AI Product Leader — Lagos · London · Remote
Six-plus years turning machine-learning research into revenue-bearing products for enterprises — from first model to board-level outcomes.
Cut claims-processing time 68% with an LLM triage copilot for a top-5 insurer.
A top-5 insurer was drowning in claims backlog, with adjusters spending most of their day routing rather than deciding.
Led product end-to-end: business case, eval framework, human-in-the-loop design, and phased rollout across three claim lines.
68% faster processing, 12-pt CSAT lift, and a regulator-ready audit trail adopted as the company standard.
Lifted underwriting accuracy 22 points with a risk model approved by regulators in three markets.
Legacy rules-based underwriting was mispricing risk and losing profitable segments to digital-first competitors.
Owned model productization, explainability requirements, and the regulatory approval pathway across three markets.
+22 pts accuracy, approval in all three jurisdictions, and a 9% improvement in loss ratio within the first year.
Scaled a multilingual banking assistant to 2.4M monthly users across West Africa.
A leading bank needed to serve customers in four languages without quadrupling its support organization.
Defined the containment-vs-handoff strategy, multilingual eval suite, and the growth loop from USSD to app.
2.4M MAU, 71% containment rate, and support cost per interaction down 55%.
Automated 80% of KYC document review for a pan-African bank — 40,000 analyst-hours saved a year.
Manual KYC review was the single largest bottleneck in onboarding, stretching account opening to eleven days.
Shaped the platform play: one extraction pipeline serving nine internal teams, with confidence-based routing to humans.
80% straight-through processing, onboarding cut to same-day, 40,000 analyst-hours redeployed annually.
Reduced retail stock-outs 31% with a forecasting suite rolled out to 1,200 stores.
A regional retail group was losing sales to stock-outs while simultaneously over-holding slow inventory.
Ran the pilot-to-platform transition — forecast UX for non-technical planners, override telemetry, and the scale playbook.
31% fewer stock-outs, 18% less excess inventory, live in 1,200 stores in nine months.
Gave clinicians 90 minutes back per shift with an ambient documentation product piloted in 12 hospitals.
Clinician burnout was driven by documentation load — hours of note-writing after every shift.
Took the product 0→1: clinical safety review process, physician champion program, and the pilot design across 12 hospitals.
90 minutes returned per clinician per shift, 94% note-acceptance rate, expansion approved system-wide.
Blocked $14M in fraud in year one with a real-time transaction-graph model.
Fraud rings were exploiting gaps between siloed rule engines across payment channels.
Defined the graph-based detection product, false-positive budget with the risk team, and the real-time serving requirements.
$14M fraud blocked in year one with false positives held under 0.8% — below the agreed customer-friction budget.
A working belief
The hard part of AI was never the model. It's earning the trust to put it in front of a customer.
Every model I ship starts as a P&L line, not a demo. I write the value hypothesis first — what changes, for whom, measured how — and kill projects that can't answer it. It's why my products survive budget season.
I treat evaluation the way great teams treat design: continuously, visibly, and owned. Regulators, executives, and users all trust a system they can see being measured — so I make the measurement part of the product.
I put a narrow slice in front of real users early, instrument everything, and scale only what the data defends. Six years in, the pattern holds: the fastest route to a big AI product is a small honest one.
Portrait placeholder — swap for a B&W or navy/orange duotone photograph.
TemiTope Kasali builds AI products that survive contact with the real world — regulators, unit economics, and users included.
Over six-plus years she has led product for machine-learning systems across insurance, banking, retail, and healthcare, taking seven products from first hypothesis to production scale. Her work has processed millions of customer interactions, cleared regulatory review in multiple markets, and delivered more than $40M in measured business impact.
She is known for translating between boardrooms and model teams: sharp on the economics, fluent in the technology, and unwilling to ship anything she can't measure. Executives bring her in when an AI initiative has to actually work.
“Temi is the rare product leader who can defend a model architecture in the morning and a business case to the board in the afternoon. We stopped debating AI strategy and started shipping it.”Chief Data Officer Top-5 insurance group
“She turned our most politically stuck initiative into the company's reference case for AI. Twice.”Chief Executive Officer Pan-African financial services group
If you're weighing an AI initiative that has to deliver — not demo — the conversation is worth an hour.
hello@temitopekasali.com