Who is actually looking after AI in your business?

Many of the business owners we speak to are in a somewhat challenging and nuanced situation between buying AI, seeing real return from it, and avoiding introducing new and unexpected risks to their business. We see this manifesting mostly as being slow to adopt AI or to realise profit level impact from their efforts.

Think about the last AI tool someone suggested your business should use. Perhaps it promised to write quotes or follow up on enquiries from customers. Before you can decide whether it will truly be impactful, avoid risk and be worth buying, what you really need to work out is what information it would need, whether you can trust it with that information, and who is able to monitor and check that the work it is doing is hitting the bottom line.

Who do those questions land with?

If it was you, the business owner, alongside everything else you already do then the tool or service has given you another job to add to your load before it has taken anything away.

That is where the conversation about a fractional AI officer should begin. Businesses that don’t have the expertise to get the right inputs for important decisions tend to defer them. Or worse, make bad decisions that cost money rather than create value.

Curiosity becomes a responsibility

We have an AI readiness assessment that asks a fairly straightforward question: who is responsible for AI in your business?

The answers range from nobody in particular to a person with defined scope and authority. Between those sits the mean answer we see in our discussions with businesses which is whoever has the time, the curiosity or enough confidence to try something new.

Curiosity is a good place to start and we love to work with these people. Curiosity is by its nature optimistic and gets people experimenting and asking questions that are important to ask. But there is a point where the experiment starts using customer information, influencing a decision or becoming something the team relies on and at that point the responsibility has changed. The business rules change, even if nobody has changed the arrangement around it.

Being comfortable using AI does not necessarily mean being equipped to make the right decisions for a business. Someone needs to understand its capabilities and limitations, recognise when a demonstration leaves out the difficult part, and know what evidence would justify trusting it with real work.

Australia’s National AI Centre makes this distinction in its guidance on AI adoption: the person overseeing AI needs sufficient authority and understanding of its capabilities and risks. Giving someone the responsibility without the knowledge, time or support leaves the business exposed.

And this is an unusual field to add to someone’s existing workload. What is possible changes quickly. At the start of the year the leading edge of developers started using OpenClaw, a powerful open-source agent that requires significant technical knowledge to operate. Now, Grok Bot and Meta’s Muse do the same for non-technical people, at a fraction of the price. Understanding the costs, the available products and the choices about what to build yourself or buy from somebody else are all things that themselves require a level of understanding of the AI landscape to be able to make good decisions. Following those changes is work in its own right. It is, we think, too much to have this be a secondary job for someone.

The result you want still needs choosing

Businesses want the enquiry followed up, the quote prepared and the customer looked after. They may be interested in how AI does those things, but buying and supervising more software is rarely the outcome they came looking for.

In his March 2026 article, “Services: The New Software”, Julien Bek described a shift towards buying completed work rather than tools to help people perform it. This is a useful way to understand what is changing. AI is expanding the choices businesses have about how work gets delivered. When you hire a person, you hire them to get a job done. AI might work more like this than like buying a tool someone in the team already is responsible for operating.

A provider being able to deliver an outcome like this does not tell you whether it is the outcome your business should pursue first.

Faster quoting might be useful. It might also produce more quotes that still wait for the same person to check them. Automated follow-up might help, unless the missing piece is knowing what was promised in the conversation before it.

This is why we start with the map. We look at the work, the decisions, the information and the places where everything keeps returning so we can identify the highest leverage points for the business to see real gain. The purpose is to establish which change would be useful and safe to make first.

The economic promise of AI becomes real through those choices. The Productivity Commission has argued that Australia’s gains will require businesses to change products, services and core work, rather than simply add AI to existing routines. We agree completely and have written about it here. Inside an individual business, somebody has to work out what that means.

Information can be written down and still be unusable

Think about a customer record that says a quote was accepted. An experienced person would know that the customer agreed only after a conversation about timing, that one item was excluded, and that a particular promise must be checked before the job starts.

A customer record often holds the transaction. The person holds the information that makes it safe to act on.

Here is where risk gets introduced. Ask an AI to prepare the next step from that record. It may produce something entirely plausible while missing the information that matters most (that is, the information the human knows that is not captured in the record).

Making information usable requires someone who understands how AI finds, interprets and applies it. Putting documents in one place is only part of the job. Someone must establish which source is authoritative, what is missing, what has gone out of date, who may access it and when the system should stop and ask a person.

This is one of the reasons practical AI expertise matters. A business owner should not have to become a specialist in how AI consumes information to know whether their customer history is fit for a proposed use.

They do need someone who can assess it properly and explain the gaps in terms the business can act on.

What would a Fractional AI officer actually own?

A Fractional AI officer provides ongoing specialist leadership that the business person responsible for operations, efficiency or growth needs as part of the working week or through an agreed engagement. The role connects what AI can do with what the business needs and helps leadership make and review the decisions that follow.

Our assessment gives this responsibility a practical shape:

Someone owns it. There is informed direction, a clear remit and access to people who can make decisions. Keeping current means judging what developments matter to the business, including which ones can be ignored.

The information is usable. The business knows what a proposed use requires and whether its information can support it. Gaps in context, quality, access and maintenance become work to resolve, rather than surprises after something goes wrong.

The work has changed. The investment produces a difference people can recognise and the business can measure. That includes checking whether time saved in one task has simply reappeared as checking, correcting or chasing somewhere else.

The officer does not need to build every solution personally. They do need to keep advice connected to delivery, make responsibilities clear and ensure someone checks whether the result justified the decision. The business always continues to retain its own accountability; bringing in expertise doesn’t (and shouldn’t!) hand that away.

Advice can’t stop at the recommendation

A recommendation reflects what is known at the time. Once something starts running, the business learns where the information was incomplete, which exceptions occur and what the team actually needs. That experience should change the next decision.

This is also where consulting has to prove its usefulness. A report can identify an opportunity. The value depends on what happens after somebody reads it: whether the change is made, whether people use it and whether it improves the work. This needs to be measured, not gut-checked.

At Catalyst, we are a consulting business. Understanding the business comes before recommending what to put into it. We map that business so we can make that one concrete recommendation about what the first piece to change should be. We don’t start with a tool or product and work backwards to the business. We start with the business and work forward to what will drive impact or profit.

A fractional AI officer addresses the continuing leadership need around those decisions. For a business considering that role, the question is whether it needs ongoing specialist oversight or a clearly bounded piece of help. A single project does not automatically require another executive title.

Who has the time to do this properly?

Go back to the person who received all the questions about that AI tool. Perhaps they are capable of answering them, have the time and capacity, and enjoy it. The question is whether the business has given them the expertise, authority and time to do it properly, including the responsibility for what happens afterwards.

If it has, there may be no need to bring someone else in. But if it has not, a fractional arrangement is one way to give the business access to experienced AI leadership without making a full-time appointment. What matters is that somebody can help choose the right work, establish what it needs and stay close enough to know whether it helped.

That is the conversation we have at Catalyst. Start with what the business is trying to achieve and what is getting in the way. Work out which outcome should come first, then decide what support is needed to deliver it.

We’d love to have that conversation with you. Talk to us about where to begin.