$6.2M invested.
17 pilots running.
Zero P&L impact.
A Big Four-aligned financial services group had spent 18 months building an AI portfolio. By every internal measure, it was thriving. By every external measure that mattered, it had produced nothing.
Reduction in processing time
Across three core operations functions
Annualised value identified
From 4 agentic workflows in production
Engagement duration
From audit to operating model handover
Agentic workflows live
Within 90 days of engagement close
Celebrating pilots instead of producing outcomes.
The group's AI steering committee met monthly. Slide decks showed green status across every initiative. Vendor relationships were strong. The innovation team was growing.
But no workflow had crossed into production at scale. No AI model was touching a live operation. The $6.2M was producing dashboards, not decisions.
When the CFO began asking for returns, nobody could answer the question. Because nobody had been asking it.
No owner. No measurement. No operating model.
AI decisions were made by whoever had the vendor relationship. Pilots were approved by innovation teams with no authority over operations. Success was defined as "completed". Not "produced value."
The organisation had no operating model for AI. It had a procurement model, a reporting model, and a vendor management model. None of them connected AI capability to operational throughput.
Four phases. Fourteen weeks.
Strategic Truth Audit
Mapped every active AI initiative against P&L, operational throughput, and executive ownership. Found 17 pilots , 14 had no clear sponsor, no success metric, and no path to production. Three were ready to scale. None were scaling.
Operating Gap Analysis
Identified two structural blockers: a data governance gap preventing model deployment to production, and an accountability vacuum. AI decisions were being made by vendor relationship managers, not operations leaders. No single executive owned the portfolio.
Agentic Operating Redesign
Built an AI operating model with defined ownership at each layer: strategic, operational, and technical. Designed three agentic workflows targeting the highest-leverage operations: document processing, exception triage, and reconciliation review. Each workflow had a named owner, a measurement framework, and a rollback plan.
Handover and Capability Transfer
Trained internal teams to operate, iterate, and extend the model without external dependency. Established a quarterly review rhythm. Exited cleanly. The group has since scaled one additional workflow without external support.
"We did not need more pilots. We needed someone to tell us which three to scale and how to do it without breaking what was already working."
Three workflows live. One operating model. No external dependency.
Within 90 days of engagement close, three agentic workflows were in production: document processing, exception triage, and reconciliation review. A fourth was in staging.
Processing time across the three operations functions dropped 31%. Annualised value from the four workflows was modelled at $2.1M. Conservative, based on throughput already observed, not projections.
The group's AI operating model now sits inside the COO function. Ownership is clear. Measurement is routine. The remaining pilot portfolio has been reduced to four initiatives. Each with a named owner, a defined success metric, and a production pathway.
Ready to stop celebrating pilots?
We audit what you have, identify what's worth scaling, and build the operating model to carry it.