Businesses are under pressure to develop an AI strategy but making AI the objective is often where projects go wrong. The better starting point is to identify the business outcomes you need to improve, establish whether your data and organisation are ready, and then decide where AI can genuinely help.
Every leadership team I speak to is wrestling with the same question, usually asked by a board, an investor or an anxious executive committee:
“What are we doing about AI?”
It’s a fair question. It’s also the wrong place to start.
The moment a business sets out to build an “AI strategy”, it quietly makes AI the objective. Budgets get allocated to a technology rather than an outcome. Committees form around a capability rather than a problem. And six months later, the honest answer to “What did we get for that?” is a handful of pilots, an impressive demo or two, and very little that has changed how the business actually runs.
The objective was never supposed to be AI. It was supposed to be a better business: lower costs, faster decisions, better service, less risk, higher productivity, easier access to the knowledge already sitting inside the organisation, or new revenue.
AI is one of the tools that can get you there. It is not the destination.
At GoRebel, our position is simple, and it shapes everything we build:
AI belongs inside your business strategy, not alongside it.
The rush to “do AI” is creating the wrong starting point
I understand the pressure. Nobody wants to be the leader who did nothing while competitors moved.
But that pressure is producing a predictable pattern, and we see it across every sector we work in.
An executive feels the heat and asks the organisation to “explore AI”. Teams start experimenting with whatever tools are in front of them. Pilots get launched without anyone agreeing what a successful outcome would look like. Technology decisions get made before the business problem has been properly understood.
Then a slick demo – the model answering a question convincingly in a meeting – gets mistaken for a working implementation.
The trouble is that “Where can we use AI?” is a seductive question that rarely leads anywhere useful.
A better question is:
Where is the business losing time, money, knowledge or opportunity?
That doesn’t start with the technology. It starts with the pain. And pain is something a business can measure, prioritise and justify spending against.
Start with the business problem, not the technology
When a genuinely worthwhile AI opportunity crosses my desk, it almost never arrives dressed as an AI opportunity.
It arrives as a business problem that someone is tired of living with.
In FinTech, it might be analysts burning hours hunting for information across systems, compliance-heavy workflows slowing everything down, or risk and support processes that don’t scale with the customer base.
In education, it might be staff endlessly searching for policies, course material, student information and institutional knowledge that lives in too many places and too many heads.
In manufacturing, it could be production inefficiencies, recurring quality issues, or maintenance knowledge trapped inside SOPs and systems that people on the floor can’t easily reach – all contributing to poor operational visibility.
In health, it may be administrative workload, the constant need to retrieve controlled information safely, or the challenge of giving the right people secure access to sensitive information without opening the door to everyone.
Notice that none of those is a request for a chatbot.
This is the distinction I push hardest on with leadership teams.
“We need an AI chatbot” is a solution looking for a problem.
“Our teams spend a fifth of their week searching for information across six different systems” is a business problem – one with a cost attached, an owner and a way to tell whether you’ve fixed it.
Only the second version gives AI something real to do.
Check whether the organisation is ready
You can identify an excellent use case, with real value and a clear owner, and still fail because the organisation underneath it isn’t ready.
Before committing to an AI initiative, there are a few unglamorous questions worth answering honestly:
- Is the data you need actually available and reliable?
- Can the systems holding it connect to anything?
- Who is allowed to access it?
- Is sensitive information properly controlled?
- Who owns the system once it goes live?
- How will you measure whether it worked?
If those questions make you wince, that’s useful information.
It tells you that the model was never the whole project.
A working AI system also depends on data, integrations, permissions, governance, adoption and ownership. Skip those foundations and even a powerful model can simply give unreliable information a more convincing interface.
Data comes before AI
Which brings me to the point I’d underline twice if I could.
Many businesses that think they have an AI problem actually have a data problem.
Their information is scattered across ERP, CRM, document stores and a graveyard of spreadsheets. It’s duplicated in places nobody fully maps. It’s poorly structured, inconsistently maintained, governed by permissions that made sense years ago and – crucially – difficult to trust.
When the underlying data is that fragmented, no amount of model sophistication rescues it.
Better AI cannot compensate indefinitely for bad data, poor access controls and disconnected systems.
A more powerful model applied to unreliable information simply produces unreliable answers faster.
That’s not progress. That’s risk.
This is the foundation much of what we do at GoRebel is built on, and it’s a big enough subject to deserve its own conversation. I’ll come back to it in a follow-up piece on why your AI problem might really be a data problem.
Choose the AI projects worth doing
Assuming the foundations are in reasonable shape, how do you decide what to do first?
I use a simple four-question test for any proposed AI project.
1. Is there a real business problem?
Not an interesting use case. Not something that would demo well.
A problem the business genuinely wants gone.
2. Is there measurable value?
Time saved. Revenue gained. Risk reduced. Errors prevented. Service improved.
If you can’t name the metric, you’ll struggle to defend the spend.
3. Is the required data available and usable?
Not “does it exist somewhere in principle?”
Is it accessible, trustworthy and usable?
4. Can it be implemented safely and realistically?
Within your actual constraints – not a hypothetical greenfield version of your business.
The projects that survive all four questions tend to share a profile: high-value, clearly defined and achievable without requiring the entire organisation to transform first.
That last part matters.
You don’t need to boil the ocean to get real value from AI.
A successful demo is not a successful AI project
There’s one more gap that catches a lot of businesses: the gap between a proof of concept and something the organisation can actually rely on.
A proof of concept answers a single question:
Can this work?
Production has to answer a much harder one:
Can this work reliably, securely and repeatedly inside our business, every day, for real users?
Everything that gets waved away in the demo becomes the actual project once you cross that line:
- integrations
- security
- permissions
- governance
- monitoring
- user adoption
- support
- cost
- scalability
- accountability when something goes wrong
I’ll say something that might sound odd coming from an AI company:
The impressive AI demo is usually the easiest part of the project.
The engineering that makes it trustworthy at scale is the work you’re really paying for.
Put AI back inside the business roadmap
So, what should leadership do instead of treating AI as a standalone strategy disconnected from the priorities the business already has?
Look at those priorities and ask where AI could accelerate them.
If the priority is reducing operating costs, ask where AI could eliminate repetitive knowledge work.
If it’s improving customer experience, ask where it could improve response times or give staff faster access to the information they need.
If it’s reducing risk, ask whether AI could improve access to policies, controls and compliance information.
If it’s productivity, find the knowledge and processes consuming a disproportionate share of your people’s time.
If it’s modernising operations, ask which of your existing systems and data could become more useful with AI on top.
Framed that way, AI stops being a separate initiative competing for attention.
It becomes an enabler of goals the business already understands and already cares about.
So, what are you doing about AI?
Replace the question.
Instead of:
“What are we doing about AI?”
Ask:
“Which of our business priorities could AI help us achieve faster or better?”
That single shift changes everything downstream.
The strongest AI programmes I’ve seen start the same way: with a clear business outcome, an honest look at the data and organisational foundations required to reach it, and only then a decision about the technology needed to deliver it.
In that order. Never the reverse.
That’s the work we do at GoRebel: helping businesses identify where AI can create real value, determine whether the foundations are ready, and build the systems needed to put it into production.
Not sure where AI fits into your business?
Start by assessing whether the foundations are ready across your data, technology, governance, people and strategy.
Have a specific AI opportunity in mind? [Talk to GoRebel]
Related: Data & AI Consulting · Coming soon: “Your AI Problem Might Actually Be a Data Problem”
