Build the Data Foundation Before the AI: How UK SMEs Can Retrofit Their IT, Systems and Data for Reliable Automation

---

Most UK SMEs are being sold AI tools before they’ve fixed their data. The result is predictable: nicely branded pilots sitting on top of half‑broken spreadsheets, duplicate contacts, and systems that do not agree with each other.

We see the same pattern in London and South East firms from 10 to 100 people. The owner is excited about AI. The ops lead is nervous about yet another integration. Finance and sales are arguing over whose numbers are “right”. Then someone asks the question that stops everything:

“If our systems don’t even match today, how can we trust an AI on top of them?”

This article answers that question with a practical, step‑by‑step way to prepare IT systems for AI, retrofit your existing stack, and clean data for automation without rebuilding everything from scratch. It is not about perfection. It is about getting your data to “good enough” so that automation is reliable, auditable and worth the money.

---

What you need in place before you touch AI

If you skip these prerequisites, any AI initiative will either:

• quietly fail after 3–6 months, or
• create more admin than it removes.

You’re ready to start building a data foundation when you have:

1. A single internal owner
   One person with authority over systems and processes for at least 4 hours per week. In our AI Readiness Scorecard, this is the Team Capacity dimension; if nobody can own the change, you’re not AI‑ready yet.

2. Basic process clarity
   You don’t need perfect SOPs. But for your top 5 operational workflows (sales, invoicing, service delivery, etc.), people should broadly agree on the steps. If everything lives in people’s heads, start