A general manager of a 60-person distribution business in the Waikato has three AI tools on the company card, a sales team using ChatGPT on their own accounts, a board asking what the AI plan is, and nobody whose job it is to answer. The IT provider looks after the laptops and the Microsoft licences. The ops manager is interested but already full. Hiring a Chief AI Officer at executive salary for a business that size would be absurd.
That gap, between needing someone to own AI and being unable to justify a full-time executive for it, is what the fractional Chief AI Officer role exists to fill. The title gets thrown around loosely, so this guide sets out what the role involves week to week, how it compares with the alternatives, when a business is ready for one, and how to tell whether it is working.
What is a fractional Chief AI Officer?
A fractional Chief AI Officer is a senior person who owns a business's AI direction for a set number of days a month, on an ongoing basis. They set the priorities, pick what gets built and what does not, hold the vendors to account, and report to the owner or board on results. The business gets executive-level ownership of AI without the cost of a full-time executive.
The operative word is ownership, the way a part-time CFO owns the numbers. They sit in the leadership meeting, and when a vendor quote arrives, it goes to them.
The ownership gap is wide. Datacom's 2026 State of AI Index found 91% of New Zealand organisations now use some form of AI, but only 15% have scaled it across the organisation. Only 22% have a dedicated AI leadership role at all, and 13% have a Chief AI Officer (Datacom State of AI Index 2026, via Geekzone). For most NZ SMEs, AI is everyone's side project and nobody's job. We covered what that looks like in practice in NZ's AI strategy gap.
What does a fractional CAIO do week to week?
Less technology than people expect, and more operations. The detail changes by business, but a month runs roughly like this.
Week one: the leadership session. Half a day with the owner or leadership team. What shipped, what the numbers say, what needs a decision. Every decision is recorded with its reason.
Week two: time with the work. Sitting with the people whose processes are changing, reading transcripts from any AI system already running, and finding the next piece of friction worth removing. Most of the value comes from here.
Week three: builds and vendors. Reviewing anything in flight, checking quotes, and saying no to the tool somebody saw at a conference.
Week four: governance and reporting. Updating the AI use policy, checking Privacy Act 2020 obligations against what is running, and writing a one-page report the board can read in five minutes.
We describe AI as an operations problem more than a technology project, and this rhythm is why. Our founder spent 16 years at Connect NZ scaling device repair from around 200 jobs a year to more than 27,000, on a custom FileMaker ERP with machine-learning assessment tooling built into it. None of that growth came from the technology on its own. It came from someone owning the process and deciding, every month, what to change next.
How is a fractional CAIO different from a consultant, a hire or an IT provider?
By what each one is accountable for, and for how long. A consultant owns a deliverable. An IT provider owns uptime. A full-time hire owns the whole function, at full-time cost. A fractional CAIO owns AI outcomes on an ongoing basis, for a fraction of the time and cost.
| Fractional CAIO | Full-time AI executive | Project consultant | IT provider | |
|---|---|---|---|---|
| Accountable for | AI outcomes, ongoing | AI outcomes, ongoing | A defined deliverable | Systems uptime and security |
| Time commitment | A few days a month | Full-time | Intensive, then finished | As tickets arise |
| Cost shape | Monthly retainer | Executive salary and on-costs | Fixed project fee | Monthly support contract |
| Sits in leadership meetings | Yes | Yes | Rarely | Rarely |
| Says no to vendors | Yes | Yes | Only within scope | Not their remit |
| Best for | 20 to 200 staff, AI in several places | Large firms with AI at the core | One clear problem to solve | Keeping existing systems running |
| Main risk | Too few days to go deep | Cost, and a long hiring search | Hands over, then leaves | Treats AI as another software install |
Is a fractional CAIO the same as an AI consultant?
No, although the same person can do both. A consultant engagement has a start, a scope and an end. You get a report or a working system, and the relationship finishes. That suits a business with one clear problem, and we set out how to choose one in choosing an AI consultant in NZ.
A fractional CAIO stays. They see the second and third projects, notice when the first one drifts, and carry the context from one decision to the next.
Why not leave AI with the IT provider?
IT providers are set up to keep systems running, and AI work changes how the business operates. A managed service provider is excellent at licences, security and support tickets. Deciding which customer conversation to automate is a different skill. Keep them close, though. Their knowledge of your systems saves a fractional CAIO weeks.
When is a business ready for a fractional CAIO?
When AI has become several things happening at once and nobody owns the whole picture. The signs are consistent: staff using unsanctioned tools, more than one AI project or vendor in play, a board or owner asking for a plan, and a leadership team with no one who has time to write it.
Size matters less than complexity. A 25-person business with AI in sales, operations and customer service can need one. A 150-person business with a single, well-run AI phone agent may not.
When is a business not ready?
When there is one clear problem and no appetite yet for more. If the need is "answer the phone after hours" or "stop re-keying invoices", a fixed-scope project does the job for less.
It is also too early if leadership will not give the role access. A fractional CAIO who cannot see the numbers or talk to the staff doing the work becomes an expensive adviser.
Five signs you are ready for a fractional CAIO
1. AI is already in the building.
Staff use AI tools, sanctioned or not. You have at least one AI system or vendor in place.
2. More than one idea is competing for attention.
Sales wants one thing, operations wants another, and nobody is ranking them.
3. Someone is asking for a plan.
The board, the bank, a major client or the owner wants to know what the AI strategy is.
4. Nobody has the time to own it.
The obvious internal candidate is already full, and AI keeps sliding to the bottom of their list.
5. You can give the role a seat.
Leadership will share numbers, give access to teams and act on recommendations.
What do the first 90 days look like?
Map, pilot, run. The first month is spent understanding the business and deciding what matters. The second month puts one thing into production. The third month proves it works and sets the rhythm for everything after. By day 90 the business should have a written plan, one live system and a monthly reporting habit.
Days 1 to 30: map
An inventory of every AI tool in use, including the ones on personal accounts. Interviews with the people doing the core work. A draft AI use policy. The output is a ranked shortlist of three to five opportunities, each with a rough cost and return, plus a list of things deliberately left out. It is the discovery step from our three-phase model, with one difference: the person who wrote the plan stays to deliver it.
Days 31 to 60: pilot
The top opportunity gets built, bought or configured. The fractional CAIO does not have to write the code. They own the specification, choose who builds it and set the escalation rules. We wrote about why the handoff to a person matters so much in the escalation experience.
Days 61 to 90: run
The pilot goes live, gets measured against its baseline and gets tuned. The monthly report starts and the second opportunity is scoped. By now there is a live system and a set of numbers to judge the role against.
How do you measure whether a fractional CAIO is working?
Three numbers, reviewed every quarter: hours returned to the business, decisions made because of the work, and systems live and in use. If all three are moving after two quarters, the role is earning its keep. If the only output is documents, it is not.
Hours back is the easiest to measure and the one people over-weight. Track it against a baseline taken from real volumes before each system went live.
Decisions made tells you whether the role has any influence. Projects started or killed, vendors dropped, a service added because of what an AI system heard. We made the case for reading AI output as intelligence in AI as an intelligence sensor.
Systems live means in production and in use. A system that launched and got quietly bypassed counts as zero.
What are the red flags when choosing a fractional CAIO?
The biggest is a candidate who leads with tools. The strong candidates ask about your margins, your bottlenecks and your staff before they mention a model or platform.
Watch for these as well:
- No production history. Ask what they have shipped that customers or staff use today. Demos and workshops do not count.
- An undisclosed reseller relationship. Someone paid commission by a platform will recommend that platform. Ask directly.
- No written measures. If they cannot say how their work will be judged at 90 days, you cannot judge it either.
- Nothing they would leave alone. Ask what they would advise you not to automate. Anyone who has done the work has a list.
- Lock-in by design. Prompts, configurations and documentation should belong to the business. We covered why in the Fable 5 shutdown and your AI vendor risk.
What does a fractional CAIO engagement look like?
Most engagements take one of three shapes. A light retainer of one or two days a month suits a business with one or two systems running that needs oversight and a roadmap. A standard retainer of three to five days a month suits a business actively building. A build-heavy arrangement pairs the retainer with fixed-price projects for anything substantial.
Keep the retainer separate from the build budget. The retainer pays for judgement and continuity. Builds should be quoted as fixed-price projects with named line items. The market ranges for those builds are in our guide to AI implementation costs for NZ SMEs: single-workflow builds from $5,000 to $25,000, advisory-only engagements from $12,000 to $35,000, and running costs of $100 to $300 a month for one or two agents.
Government funding can cover part of the start. The MBIE AI Advisory Pilot co-funds up to $15,000 towards an AI roadmap and putting it in place, delivered through the Regional Business Partner Network. It was extended in September 2026 to run until 30 June 2027 for a further 120 businesses (Beehive). The mapping month is the part it fits best.
Plan on two quarters as a minimum. One quarter maps and pilots. The second shows whether it lasts.
Where do you start?
Write down every AI tool your staff use, every AI idea on the table and who is responsible for each. If the last column is mostly blank, you have an ownership gap.
Language became programmable, so the calls, emails and forms a business runs on can now be read and acted on by software. Somebody has to decide which of them are worth it. See our services for how we approach AI strategy and delivery, read about us, or get in touch if you want to talk through whether your business needs a fractional CAIO or a smaller piece of work.