Somewhere in the last year, your board probably asked a version of this question: what is our AI strategy?
It is a fair question to ask. It is also, for most small organisations, the wrong question to ask first. The honest answer, in a surprising number of cases, is closer to: we do not yet have a website strategy, a data strategy, or a governance strategy, and AI sits on top of all three.
A recent MIT Technology Review Insights report on AI adoption contains a finding that gets far less attention than the technology itself, and matters more than almost anything else in it. Every organisation profiled in the report that genuinely benefited from AI had done the unglamorous work of cleaning up its foundations first. Not as a side note. As the precondition.
What the case studies actually show
Take Maps Credit Union, one of the organisations featured in the report. Before any automation of customer-facing processes, the credit union modernised its ageing telephony infrastructure. Only once that foundation was in place did the automation produce results worth reporting: roughly 90% of balance enquiries handled through IVR, and after-call work reduced by 90 to 94%. Those are striking figures, but they were not produced by the AI layer alone. They were produced by AI applied to infrastructure that had already been made coherent.
A higher education case in the same report followed a similar pattern from a different starting point. The institution consolidated four separate record-keeping systems into one before it saw any improvement in outcomes. One person involved described the consolidation itself, not the AI that followed it, as the thing that gave the organisation a clear roadmap for the first time.
The sequence in both cases is the same, and it is not incidental. Foundations first. Intelligence second. Reverse the order and the intelligence has nothing solid to stand on.
Why the sequence cannot be skipped
This is worth stating plainly, because it cuts against how AI is usually sold: intelligence does not fix fragmentation. It amplifies whatever it is layered onto, mess included. A recommendation engine built on inconsistent data will make confident, inconsistent recommendations. A chatbot answering from contradictory content will answer confidently and contradict itself. An AI-powered search feature indexing two unreconciled websites will surface whichever version it found first, with no way to know that a better answer exists on the other domain.
None of this is a flaw in the AI. It is a faithful reflection of what was already there, delivered with more confidence and at greater speed than a person would have delivered it. For a small organisation without deep technical resourcing, that combination, speed plus false confidence, is a genuine risk rather than a genuine gain.
What this means at small-organisation scale
Translated down from credit unions and universities to a community organisation with a handful of staff, "foundations" means something achievable rather than something requiring an enterprise IT department. It means analytics that are unified rather than duplicated or missing. It means content architecture that says one consistent thing about your services, rather than three slightly different things across three pages written at three different times. It means accessibility compliance that has actually been checked, not assumed. It means documented governance: who owns what, what gets reviewed and when, what happens if something breaks.
That is not a lesser version of AI readiness. It is the whole of it. An organisation with those four things in order is ready for whatever comes next, AI included. An organisation without them is not made ready by adopting a chatbot; it is made more visibly unready, faster.
Being a late mover is not a weakness
There is a reassurance buried in the same body of research, and it is worth saying out loud because the pressure to act immediately is real and rarely helpful. Most organisations, including well-resourced ones, are not early movers on this. The report notes that lagging on AI adoption puts an organisation in the company of most of its peers, not apart from them. Late and stable is a perfectly respectable place to be. Early and fragmented is not, whatever the sales material suggests.
The practical implication is that there is no cost to sequencing this properly. Getting your foundations in order first does not put you behind an AI curve that is moving faster than you. It puts you in a position to use whatever comes next well, rather than using it badly and discovering why only after the fact.
The practical first step
Foundations before intelligence starts with knowing, specifically, where your own foundations currently stand: what is unified and what is fragmented, what is documented and what lives only in someone’s memory, what would happen to any automation you layered on top of the site as it exists today.
A Digital Capacity Diagnosis is built to answer that question directly, across security, accessibility, search visibility, performance, user journey, and governance, before any AI or automation decision gets made. Once the foundations are understood, ongoing Digital Stewardship is what keeps them from drifting again.
Your board’s question about AI strategy is a good one. It is just worth answering the foundations question first.
