Temitayo Bewaji

Physician, entrepreneur, loves God, family, sports, music and tech.

[Temitayo Bewaji @drtemib]

A scheduling system that deprioritizes the wrong patient has made a clinical decision. Most orgs don't govern that because it happened in billing, not bedside. Start with one AI tool. Ask: what happens when it's wrong? Can you override it?

[Temitayo Bewaji @drtemib]

"Human in the loop" means nothing if overriding the AI costs you more than following it.

That asymmetry is your real governance policy. Everyone on the team already knows it.

Thoughts?

[Temitayo Bewaji @drtemib]

The algorithm stays the same. What changes is everything downstream of it.

A deterioration model built near specialists behaves differently when the nearest one is an island away.

Governance has to account for the system, not just the tool.

[Temitayo Bewaji @drtemib]

You'd never accept a colleague using an intervention they can't explain or describe the failure modes of.

"AI" is not a reason to lower that standard.

The scrutiny comes before the decision. Not after.

[Temitayo Bewaji @drtemib]

A clinical summary that omits a pending result looks complete.

The reviewing clinician can't detect what's missing from inside the summary.

That's why AI governance has to catch the gap before the tool goes live.

Thoughts?

[Temitayo Bewaji @drtemib]

The alert that fires wrong is visible. The one that never fires is silent.

Automation bias means a quiet screen can move a patient down your list with no one noticing.

That is a clinical decision. Govern it like one.

#AIGovernance

[Temitayo Bewaji @drtemib]

At 20 messages a day, a human reviews every chatbot reply.

At 200, review becomes click-and-accept.

That is where AI governance breaks down. Not in the clinic. In the volume.

Thoughts?

[Temitayo Bewaji @drtemib]

"Administrative" AI still makes clinical decisions.

A scheduling system that bumps the wrong patient down the list just made a choice with clinical consequences. No white coat in the room.

Governance belongs at rung 3. Not just rung 5.

[Temitayo Bewaji @drtemib]

"95% accurate", which column does that go in?

Demonstrated. Stated. Open.

A vendor showing you something is not the same as it working in your environment.

Ask all three before you govern any AI touching patient care.

[Temitayo Bewaji @drtemib]

AI gives you the draft without the person.

The confidence without the track record.

A junior colleague, you know what they read, what they miss, and when to stop asking them.

With AI, you get none of that. Go to the source. Every time.

[Temitayo Bewaji @drtemib]

"Can I put patient data in this tool?" is the wrong first question.

Ask this instead: do you know where that data goes?

Storage. Access. Breach response. Every AI rung demands the same answer.

Governance starts with knowing, not assuming.

[Temitayo Bewaji @drtemib]

"Are you using AI?" is too broad a question.

Ask this instead: How directly can this tool shape what happens to a patient?

That's where governance starts.

[Temitayo Bewaji @drtemib]

Start with demand, setting, role, and workflow before any system shortlist. Vendor defaults follow clinical need, not the reverse. Audit where scarce clinical time drains into unnecessary work, rework, and avoidable admin, then redesign one high-friction workflow.

[Temitayo Bewaji @drtemib]

Technology frees clinical capacity by removing work entirely. Audit where time disappears into rework: auto-fill verification, low-signal alerts, referral gaps chasing. Find your highest-friction workflow and redesign it.

[Temitayo Bewaji @drtemib]

Workflow design is a retention strategy. Avoidable burden, unclear handoffs, and duplicated admin wear clinical staff down shift by shift. Scarce labour makes badly designed work accelerate attrition. Audit where clinical time is lost, then redesign one high-friction workflow.

[Temitayo Bewaji @drtemib]

Skill mix is a genuine workforce strategy. Highly trained staff doing work a supported role could safely handle quietly shrinks clinical capacity. Audit where scarce clinical time goes, then redesign one high-friction workflow.

[Temitayo Bewaji @drtemib]

Rework costs clinical time your team cannot spare.

Incomplete referrals, wrong bookings, and missed handoffs force staff to repeat necessary work.

Standardize the referral. Fix the handoff. Then audit one high-friction workflow and redesign it.

[Temitayo Bewaji @drtemib]

Scarcity raises the cost of badly designed work. A duplicated form or failed handoff wastes time that now competes directly with patient care. Fix recruitment and fix the workflow. Audit where clinical time drains into rework, then redesign one high-friction process.

[Temitayo Bewaji @drtemib]

The global health worker shortfall was 15.4M in 2020. That crisis is already here. Audit where scarce clinical time drains to unnecessary work, rework, poor role design, and admin burden, then redesign one high-friction workflow.

[Temitayo Bewaji @drtemib]

Treat patient transparency as a provider obligation, separate from product approval. Pointing to regulatory clearance does not discharge it.