The tension
The ideas were there. The operating method was not.
The team could already name the friction: voice-of-customer tickets were difficult to connect to the opportunity registry; competitive research arrived ad hoc; a product brief could take hours to turn into a branded slide; accessibility checks happened too late; and meeting notes varied from person to person.
None of those problems needed another generic AI presentation. They needed a shared way to move across Notion, Slack, Figma, Zoom and Google Workspace without losing judgement, provenance or the human review step.
The intervention
Thirty minutes of framing. Two hours of building. Six live prototypes.
Discovery interviews before the workshop meant every participant arrived with a problem worth solving. After a short framing session, the group built for two hours with Claude Code and MCP against Modern Health's actual sources and standards, then showed the work to one another.
The prototypes included a VOC-to-Opportunity Matcher with human review; a competitive-intelligence scanner; a brief-to-branded-slide workflow; a Figma library prototype; a meeting-notes agent; and an accessibility auditor checking WCAG AA contrast, target size, hierarchy, labels and off-library components.
The change
Reusable patterns, known failure cases and a Monday-morning plan.
Across the six builds, three patterns kept returning: synthesise across systems; generate from structured data; and configure the tools the company already trusts before commissioning a custom application.
Every track also surfaced a failure case. The group saved shared prompts, documented the next action for each prototype and scheduled follow-up office hours within two weeks. The value was not a polished demo—it was a set of useful systems the team understood well enough to keep improving.
