some governance work for clients. Read this page with that fact in mind.

What does each role own?

The difference is the main audience. A CAIO works mostly with business leaders. An AI CTO works mostly with engineers. Both roles must understand the business and the technology. But each role spends most of its hours in a different place.

What does a fractional Chief AI Officer deliver?

A good fractional CAIO delivers these items in the first months:

• An inventory of AI use. A list of all AI tools, the teams that use them and the data that goes into them. This list includes tools that employees pay for themselves.
• An AI usage policy. Rules for data, approvals and output review. The policy has a fast path for approval. A slow path causes people to use unapproved tools.
• An approved tool list. The tools that the company pays for, with the data that each tool can receive.
• A risk register. The AI risks, with an owner and an action for each risk.
• A roadmap. The business processes where AI can give the most value, in order of priority.
• Training. Sessions for staff and for managers, about the approved tools and the rules.
• Regulatory preparation. For example, the EU AI Act obligations that apply to the company as a deployer of AI systems.

These deliverables are necessary in a company with many teams and many tools. They are not sufficient. A policy does not put a system into production.

What does a fractional AI CTO deliver?

A good fractional AI CTO delivers these items:

• An architecture. The design of the AI system, with the data flow, the model, the tools and the points where a human approves an action.
• A model decision. Which model to use for each task, with the reasons and a plan to change the model later.
• An eval suite. Tests that compare the model output with known good answers. The team runs them before each change.
• A cost model. The cost of each task, not only the monthly bill.
• Monitoring. Traces, alerts and dashboards for the system in production.
• Code reviews and hiring. Reviews of important changes. Help with the selection of AI engineers.
• A system in production. Built by the AI CTO, by your team or by an agency, but reviewed and owned by the AI CTO.

These deliverables put AI into production. They do not control how the whole company uses AI.

How do I decide which role I need?

Answer these questions. Count the “yes” answers in each column.

If most answers point to one column, start with that role. If the answers are equal, start with the AI CTO. The reason is simple. A governance program needs real systems to govern. A policy written before the first system is often wrong. It describes risks that you do not have, and it misses the risks that you do have.

Can one person do both roles?

Yes, in a small company. A company with fewer than 100 people usually has one AI system or a few AI systems. One person can own the architecture and also write the policy for those systems.

Look for these signs that one person can do both:

• The person can show a system that they put into production and operated.
• The person can explain a data protection rule and also a retrieval design.
• The person speaks to your CEO and to your engineers without a translator.

Look for these signs that one person cannot do both:

• The person only shows slides and frameworks. No code, no traces, no evals.
• The person only shows code. They cannot explain the business case or the risk.
• Your company has many departments and a high regulatory risk. The governance work alone can fill the hours.

In a large company, the two roles usually separate. The CAIO owns the program across the company. One or more technical leaders own the systems.

Who should each role report to?

A fractional CAIO should report to the CEO or the COO. The role needs authority over departments. If the CAIO reports to the CIO, the role often becomes an IT project. The CAIO then cannot change how sales or finance use AI.

A fractional AI CTO should report to the CTO, or to the CEO when the company has no CTO. The role needs authority over technical decisions. If the AI CTO reports to a product manager, the role often becomes a feature factory. The AI CTO then cannot stop a feature that has no eval or no cost model.

Write the reporting line in the contract. Write the decisions that the role can make without approval. Also write the decisions that need approval.

What does each role cost?

Prices for fractional AI leadership are not standard. In public videos from 2026, sellers of fractional CAIO services quoted from about 5,000 to 15,000 US dollars per month. Other sellers quoted 8 to 20 hours per month. These are the sellers’ own claims. They do not include the deliverables, so you cannot compare them directly.

Compare the deliverables, not the hourly rate. Ask each candidate:

• What will exist in 30, 60 and 90 days that does not exist today?
• Who builds the systems? You, my team or a third party?
• What do I keep when the engagement stops?

The fractional CAIO cost guide explains how to compare offers. My own engagement shapes are on the rates page.

What goes wrong with each role?

Each problem has the same fix. Write the authority of the role in the contract. Give the role a budget, engineering time or both.

A hypothetical example

A company with 60 employees sells software to logistics companies. The sales team uses three AI writing tools. The support team tested a chatbot, but it is not in production. The product team wants an AI feature that reads shipping documents. The CEO asks: “Do we need a Chief AI Officer?”

Use the questions above:

• Departments that buy tools: two. This is not a strong CAIO signal.
• Regulated industry: no.
• AI in the product: yes. This is an AI CTO signal.
• Pilots that never went into production: yes. This is an AI CTO signal.
• AI production experience on the team: no. This is an AI CTO signal.

This company needs a fractional AI CTO first. The AI CTO puts the document feature into production with evals and a cost model. The AI CTO also writes a short usage policy for the sales tools. After a year, if AI use grows across many departments, the company can add a governance role.

How do I write the contract?

For either role, include these items:

1. Scope in deliverables. Not “AI advice.” Write the documents and systems with dates.
2. Authority. The decisions that the role makes alone.
3. Hours and reply time. A monthly block, and the reply time for urgent problems.
4. Ownership. All code, prompts, evals and documents belong to the company.
5. Access. The person works in your accounts. You can remove the access in one step.
6. Conflicts. The person tells you about each vendor relationship and each referral fee.
7. Exit. The handoff documents and the notice period.

The exit plan guide lists what a good handoff contains.

How do the two roles work together?

Some companies need both roles at the same time. This is common in a company with 200 or more employees and AI in the product. Divide the work clearly, or the two people will compete.

Use this division:

• The CAIO decides where AI goes. The CAIO selects the business processes, sets the priorities and owns the policy.
• The AI CTO decides how AI works. The AI CTO selects the model, designs the system, writes the evals and owns the production quality.
• They share the risk register. The CAIO owns business and legal risks. The AI CTO owns technical risks, for example prompt injection, data exposure and cost overruns.
• They give one report to the CEO. One page each month, with systems, cost, incidents and decisions.

Write one rule in both contracts: a new AI system does not go into production without approval from both people. The CAIO approves the business case and the policy fit. The AI CTO approves the evals, the monitoring and the cost estimate.

This rule prevents two common failures. The first failure is a business team that ships an AI tool without tests. The second failure is an engineering team that ships a correct system that breaks a policy.

Which role is easier to replace later?

Think about the end of the engagement before you start it.

A fractional CAIO leaves documents: a policy, a roadmap, a risk register and training material. An internal manager can usually own these documents after a handoff.

A fractional AI CTO leaves systems: code, evals, monitoring and a cost model. An internal engineer can own these systems only if the engineer worked on them during the engagement. Thus, plan the knowledge transfer from the first month. Put one of your engineers next to the AI CTO in all design reviews.

Next step

Answer the nine questions in the table. If most answers point to the AI CTO column, read how my fractional AI CTO engagement works. If you are not sure that you need either role, read do we need a Chief AI Officer? first.