Ola Sobo
Senior product leader
Search portfolioSearch

Search my portfolio

Search cases, decisions, results and experience.

Browse all casesAsk the Evidence Guide

Experience snapshot

My experience, in view.

I lead enterprise workflow products at HCA Healthcare. My career connects operating analysis, product ownership and platform delivery.

Download résumé Discuss a role

I use generalized enterprise workflows and fictionalized examples to explain product decisions.

Nashville, Tennessee

Read the full record

My experience, in view.

I lead enterprise workflow products at HCA Healthcare. My career connects operating analysis, product ownership and platform delivery.

My current work spans platform strategy, migration and monthly delivery. Across roles, I have managed up to 10 direct reports and provided matrix leadership for up to 20 contributors.

I lead a production healthcare worklist platform and have shaped delivered payer workflows that make the next action and outcome visible. I am applying that experience to MIN–MAX Living, my private renter-decision prototype. It keeps source quality and uncertainty visible. Renter research and an external beta are next.

I used ChatGPT, Copilot, Redgate SQL Monitor and Dynatrace to focus the performance investigation on high-impact queries and actual search behavior. The team UAT comparison recorded average request response time falling from 0.910 seconds on 28 June to 0.228 seconds on 6 July 2026, a 74.9% reduction, with 1,000 virtual users and 80 CPUs recorded. I shaped the priorities and release decisions; this team result does not isolate the effect of AI assistance.

I built and released the Evidence Guide through agent-assisted implementation, read-only evidence tools, source validation and controlled evaluations. My experience snapshot links these responsibilities to the cases, implementation records and full résumé.

Professional experience

My scope, over time.

Select a role to inspect my responsibilities and the work behind them.

2012–2015PolarisBusiness Analyst2015–2018PolarisTechnical BA / Product Owner2020–2023NikeProduct Manager2023–presentHCA HealthcareData Product Manager
Business educationEconomics BSc · 2009 MBA · 2019 Business Analytics master’s · 2020
Operating analysis

Business Analyst

June 2012 to June 2015

I translated operating trends into reporting, data-quality and process-improvement requirements for financial services.

I connected business needs with the information and processes required to act on them.

My career record
Product requirements + analytics

Senior Technical Business Analyst / Technical Product Owner

June 2015 to August 2018

I turned business and technical requirements into prioritized roadmaps and user stories for workflow, data and platform improvements.

I used SQL, Python and analytics to clarify workflow and data requirements and improve decision support.

My career record
Enterprise delivery + adoption

Product Manager

March 2020 to November 2023

I coordinated enterprise roadmap execution and release readiness across business and technical teams.

I connected UAT, issue triage, product analytics and enablement to release quality and adoption.

Delivery, quality and adoption My career record
Platform strategy + recurring delivery

Data Product Manager, DNA/Helix

November 2023 to present

I lead product strategy and monthly delivery for DNA/Helix, guiding legacy migration and expanding a shared workflow platform across operational teams.

I connect reusable workflow controls, API requirements, team adoption and release readiness.

Platform and migration My career record

Enterprise work + independent building

Working range.

I own product direction, evidence standards and release decisions. Independent implementation and testing were agent-assisted.

Build AI systems

I built the Guide around model-directed evidence selection and application-owned validation.

Context · Model selection · Tool contracts · Recovery

Evidence GuideReleased · 7 October 2026
Use the Guide My implementation contribution
Context, tools and recovery

I keep question context within the request and bind selections to approved source identities and withdrawal status.

I expose read-only MCP search and fetch through a separate assistant connection. Application validation and source-only recovery keep evidence available when a model selection cannot be used.

Connect the MCP tools Run the historical reference

Govern and lead

I apply production healthcare delivery practice to MIN–MAX, my private renter prototype, with distinct responsibilities for builders and independent reviewers.

Policy · Release authority · Audit · People

MIN–MAX LivingPrivate prototype · Delivery and release model

Next for MIN–MAX: renter research and external beta.

Explore the product case
People leadership · Across roles
Direct reports
Up to 10Average six
Matrix contributors
Up to 20
Junior Product Owners hired
3
Employees coached
15
See how I lead
Full career recordContributions, education and certifications

Where I have done the work.

November 2023 to present

HCA Healthcare

Data Product Manager, DNA/Helix

Enterprise workflow platforms, payer automation and applied AI practice.

Inspect my contribution
  • Lead product strategy and monthly delivery for DNA/Helix, expanding a production workflow platform across operational teams and guiding migration from legacy applications.
  • Designed a reusable workflow model for eligibility, prioritization, ownership and exception recovery, adapting shared controls to distinct team workflows.
  • Led product definition and initial delivery of payer-response automation, with an internally reported estimate of more than 5,000 staff hours saved annually following implementation in 2025.
  • Redirected performance investigation toward high-impact SQL queries and actual search behavior; team UAT tests recorded 74.9% lower average request response time with 1,000 virtual users in June–July 2026.
  • Defined API and interface requirements across payer, clinical-documentation and EHR workflows, making acknowledgements, exceptions and final outcomes visible across systems.
  • Established release controls using feature flags, end-to-end validation and post-launch monitoring; internal UAT reporting recorded 98% functionality compliance.
Platform and migration Payer automation Performance and test conditions
March 2020 to November 2023

Nike

Product Manager

Enterprise delivery, release quality and adoption. Team results are internally reported.

Inspect my contribution
  • Coordinated roadmap execution and release readiness across business and technical teams; 95% of approved platform enhancements were delivered on time.
  • Facilitated UAT and issue triage to improve release quality; the enterprise delivery scope reported 20% fewer post-implementation defects.
  • Connected product analytics, user feedback and enablement to adoption improvements; the enterprise delivery scope reported a 20% increase in product adoption.
Delivery, quality and adoption
June 2015 to August 2018

Polaris Limited

Senior Technical Business Analyst / Technical Product Owner

Product ownership, analytics and operational improvement.

Inspect my contribution
  • Turned business and technical requirements into prioritized roadmaps and user stories for workflow, data and platform improvements.
  • Used SQL, Python and analytics to clarify workflow and data requirements and improve decision support.
June 2012 to June 2015

Polaris Limited

Business Analyst

Financial services, reporting and process improvement.

Inspect my contribution
  • Analyzed operating trends and translated business needs into reporting, data-quality and process-improvement requirements for financial services.
Read the full résumé Download résumé

Where I can contribute

I’m interested in Senior and Lead Product Manager roles involving enterprise platforms, workflow automation, data products and AI-assisted operations.

My experience combines platform strategy, API and data fluency, initial feature delivery, migration and formal people management. The flagship cases show the choices and artifacts behind that scope.

See how I lead

Education and certifications

March 2020

Master's in Business Analytics

Hult International Business School, San Francisco

September 2019

MBA

Hult International Business School, Boston

June 2009

Bachelor of Science in Economics

Olabisi Onabanjo University

Certified ScrumMaster (CSM) · Certified Scrum Product Owner (CSPO)

Four ways to inspect the work

Production platform

Enterprise migration

Shared controls, team-specific workflows and a controlled move from legacy applications.

Delivered AI-assisted practice

Performance investigation

Evidence synthesis, findability and the choice to keep emergency release scope bounded.

Independent system

Evidence Guide evaluation

A same-question comparison, useful-answer criteria and a decision about model cost.

Practice reflection

Agent-assisted delivery

How I think about specialization, handoffs and the cost of review.

Decision and implementation detailExisting comparisons, system walkthrough and evidence

Enterprise delivery → applied AI

Three decisions.
One way of working.

I identify where to intervene, make the constraints explicit and test whether the solution earns its complexity.

01 / Focus the investment

Delivered practice · Team UAT result

I changed where
the team looked.

The team had worked on recurring performance issues for about three years. I connected the highest-impact queries to how people actually searched.

Account search was reported to consume 40% of CPU. I found joins and filters that did not serve the user’s task, using Copilot, ChatGPT and Redgate SQL Monitor to structure the investigation.

Optimize
Faster searches that still let people find the right record.
Protect
Record findability, read/write behavior and a bounded release.
My decision
Prioritize high-impact queries and challenge unnecessary query behavior against actual use.
Tradeoff
An exact-match query can be faster and harder to use. I balanced hotfix relief with the longer-term investigation.
What the team measured

Same reported load.
A different response.

74.9%lower average request response time
in the documented UAT comparison

1,000 virtual users · Test configuration: 80 CPUs
Source: July 2026 internal performance update.

Query impactUser behaviorControlled test
Continue the connectionNext: protecting priority

02 / Protect the priority

Correction working in production

I put priority
before fit.

A seemingly efficient batch could use the only qualified reviewer on easier work, leaving an urgent restricted case waiting.

I connected the UAT failure to the decision hierarchy: protect the priority outcome first, then compare valid matches. The corrected workflow is working in production within its operating constraints.

Optimize
Cover time-sensitive work with eligible reviewers.
Protect
Qualification, availability and capacity remain binding.
My decision
Make business priority the first comparison among eligible allocations, before geographic fit and match preferences.
Tradeoff
Accept a less convenient valid match to protect urgent work. Lower-priority work stays visible when capacity is full.
Fictional reconstruction

The same capacity.
A different decision.

Three reviewers. One slot each. Four cases.

Fit-led batch

1/2 cases due today assigned · 3/4 overall

R1 waits while tomorrow’s general case uses a slot.

  1. R1Restricted · TodayWaiting for capacity
  2. R2General · TodayAssigned to Mara
  3. R3Licensed · TomorrowAssigned to Theo
  4. R4General · TomorrowAssigned to Jules
  1. MaraLicensed + pediatric1 slot
  2. TheoLicensed1 slot
  3. JulesGeneral work1 slot

Priority first

2/2 cases due today assigned · 3/4 overall

R1 reaches the only reviewer with both qualifications. R4, due tomorrow, waits.

  1. R1Restricted · TodayAssigned to Mara
  2. R2General · TodayAssigned to Jules
  3. R3Licensed · TomorrowAssigned to Theo
  4. R4General · TomorrowWaiting for capacity
  1. MaraLicensed + pediatric1 slot
  2. TheoLicensed1 slot
  3. JulesGeneral work1 slot

Illustrative records and allocations. Production status confirmed by me, 8 October 2026.

From operating constraints to AI valueNext: testing the extra AI step

03 / Test the added value

Historical paired experiment · Agent-scored

I made the extra
AI step earn its place.

A model selector could rescue an answer and spoil another. I tested whether the complete task improved enough to justify the extra step.

I compared both paths on the same 120 questions. The selector recovered seven answers and degraded eight. I recommended source-only for that candidate and focused subsequent work on request references, evidence applicability and answer composition.

Optimize
A useful first answer with a source that supports the requested claim.
Protect
Reviewed evidence, source identity, read-only tools and a bounded execution budget.
My decision
Keep the simpler path as the recommendation for this experiment. Repair the application contract before adding more complexity.
Tradeoff
The provider spend was small. The added delay and absence of a net usefulness gain were the stronger reasons to keep the simpler path.
Same questions, two paths

Did the extra step
improve the task?

Useful first answers of 120 questions

Source-only96 / 120
Model selector95 / 120

Local p95 latency 0–2,000 ms scale

Source-only224 ms
Model selector1,858 ms
Provider cost / full run$0 vs $0.0389

Frozen build 06226e4 · Independent agent scoring
Warm-process local timings; provider cost for all 120 questions.

Inspect the recorded measurements
Historical paired experiment
MeasureSource-onlySelector
Useful answers96/12095/120
Local p50138 ms1,172 ms
Local p95224 ms1,858 ms
Provider attempts0111
Provider cost$0$0.0388617

Both paths had to reach 41 useful answers in each 60-question set. The selector also needed positive combined gain and no lower score in either set.

Latency excludes browser rendering, deployed network transit and cold startup. Provider cost excludes hosting, engineering and review. Later configurations have separate evaluation records.

From the investment decision to implementationNext: inside the product

Inside the product

One question.
Explicit responsibilities.

I built the Guide around reviewed evidence and controlled tool access. This walkthrough follows its released architecture; model selection is an optional path.

“What did I change in the performance investigation?”

Guide architecture walkthrough · Released 7 October 2026

01ReceiveQuestion + approved contextApplication

I preserve the original request, approved evidence references and execution budget. Context is request-local; the Guide does not retain persistent visitor memory.

02SelectEvidence + allowed toolsOptional model path

I allow the model to select approved evidence and permitted read-only operations. The public MCP service exposes search and fetch; it does not publish changes.

03ValidateRequest + source identityApplication

I require selections to match the request, an approved source and its status. Application checks preserve contribution, maturity and measurement context; a failed selection has a source-only recovery path.

04PresentReviewed facts + exact sourceApplication + my release authority

I keep factual wording in reviewed records and application code. The answer exposes its source and any specific evidence gap. I retain evidence approval, withdrawal and release authority.

The local Guide uses reviewed sources. Model-assisted answers open the published Guide, as configured for this preview.

Try the Guide Inspect the architecture Inspect the controls and risks Open the execution traces Connect the read-only tools Inspect the delivery contracts

The reusable foundation

I use shared controls for state, eligibility, timing, ownership and recovery, then adapt behavior to the team’s workflow. That pattern connects my enterprise platform work to my independent AI building.

See the shared workflow controls

Enterprise product delivery

From first delivery
to a shared platform.

I bring the same decisions into real operating workflows: define the outcome, connect the systems and improve what happens after release.

Payer automation · Initial delivery in 2025

5,000+ hours

Estimated annual effort saved

I led workflow definition and initial delivery, joining payer events with staff actions. The annual estimate was circulated internally after implementation.

Production healthcare platform

Multiple teams.
A common foundation.

I lead product strategy and monthly delivery for DNA/Helix. I guided migration from legacy applications and expanded shared orchestration into team-specific workflows.

Leadership record

Inspect my leadership scope

Continue the story

Let’s discuss the responsibility ahead.

I’m interested in senior and lead product roles spanning enterprise platforms and applied AI.

Discuss a senior product role Inspect the leadership examples
Sources and measurement

Healthcare renter contrast

I lead a production healthcare worklist platform and have shaped delivered payer workflows that make the next action and outcome visible. I am applying that experience to MIN–MAX Living, my private renter-decision prototype. It keeps source quality and uncertainty visible. Renter research and an external beta are next.

Recruiter ai title

AI-assisted product operations

I used ChatGPT, Copilot, Redgate SQL Monitor and Dynatrace to focus the performance investigation on high-impact queries and actual search behavior. The team UAT comparison recorded average request response time falling from 0.910 seconds on 28 June to 0.228 seconds on 6 July 2026, a 74.9% reduction, with 1,000 virtual users and 80 CPUs recorded. I shaped the priorities and release decisions; this team result does not isolate the effect of AI assistance.

I built and released the Evidence Guide through agent-assisted implementation, retrieval, read-only MCP tools and controlled evaluations. I use the enterprise performance case to show applied product judgment and the Guide to show implementation and evaluation decisions.