What is a data strategy and why does your business need one before AI?
Most organisations at this scale have already spent money on AI. A pilot here, a vendor contract there, perhaps a proof of concept that the data team is still trying to operationalise six months after it was declared a success. The results have been underwhelming - not because the technology failed, but because the foundation was never there.
The mistake firms at this stage consistently make is treating AI as the starting point. It is not. AI is a consumer of data. If your data is incomplete, inconsistent, siloed or poorly governed, AI does not fix that. It amplifies it. You end up with fast, confident, expensive answers that are wrong.
A data strategy is the foundation that makes AI viable - and valuable. Without one, you are building on sand regardless of the vendor, the model or the budget you commit. This article explains what a data strategy actually is, why most mid-market firms do not have a real one and what you need to get right before AI investment can deliver a return.
What a data strategy is - and what it is not
A data strategy is not a technology roadmap. It is not a list of tools you plan to buy or a description of your data warehouse architecture. Those are outputs. A data strategy is a set of deliberate decisions about what data your business needs to compete, how you will collect and maintain it, who owns it and how it connects to commercial outcomes.
It answers four questions:
- What decisions in our business should be data-driven that currently are not?
- What data do we need to make those decisions reliably?
- What is the current gap between the data we have and the data we need?
- What governance, infrastructure and capability do we need to close that gap?
A retailer turning over £800m might have transaction data, loyalty data, supplier data and web analytics all sitting in separate systems, owned by separate teams, with no consistent customer identifier across them. They might believe they have rich customer data. What they actually have is four partial pictures that cannot be joined without months of remediation work. That is a data problem. Buying an AI personalisation platform does not solve it - it surfaces it, expensively, in production.
Why most mid-market firms lack a genuine data strategy
The firms we work with are not naive. Their leaders understand data matters. The issue is that data strategy tends to emerge reactively rather than being designed deliberately.
It typically develops in fragments. The finance team builds a reporting layer around the ERP. The marketing team buys a CDP. The commercial team runs their own spreadsheets because the official dashboards are three weeks behind. The IT function maintains infrastructure but has no mandate to enforce standards. Nobody has drawn a line connecting any of this to where the business is trying to go in the next three years.
This fragmentation is compounded by turnover. A CDO or Head of Data who joined two years ago inherited decisions made by three predecessors. The documented rationale for those decisions no longer exists. What remains is a set of technical constraints that the current team works around rather than addresses.
The result is what we call "data debt" - the accumulated cost of undocumented pipelines, inconsistent definitions, ungoverned access and tribal knowledge that lives only in the heads of people who may not be there next year. For a firm at £1bn revenue, the operational cost of data debt typically materialises in analyst time wasted reconciling conflicting numbers, leadership decisions made on stale or incomplete information and failed technology projects that could not be scoped accurately because nobody knew what the data actually looked like.
The six decisions a data strategy must make explicit
A working data strategy forces clarity on decisions that most organisations leave implicit. Those decisions are:
- Data ownership. Which team or individual is accountable for the accuracy and maintenance of each critical data domain - customer, product, supplier, financial.
- Definitions. What does "active customer" mean? What counts as revenue in a given period? Consistent definitions, agreed at leadership level and enforced at source, are the difference between a business that can trust its numbers and one that spends the first twenty minutes of every board meeting arguing about them.
- Collection and quality standards. What data must be captured, at what fidelity, and what happens when it is not? This is where most strategies have a gap - they describe what they want but not how they will maintain it.
- Access and governance. Who can access what, under what conditions, with what audit trail? Relevant both for regulatory compliance and for controlling how AI systems interact with sensitive data.
- Integration architecture. How do data domains connect? What is the canonical system of record for each? This determines whether your AI systems are working with a coherent view of the business or stitching together conflicting inputs at runtime.
- Prioritisation. You cannot fix everything at once. Which data domains are most critical to the commercial decisions that matter most right now?
A private equity-backed business preparing for a value creation programme needs to answer all six of these before it can credibly scope an AI initiative. If the data team cannot answer them, that is the engagement that needs to happen first.
How a data strategy connects to AI readiness
The connection is direct. Every AI application - whether predictive, generative or agentic - depends on input data that is accurate, current, consistently defined and accessible in the right form. A language model summarising contract terms needs clean, structured contract data. A demand forecasting model needs a consistent, unbroken history of sales and inventory at the right granularity. An agentic system taking commercial actions on behalf of a business needs a trustworthy, real-time view of the state of that business.
None of those things exist by default. They exist because someone made deliberate decisions upstream.
The firms that are getting genuine commercial return from AI right now are not the ones that moved fastest on model deployment. They are the ones that spent 12 to 18 months quietly doing the unglamorous work of data remediation, domain ownership and integration architecture before the AI layer went anywhere near production.
A useful diagnostic question for any leadership team: if a new AI system needed a reliable, joined view of your customer, product and transaction data tomorrow, how long would it take you to produce it? If the honest answer is "months" or "we are not sure", the data strategy comes first.
Rodan's diagnostic engagements are specifically designed to answer that question with precision - mapping current data state against commercial ambitions, identifying the highest-leverage gaps and producing a prioritised action plan that AI investment can then be built on top of.
The cost of treating data strategy as something you do later
The temptation is to start with the AI use case and retrofit the data work. It feels faster. It shows momentum to boards and investors. In practice, it produces the worst outcome: a visible, funded initiative that cannot deliver because the underlying data cannot support it.
That failure has a compounding cost. It consumes budget. It consumes the goodwill and credibility of the data team. It teaches the organisation that AI does not work - when the actual lesson is that AI without data strategy does not work. The next initiative faces a harder internal sell and a more sceptical leadership team.
For firms at £500m to £1.5bn revenue, where a transformation programme typically carries an eight-figure price tag and a 24-month timeline, getting the sequencing wrong is not an inconvenience. It is a material commercial risk.
The right sequence is: data strategy first, AI infrastructure second, use case deployment third. That is not a conservative position. It is the position of every organisation that has actually delivered AI at scale.
What to do now
If you do not have a written, agreed data strategy - one that names owners, defines critical domains and connects explicitly to your commercial priorities - you have a gap that no AI investment will close on your behalf.
Start by commissioning an honest assessment of where you actually are. Not a vendor-led discovery workshop designed to justify a platform sale. A neutral, commercially grounded diagnostic that tells you what your data can and cannot support, and what it would take to change that.
That is precisely what Rodan's AI readiness assessment delivers. It is a fixed-scope, fixed-price engagement that gives senior leadership a clear picture of data maturity, the specific barriers to AI value and a prioritised roadmap to address them.
If AI is on your roadmap for the next 12 months, book the diagnostic now - before the AI programme starts, not after it stalls.
