Part 2 · The way of working

AI improved speed. Product decisions remained human.

AI helped process information, but strategy, prioritisation, and trade-offs were my decisions.

Research Synthesis Positioning Deck creation

How I approached the work

Start with domain context, not just the data

Find the key pattern in the data before jumping to solutions

Use AI to accelerate analysis, clustering, and synthesis

Keep product decisions human — especially prioritisation and trade-offs

The workflow was designed to get to a credible product bet quickly, then express it clearly in both the landing page and the decks.

1 · Domain understanding

I grounded the work in how hospitals and cardiology teams actually operate.

What I needed to understand first

How hospitals operate and where cardiology sits inside them.
How cardiology teams make decisions in real clinical workflows.
How clinical software is used at the point of care.
How hospital software buying and deployment decisions get made.

Why it mattered

It changed how I interpreted the product and churn signals.
It made mobile workflow and alert fatigue look more important.
It made me treat reliability as a workflow issue, not just a technical issue.

Key insight: a lot of clinical decisions happen in fast, mobile, interrupted environments. That changed the product read.

2 · Data analysis

I combined structured metrics with qualitative signals instead of relying on one source.

Inputs reviewed

Product metrics: engagement, alerts, usage.
Commercial metrics: churn, ACV, pipeline.
Sales data: wins, losses, objections.
Customer feedback: qualitative insights.

Quantitative pass

Compared large vs community hospitals.
Looked at engagement, retention, and feature usage.
Checked where the commercial upside and risk overlapped.

Qualitative pass

Clustered comments into alert fatigue, integration, mobile, and reporting themes.

This revealed a clear pattern: strong performance in community hospitals, weak performance in large hospitals.

3 · Key insight

The core problem: MedBuild is strongest where revenue is smaller.

This creates a direct risk to retention and future growth.

Key pattern

MedBuild performs well in community hospitals.
It struggles in large hospitals where the revenue concentration sits.
The biggest opportunity is fixing large-hospital adoption and retention.

Core problems to solve

Alert fatigue.
Data reliability.
Poor mobile workflow.
Weak proof of value.

4 · Strategic options

I looked at multiple directions and chose the one with the strongest six-month leverage.

Option 1

Focus on community hospitals

Easier growth, but weaker long-term leverage against the current revenue concentration.

Risk: Leaves main revenue at risk.

Option 2

AI-first transformation

Strong narrative, but too risky while trust and workflow basics are still weak.

Risk: Builds on weak foundations.

Option 3

Fix large-hospital adoption

Highest revenue impact and directly addresses churn. This was the chosen direction.

Why chosen: Directly improves retention and revenue.

5 · Core decision

The decision was to fix large-hospital adoption first.

That is where MedBuild has the most revenue, the most churn risk, and the weakest product fit.

What I chose

Fix trust first.
Fix workflow second.
Add AI only after that.

Why this direction

It targets large-hospital churn.
It improves retention and usage.
It protects enterprise revenue.

Why not the others

Community focus was easier, but lower impact.
AI-first was exciting, but too early.
Expansion added scope before fixing the core issue.

6 · Landing page

I used the landing page to show positioning, structure, and the key messaging choice.

The landing page reflects the strategy: focus on core pain to improve enterprise conversion.

Why I built a landing page

Show product positioning.
Show the target audience.
Make the core problem clear.
Demonstrate the value proposition.

Landing page structure

Hero with a clear value proposition.
Problem grounded in real data.
Solution and how it works.
AI section kept focused and realistic.
Impact section with measurable outcomes.

Key positioning decision

I avoided leading with AI to keep focus on real clinical pain. Instead, I led with a more concrete promise: turn alert overload into real clinical action.

Why this works

Matches the real user pain.
Feels more credible.
Differentiates from generic AI messaging.

7 · Business impact

How this helps the business.

This strategy directly improves retention, engagement, and enterprise revenue.

Retention Reduce large hospital churn
Improve renewal confidence

Usage Increase alert engagement
Increase clinician activity

Trust Reduce data delays
Improve reliability

Revenue Protect enterprise ARR
Enable expansion

8 · Way of working

How I built the exercise output from start to finish.

I moved from data → problem → decision → execution.

01

Frame the problem

Defined the business goal and constraints.
Identified the key question: where is the biggest risk?
Focused on outcomes, not features.

02

Understand the domain

Mapped how cardiology teams actually work.
Looked at real clinical workflows and decision points.
Considered how hospitals buy and use software.

03

Analyse data

Reviewed usage, engagement, and churn.
Compared large vs community hospitals.
Combined metrics with customer feedback.

04

Choose direction

Evaluated multiple strategic options.
Prioritised highest business impact.
Selected one clear, focused direction.

05

Turn into execution

Defined what to build and in what order.
Aligned product, team plan, and messaging.
Translated strategy into landing page and roadmap.

The goal was speed with clarity, not perfect analysis.

9 · AI vs Product judgment

What AI was used for, and what stayed human.

ChatGPT

Used for domain research and for analyzing the MedBuild data across commercial, product, feedback, and sales inputs.

Lovable

Used to create the landing page and turn the strategic direction into a market-facing narrative.

Codex

Used to build the HTML presentations so they matched the landing page design and kept the overall feel consistent.

AI used for

Research.
Pattern detection.
Drafting.
Clustering feedback themes.
Structuring the landing page and slides.

Human decisions

Strategy.
Prioritisation.
Trade-offs.
Which segment to focus on.
What not to build.

AI was most useful as a fast analyst and packaging assistant. The strategic choice itself stayed human.

10 · Closing thought

What this exercise shows.

How I worked

Started with the domain.
Analysed the data.
Chose one clear direction.
Turned it into a landing page and decks.

What mattered most

Clear prioritisation.
Realistic scope.
Consistency between strategy and messaging.
Focus on the biggest business risk.

I used AI to move faster, but the outcome still depended on product judgment, prioritisation, and trade-offs.

Bottom line