The partnership had grown to 28 doctor-owned practices across the Mountain West, and with growth came fragmentation. Each practice ran its own experiments, so an office in one town might be trialling a radiograph-analysis tool while its neighbour tested a scheduling assistant, with no way to compare notes. Patient data, the raw material for any of it, sat in three separate practice-management systems that did not talk to each other.

Vendor pressure made the noise worse. Sales teams pitched overlapping radiograph-analysis and scheduling products practice by practice, each with its own contract, pricing and data demands. Because the partnership is owned by its doctors, every meaningful decision needed partner agreement, and the partners did not agree: some wanted clinical AI first, others wanted the front desk fixed, and nobody had evidence to settle the argument.

Two early pilots had already stalled without measurable outcomes, and the failures cast a long shadow. With no agreed starting point, no shared evaluation method and no owner for the question, the group's AI ambition was producing subscriptions and meetings rather than results, while each practice kept paying for tools the one next door had already tried and quietly dropped.

We started with evidence rather than opinions: a structured AI readiness assessment covering all 28 practices, auditing data quality in the three practice-management systems, mapping day-to-day workflows and reviewing every active vendor contract. That gave the partnership its first complete picture of what it was already paying for and what its data could actually support.

With the groundwork done, we facilitated scoring workshops where partners and practice managers ranked more than 40 candidate use cases on impact and effort, in the open and against the same criteria. Ideas with powerful champions but weak cases fell away; ideas nobody had championed rose. The shortlist that emerged was one the room had built together, which mattered in a partnership where the owners vote.

The final deliverable was a phased eight-quarter adoption roadmap: 6 initiatives sequenced by dependency, each with governance guardrails, a named owner and a budget. We then helped launch the first two pilot programmes ourselves, so the roadmap left the boardroom with momentum rather than as a document.