Ten years of shipping — and the lessons that came with it.

The challenge

Enterprise healthcare implementations combine strict regulatory requirements, legacy interfaces (HL7 and lab instruments), and hospital stakeholders who have very little time. Scope drift and unclear ownership were slowing delivery and eroding client confidence.

What I did

• Established a single program cadence: weekly steering, RAID log, and a milestone plan each client could see and sign off on.
• Rebuilt requirements practice around traceable user stories with acceptance criteria, linking each to test cases and release notes.
• Ran gap analysis and process-mapping workshops with lab directors to convert workflows into configurable system behaviour rather than custom code.
• Introduced structured change control so late requests were costed and scheduled instead of silently absorbed.

Outcome

The challenge

Requirements lived in people's heads and in paper forms that varied by office. Stakeholders spanned senior officials, front-line staff and IT — each with different definitions of success. A "big bang" launch was not an option.

What I did

• Led discovery: stakeholder interviews, current-state BPMN process maps and a service blueprint that exposed duplicated steps and approval bottlenecks.
• Wrote the functional and non-functional requirements (security, audit, accessibility) and a data-migration plan for historical records.
• Designed a phased rollout — one department at a time — with training, feedback loops and a defined go/no-go checklist for each phase.
• Acted as the bridge between officials and the engineering team, translating policy constraints into build decisions.

Outcome

The challenge

The backlog was a list of everything competitors did. There was no shared definition of the MVP, no success metrics, and engineering was building without acceptance criteria — leading to rework and missed dates.

What I did

• Ran problem-framing and customer-interview sessions to define the core job-to-be-done and the smallest release that would prove it.
• Wrote the PRD, user flows and API specifications; set up Jira with a lean Scrum workflow and definition-of-done.
• Prioritised with RICE, cut scope ruthlessly, and protected the team from mid-sprint changes while keeping the founder informed weekly.
• Defined launch metrics (activation, first transaction, retention) and instrumented them before release.

Outcome

The challenge

Requests arrived by email, chat and phone; triage depended on two people's memory; weekly reports were assembled by hand. Any automation built on top of that chaos would have automated the chaos.

What I did

• Mapped the end-to-end intake-to-delivery process and quantified time spent per step to find the highest-value automation candidates.
• Specified an LLM-based classification and summarisation step with human review, plus rule-based routing into the CRM.
• Defined guardrails: confidence thresholds, fallbacks to manual handling, and an audit log for every automated decision.
• Delivered in increments — intake first, then triage, then reporting — measuring hours saved at each step.

Outcome