A private exploration for Harsh Mulik, HumLife360.

One updating view of every patient, so a Nudge Friend can act, not assemble.

You told us the biggest manual job you have is pulling a patient's whole story back together — across the app, WhatsApp, calls and workshops — before deciding their next nudge cycle. We explored what that looks like solved: not a product we bolt on, but how the idea would work for HumLife360 specifically.

“How do you bring all of that together — an integrated, summarised analysis of what the customer has been going through — that determines the next nudge cycle. That's the biggest manual thing we have going on right now.”
Harsh Mulik, HumLife360 — discovery call, 4 September

The work isn't the nudging. It's the stitching.

Every two-week cycle, a Nudge Friend engages one patient across five surfaces — then reconstructs the full picture by hand before choosing the next move. The judgement stays human. The assembly is what eats the hours.

App messagesWhatsAppCoach SessionsEmailWorkshops
Re-reading, per patient

Threads across five tools, stitched back together to rebuild one patient's fortnight.

Held in your head

Each patient's behavioural, motivational and physiological picture, carried from memory.

Decided under pressure

The next cycle chosen from that memory, patient after patient, against the clock.

The ceiling

How many patients one Nudge Friend can hold before the quality slips.

Three views.

Assembly offloaded, judgement kept — shown three ways.

How it works

The flow end to end — five channels pulled into one updating context layer per patient, structured into three living profiles, summarised for the next cycle. Consent-gated, human in the loop.

Trace the flow

Nudge Friend

The screen a Nudge Friend opens: a caseload at a glance, then one patient's whole fortnight assembled in seconds — timeline, three profiles, risk flags, and a drafted next cycle waiting for a human yes.

Open the view

Leverage & ROI

The numbers on paper — old versus new, all costs in. Where the twenty use cases rank by leverage, which one to build first, and why the return compounds as you grow toward the insurer play.

See the numbers

A note on how we're framing this.

The rest of the world dumps money into AI to look current. You asked for objective results on paper — maintenance, running cost, who owns it when needs change. Every view here is built to answer that, not to impress.

The system is spec-driven and rebuilt in days, not months — so when your model changes in six months, it changes with you. It's configuration, not a frozen product.