an AI that runs the campaign, and shows its working

The client is a UK SME with a real opportunity in front of it and not enough hands to work through it. Somewhere in publicly available data were the prospective customers most likely to buy, and in a growing spreadsheet were the leads already gathered by hand. What was missing was the capacity to turn all of that into a planned, running campaign, instead of a list nobody had the time to action.

The raw materials were unglamorous: openly available information about potential markets, and a spreadsheet the business was already keeping itself. The first project, AIM-DSH-P01, established what could be trusted from each source, how a prospect should be scored, and what a genuine opportunity actually looked like, before any AI was pointed at a single decision.

The campaign runs through a dashboard built under our PLAN, DESIGN, BUILD, STATUS methodology, where the AI model works one phase at a time and reports back at each one. It reads the public data and the spreadsheet, proposes which markets and prospects are worth pursuing, drafts the outreach for that phase, and then waits. Nothing leaves the building on the model's say-so alone. Every phase is visible on the dashboard as planned, in progress, or done, so the people running the campaign always know exactly what the AI has done and what it is about to do next.

The result is a full marketing campaign managed end to end from one screen. The AI segments the market from public data, pulls and enriches leads from the spreadsheet, plans each phase of outreach, and writes back its progress so the dashboard is always current. Instead of a person chasing the model for updates, the model reports to the person. The campaign became something you could watch happen, phase by phase, rather than something you had to assemble by hand.

Left to its own confidence, the model was ready to treat a promising-looking pattern in the public data as a proven opportunity and market against it at scale. On a thin evidence base, that would have meant real spend and real outreach aimed in the wrong direction. The governed build put a human approval gate between the AI's recommendation and any action, and the STATUS document records where the evidence was strong and where it was still a guess. The AI proposes; a person decides. That line is deliberate, and it holds.

The dashboard is live and the campaign runs through it, phase by phase, with the AI reporting progress and the client approving each move. It's already surfacing prospective clients from data the business could never have worked through by hand, and doing it in the open, where every recommendation can be seen and questioned. A campaign that manages itself, and still answers to people.