The board asked about AI. The CEO is asking what the AI strategy is. Six months later, the answer is a pilot that did not scale, a Microsoft Copilot rollout that the team barely uses, and a frustrated leadership team wondering where the ROI is.
We have seen this cycle multiple times, and the root cause is almost always the same. The AI part is fine. The data the AI has to work with is broken.
Enterprise data is rarely clean. Customer records are duplicated across three systems with slightly different spellings. Product codes have drifted over 15 years of mergers. Historical contracts are PDFs without structured metadata. Processes live in people's heads because the documentation was never written down, or was written down but is three years out of date.
If you point an AI at messy data, you get AI-generated answers that are confidently wrong. Users lose trust fast. Once they have lost trust, they do not come back even if you fix it later. First impression is expensive.
The unglamorous truth is that the pre-work for enterprise AI is usually 60 to 80 percent of the project. Data consolidation. Master data management. Document digitization. Pipeline building. Access control. Then, and only then, AI on top. The order is not negotiable.
The mistake most organizations make is inverting the order. They pick an AI product, deploy it, and assume the data will sort itself out. It does not. The AI reveals all the data problems nobody knew they had, and the team that rolled out the AI gets blamed for the data problems they inherited.
When we run an enterprise AI engagement, the first 3 to 6 months are almost never AI work. They are data and infrastructure. Clients who embrace this sequence get to a place where every subsequent AI use case is quick to ship. Clients who try to skip it end up doing the data work anyway, but in a fire drill after the first pilot failed publicly.
If you are being pitched an "AI solution" that promises to deploy in 30 days with no data work, ask what happens if the data is bad. The answer will tell you whether the pitch is serious.