The Challenge
A mid-size capital markets firm had more than 40 AI ideas scattered across trading, operations, compliance, and client service desks. Three desks were independently piloting near-identical document-intelligence tools. Two pilots had consumed budget for over a year with no production path and no defined conditions for shutdown. The head of technology could not answer the board’s simplest question: what is the AI portfolio worth, and why these bets?
There was no inventory, no shared scoring model, and no kill criteria. Funding followed sponsor seniority, not portfolio logic.
Our Approach
We deployed AI Portfolio Explorer as the single system of record for the firm’s AI portfolio:
Inventory: Captured all 47 use cases in one workspace, seeded from the platform’s capital-markets content pack and enriched through structured working sessions with each desk.
Score: Applied sector-specific dimensions (revenue impact, regulatory exposure, data readiness, execution complexity) to every use case. Scoring arguments happened in the open, against written criteria, with the Claude strategy co-pilot drafting rationales for the investment committee.
Decide: The priority matrix and maturity curve separated the portfolio into explicit decisions: 12 funded, 9 killed, 8 near-duplicates merged into 4 platform bets that were funded within the 12, and the remainder placed in a scored backlog with revisit dates.
Govern: Jira and ServiceNow connectors pushed every funding decision into delivery queues, and signed webhooks kept downstream systems current. Quarterly portfolio reviews now run from the dashboard instead of a month of slide assembly. The funded initiatives entered BenchMark for infrastructure calibration.
The Results
- 47 ideas, 12 funded initiatives: Each with an owner, a written scoring rationale, and pre-agreed kill criteria.
- 9 zombie pilots terminated: Including the two pilots that had run for over a year without a production path, releasing engineers and budget within one quarter.
- ~30% of AI budget reallocated: From duplicated and unviable pilots to the funded portfolio.
- Governance cadence established: Quarterly portfolio reviews compressed from weeks of preparation to a two-hour dashboard session.