What a fractional CTO is actually accountable for, the warning signs that data is not ready for AI, and where Microsoft Fabric fits.
When a client brings me in as a fractional CTO, they are rarely short of ideas. They usually have more of them than they can fund, a strategy they want validated, or a period of investment ahead that they are not sure how to approach. What they are asking me to take responsibility for is the sequence: what to do first, what can wait, and what to stop doing altogether.
A lot of their questions are: how do we effectively use the technologies we have at our disposal? What do we have to do to use those technologies safely? And what are the foundational elements that give us the most return on investment? Often they have all of those questions swirling around and no sense of where to start. Our job is to come in and help make sense of that.
Every voice heard, then a decision
In the private wealth organisation where I currently work as fractional CTO, different teams have very different views about what needs to happen. Some are moving quickly. Some have barely started. Introducing a new foundation and set of technologies into that environment means listening to all of it before you sequence anything.
You will not get everyone to agree. What you can do is make sure every team has been heard, show them where their concern landed in the plan, and explain why one piece of work comes before another. People need to see visible progress, and they need to see that the strategy is holding firm while it happens. Both matter. Progress without a strategy is noise. A strategy without progress loses the room.
Start with the business problem
Investment discussions start with the value stream, not the technology. What are we trying to improve? What is preventing it today? What would be different if we fixed it? Once that is clear, the steps along the way tend to reveal themselves, and so do the activities that are consuming effort without moving us towards the objective. Those are the candidates to stop.
Whenever you start with technology, you invariably have a solution that’s looking for a problem.
The pressure to adopt AI makes this harder, because it tempts organisations to tick a box. Deploying AI needs a business reason: a customer interaction that could be better, a process that could be faster or more consistent, or a task that is eating hours of someone’s week. Over the next two years, the businesses that get lasting value from technology will be the ones that understand their own operations, customers and employees well enough to point AI at a real problem. The ones that spin their wheels will be the ones trying to get ahead of the curve without that understanding.
Three warning signs your data is not ready
Before committing to an AI investment, I look for a few things. If the answers are unclear, there is foundational work to do first.
Do you know where your data is?
Siloed data, data in unexpected places and teams that cannot find the information they need are the first sign that the foundations need work before AI can use them.
Do you know who has access to it?
Data in a lakehouse or a database without proper controls has been a fairly benign problem until now. AI does not care about a lack of controls. It will find and use whatever it can reach, which is exactly why identity and governance have to sit around it.
Does every AI application or agent have an identity?
Like a person, an agent needs an identity so you know what it is accessing and what it is permitted to access. It also needs monitoring so you can see how it behaves. Those are design decisions, and they belong in the solution from the start, rather than being bolted on after a pilot.
Most of the metrics about organisations struggling with AI adoption trace back to skipping this groundwork. Getting it right is what gives a business confidence that its data is secure and is not being used for purposes the business is unaware of.
The cost of moving too fast, and too slow
Move too quickly and you tend to end up with a proof of concept that cannot leave the lab. The demonstration works, but nobody can say what it will cost at scale, whether it will deliver value, or whether the data behind it is secure, governed and observable. The pilot stalls, and the organisation concludes that AI does not work for it.
Move too slowly and the cost is competitive. The market is changing quickly, and the businesses that wait for perfect data watch competitors learn faster, build faster and serve customers better in the meantime. The practical path sits between the two: pick a worthwhile use case, establish what it needs, put the foundations in place for that use case, and measure the value as the work progresses rather than at the end.
Where Microsoft Fabric fits
Whether Fabric is the right investment depends on where a client’s data sits today, the technologies they already use, how they intend to use the data, and what they are prepared to pay. For a business already on Microsoft, it is usually the first platform we assess.
Its value is as a landing place. Fabric gives us a foundation to bring data together and build a medallion architecture that enriches that data and presents it back to applications and people in a structured way. It also carries the guardrails with it: identity, governance, security and observability on the platform rather than assembled from parts.
The second half of the value is what it connects to. From Fabric we can plug data into AI through Foundry, integrate third-party sources, and surface what matters in the places people already work, such as SharePoint, Outlook and Teams. The platform choice is one decision. It still takes engineering and clear operating responsibilities to make the result dependable.
What a client should be able to do without us
I want a client to come away from our work together understanding their data far better than when we started: where it is, how it is used, and the controls, governance and architecture around it. They should also be more ready for AI than they were, with the use cases that will deliver results already identified.
They should have the groundwork to develop those opportunities quickly and independently of us. They can identify the right thing to focus on, measure its value and get it into production in an accelerated timeframe, because we have been there building the patterns, the structure, the governance and the observability of the platform behind it.
Giving people time for better work
When we measure the value of AI, we look at it across employees, processes and outcomes. On the employee side, the question is where people spend their time and where it would be valuable for them to spend it. At my current client organisation, people lose hours to booking meetings, filling out spreadsheets and hunting for information that should be easy to find. Those are tasks AI does well.
On the process side, most organisations run hundreds of processes that have existed for years without anyone breaking them into their component steps. Doing that, and asking where AI can add speed or consistency to a step, is where a lot of the measurable value sits.
The time this returns is only valuable if people use it on the work that needs them: the interesting problems, the decisions that require judgement, the critical thinking. AI should not take that away, and in a well-designed system it does not. The human stays in the loop, questioning what the model produces and guiding agents on how they think and what they surface.
What keeps me awake
Security. The pace of AI means new vulnerabilities and incidents are surfacing every week, and I ask myself whether we and our clients are doing enough due diligence. Do we know what our critical systems are and what they run? Have we mapped the surface area we are exposing? Can we move fast enough to patch when an exploit appears, because it will? AI is an advantage in how we work. It also demands a serious response plan, and the organisations that treat security as a foundation rather than an afterthought are the ones that will be able to keep moving when something goes wrong.
How client work shapes nexwave
You cannot decide what to build from inside your own echo chamber. Our clients tell us what they are struggling with day to day. We combine that with what we see across the industry, and the solutions and products we are building now all trace back to a problem a customer described. Our part is to do it faster and better than anyone else, and to add our experience to it. The direction comes from them.
The same applies to how our consultants work. When someone on the team finds a better method, part of my job is to work out whether we can make it repeatable across our client partnerships. I keep up by following the people at the frontier labs and, just as usefully, the engineers on the fringe who are orchestrating agents in ways nobody has written up yet. Individuals have a louder voice in this moment than large organisations do, and a lot of the best ideas start there.
I also try things myself. I am building a personal command centre to manage commitments across client work, my CTO responsibilities and product development. The requirement is that I can read and update it, and so can an agent. It is still a work in progress, and it keeps me honest about what we ask clients to do.
A client should always know what we recommend doing next, why it is worth doing, and what needs to be in place for it to work. That is the job.
To discuss fractional CTO advisory or the data foundations for your AI plans, contact discover@nexwave.com.au