"If you found €1.4M in one department, what's hiding in the other seven?"

A European tower company managing 12,000+ sites across four countries had been experimenting with AI — an internal chatbot here, a predictive maintenance proof-of-concept there — but had no systematic picture of where AI would actually pay off. Three enterprise vendors were pitching them. The board was asking for an AI strategy. They had no objective way to decide.

The COO wanted an independent assessment before committing budget to any vendor. They engaged Kerf to begin with Operations — their largest team and the area they suspected was bleeding the most time.

The brief was straightforward: map what actually happens in Operations, identify where AI would genuinely help, and put numbers on it. The full-organisation scope was not on the table at the start.

One week embedded in the Operations team. Structured process-mapping interviews with every role that touched field dispatch, maintenance scheduling, and site reporting:

All process maps were validated with interviewees at the end of each session. Discrepancies between what different people described as 'the process' were flagged explicitly — these were consistently the highest-value areas.

The AI Opportunity Register was built in parallel with the process mapping: each identified gap was immediately assessed for automation potential, estimated ROI, and implementation feasibility before the week was out.

The highest-impact single opportunity was an AI-powered field dispatch system. The current process required a human dispatcher to manually match available engineers to open work orders, cross-referencing skill certifications, location, and equipment availability across three systems. The matching logic was well-defined — it had just never been automated. Estimated saving from AI-assisted dispatch: €340,000 per year.

The site reporting workflow was the second major finding. Monthly performance reports were compiled manually from three disconnected data sources by a team of analysts — each analyst doing it slightly differently, each month taking between 12 and 18 hours of analyst time. An automated report generation pipeline would eliminate this entirely. Estimated saving: €280,000 per year.

Five further opportunities were identified across maintenance planning, contractor management, and anomaly detection — collectively worth an additional €780,000 in annual savings.