# Alvin Omozokpia

> AI product manager turning ambiguous opportunities into useful, measurable, and trustworthy products.

The portfolio of Alvin Omozokpia, AI Product Manager. Case study, About and Resume links point to Markdown copies generated from the same published content as their pages; drop the `.md` for the HTML page. [SKILL.md](https://alvn.io/SKILL.md) summarizes how Alvin Omozokpia works and the published evidence for it, with a link to every source.

## Pages

- [Home](https://alvn.io/): Introduction, featured case studies and articles, and operating principles.
- [Work](https://alvn.io/work): AI products and experiments led and built by Alvin Omozokpia.
- [Blog](https://alvn.io/blog): Notes and case studies on AI product management, growth, and product craft.
- [About](https://alvn.io/about.md): About Alvin Omozokpia, an AI product manager based in Los Angeles.
- [Resume](https://alvn.io/resume.md): Experience and capabilities of AI product manager Alvin Omozokpia.
- [Release Gatekeeper](https://alvn.io/evals): Interactive walkthrough: four AI release decisions, scored against what shipped.

## Case studies

- [Bragora](https://alvn.io/work/bragora-empowering-career-growth-through-ai-enriched-achievement-tracking.md): An AI-powered career workspace for capturing achievements, managing job applications, and turning work history into actionable insight. Visit: https://bragora.com
- [Alvin’s Library](https://alvn.io/work/alvins-library.md): A walkable 3D library with 628 shelved books, a catalogue-grounded keeper, and a searchable view of all 969 editions. Visit: https://library.alvn.io
- [Living Fieldbook](https://alvn.io/work/living-fieldbook.md): A daily practice for product judgment: observe something real, keep it apart from what you think it means, and watch your beliefs change in a living habitat. Visit: https://fieldbook.alvn.io
- [Product Lab](https://alvn.io/work/product-lab.md): Three mobile-first simulations that turn experiment calls, stakeholder tradeoffs, and AI exception handling into playable product decisions. Visit: https://alvns-productlab.vercel.app
- [Release Gatekeeper](https://alvn.io/work/evals-walkthrough.md): An interactive walkthrough of a real AI release pipeline: four decisions, real evidence, and a verdict scored against what actually shipped. Visit: https://alvn.io/evals
- [Tic Tac Tile](https://alvn.io/work/tic-tac-tile.md): A lightweight web game exploring adaptive AI difficulty, fast performance, accessibility, and short-session retention. Visit: https://tictactile.lovable.app

## Articles

- [Your AI Rollout Needs a Review Budget](https://alvn.io/blog/your-ai-rollout-needs-a-review-budget): Strategy, 2026-10-18. More drafts can mean more waiting. A five-case capacity model shows what an AI rollout must prove before it promises faster delivery.
- [No Watermark Is Not Proof](https://alvn.io/blog/no-watermark-is-not-an-authenticity-check): Trust, 2026-10-11. A media-provenance result can describe a file without settling the claim it supports. Six refund-photo cases show where a product must keep those decisions separate.
- [Your Agent Can Win the Wrong Deal](https://alvn.io/blog/your-agent-can-win-the-wrong-deal): Strategy, 2026-10-04. Before improving an agent's negotiating skill, check whose preferences it is optimizing. A five-book example shows how 90% agreement can still miss the choice that matters.
- [The Call Can End Before the Work Does](https://alvn.io/blog/the-call-can-end-before-the-work-does): Shipping, 2026-09-27. A voice agent can stop speaking while its last action is still unresolved. Three interruption replays show what a product must preserve before work continues in text.
- [My Library Has a Genie in It](https://alvn.io/blog/let-the-search-decide-what-exists): Building, 2026-09-26. Books, candlelight, a brass lamp, and the fun of building a little world you can walk into.
- [When the Tutor Leaves the Room](https://alvn.io/blog/when-the-tutor-leaves-the-room): Strategy, 2026-09-20. AI can make a study session feel successful without showing what the learner can do alone. A worked assessment brief separates useful assistance from a credible learning claim.
- [Your AI Analyst Needs to Name the Denominator](https://alvn.io/blog/your-ai-analyst-needs-to-name-the-denominator): Strategy, 2026-09-13. Correct arithmetic can answer the wrong business question. A reproducible activation example shows when an AI analyst should use a governed definition, compare interpretations, or ask.
- [The System Prompt Is Not a Permission Boundary](https://alvn.io/blog/the-system-prompt-is-not-a-permission-boundary): Trust, 2026-09-06. If the prompt says one thing but the runtime can reach another, the runtime wins. Consequential agents need a machine-verified environment receipt before they start.
- [Your Agent Eval Should Survive a Vendor Swap](https://alvn.io/blog/your-agent-eval-should-survive-a-vendor-swap): Shipping, 2026-08-30. Models and frameworks will change. The task, failure rules, trace evidence, and release threshold should travel together, or every migration resets what the team knows.
- [Teardown: The Leaderboard Is a Product. You Are Not the User.](https://alvn.io/blog/teardown-leaderboard-is-a-product): Teardowns, 2026-08-17. A 16-month study of Chatbot Arena found private variant testing, selective score disclosure, and unequal data access shaping the rankings teams treat as ground truth. The teardown lesson: leaderboards gate attention, and your release pipeline has to gate everything else.
- [Decision Doc: A PRD Where the Core Feature Is Saying No](https://alvn.io/blog/decision-doc-refusal-prd): Decision Docs, 2026-08-16. The sanitized PRD for teaching a compliance-bound research assistant to refuse low-confidence requests, which cut unsupported claims from 14 percent to 1.8 percent at the cost of a 9 percent refusal rate.
- [Decision Doc: The Experiment Brief That Replaced Our Offline Evals](https://alvn.io/blog/decision-doc-shadow-mode-eval-brief): Decision Docs, 2026-08-16. The actual experiment brief I wrote when our offline evaluation scores turned out to have near-zero correlation with production outcomes, and the shadow-mode replay design that replaced them.
- [Decision Doc: The Memo That Stopped a CTO-Sponsored AI Initiative](https://alvn.io/blog/decision-doc-ship-or-hold-memo): Decision Docs, 2026-08-16. The actual hold memo I wrote after a three-week feasibility test found a 71 percent accuracy ceiling against a 95 percent requirement, and how the team was redirected without losing the sponsor.
- [Teardown: AI Search Ships Answers. The Missing Feature Is Doubt.](https://alvn.io/blog/teardown-ai-search-confidence-layer): Teardowns, 2026-08-16. Eight AI search engines answered wrong more than 60 percent of the time on a basic citation task, and almost never declined to answer. The teardown lesson for product teams is that abstention is a feature you have to build, budget, and own.
- [A Product Note Is Not Evidence Yet](https://alvn.io/blog/a-product-note-is-not-evidence-yet): Strategy, 2026-07-26. A note becomes useful when the observation, interpretation, and affected assumption stay connected. Living Fieldbook is a prototype for making that chain visible.
- [When Helpful Memory Becomes Surveillance](https://alvn.io/blog/when-helpful-memory-becomes-surveillance): Trust, 2026-07-20. AI memory earns its place when people can see what was retained, why it matters, where it is used, and how to change or remove it.
- [The Exception Queue Is the Product](https://alvn.io/blog/the-exception-queue-is-the-product): Trust, 2026-07-16. Human review is not a safety feature by itself. The queue must show why a case arrived, what could go wrong, who can act, and what the system should learn from the decision.
- [The Compliance Boundary PMs Must Draw Before Roadmapping](https://alvn.io/blog/the-compliance-boundary-pms-must-draw-before-roadmapping): Trust, 2026-07-13. A career assistant that helps a person tell their story and a system that ranks people for an employer may use similar technology, but they create different product, evidence, and regulatory responsibilities.
- [The AI Feature Ship Gate](https://alvn.io/blog/the-ai-feature-ship-gate): Shipping, 2026-05-12. An eval should behave like an executable product requirement: it defines the expected outcome, the unacceptable failures, and the evidence required to launch.
- [Adoption Is Not an AI Value Metric](https://alvn.io/blog/adoption-is-not-an-ai-value-metric): Shipping, 2026-03-24. Adoption shows that people used an AI feature. This metric tree tests whether the feature completed useful work at an acceptable cost and risk.
- [The Cost of a Correct Answer](https://alvn.io/blog/the-cost-of-a-correct-answer): Shipping, 2026-01-13. AI economics should be measured through correct outcomes, recovery burden, latency, and reliability, not a token price or unit-cost headline in isolation.
- [The Agent Autonomy Budget](https://alvn.io/blog/the-agent-autonomy-budget): Trust, 2025-11-18. A practical ladder for deciding what an AI agent may recommend, draft, or do, before convenience turns into an unacceptable product risk.
- [The Incident-Ready AI Product Manager](https://alvn.io/blog/the-incident-ready-ai-product-manager): Trust, 2025-10-07. When an AI feature fails in production, the PM needs a severity model, containment decision, user remedy, replay path, and a way to turn the failure into a stronger ship gate.
- [Red-Team the Product, Not Just the Model](https://alvn.io/blog/red-team-the-product-not-just-the-model): Trust, 2025-08-19. A model can refuse a malicious prompt in a lab and still sit inside a product that leaks data through tools, permissions, identity, or recovery gaps.
- [Your AI Product Is Only as Ready as Its Data](https://alvn.io/blog/your-ai-product-is-only-as-ready-as-its-data): Shipping, 2025-07-01. Choose the use case before the model. This Data Readiness Contract makes purpose, provenance, permissions, quality, freshness, evaluation, retention, and ownership explicit.
- [Designing AI That Knows When to Stop](https://alvn.io/blog/designing-ai-that-knows-when-to-stop): Trust, 2025-05-13. A trustworthy AI experience does more than display confidence. It helps people inspect evidence, recognize uncertainty, stop safely, and recover when the system is wrong.
- [A Product Why Must Change a Decision](https://alvn.io/blog/product-why-must-change-a-decision): Strategy, 2025-04-10. Purpose is useful only when it changes scope, sequencing, evidence, or a stop decision. This brief turns an inspiring why into an operating constraint a product team can actually use.
- [Building Bragora Around User Control](https://alvn.io/blog/building-bragora-around-user-control): Shipping, 2025-03-19. Bragora began with a simple product problem: useful career evidence disappears when people need it most. Building the product taught me to treat capture, reuse, and AI control as one connected workflow.
- [The Art of the Ask: Prompting as a Product Discipline](https://alvn.io/blog/prompting-as-a-product-discipline): Shipping, 2025-03-10. The best prompts are not clever incantations. They are small product specifications that define the task, context, constraints, evidence, and acceptance test before an AI system begins.
- [Duolingo Designed for the Return, Not Just the Streak](https://alvn.io/blog/duolingo-designed-for-the-return): Strategy, 2025-02-28. Duolingo's product lesson is deeper than gamification. Its habit system reduces the cost of starting, makes continuity visible, creates recovery paths, and keeps the learning outcome close enough to matter.
- [Product-Led Growth Starts After Sign-Up](https://alvn.io/blog/product-led-growth-starts-after-sign-up): Strategy, 2025-02-18. A free tier can generate accounts without creating a growth engine. Product-led growth works only when users reach value, return for it, and encounter an expansion path that fits the job they are already doing.
- [AI Can Do PM Tasks Without Owning the Product Decision](https://alvn.io/blog/ai-pm-tasks-without-owning-the-decision): Strategy, 2025-01-14. The useful question is not whether AI can write a PRD or analyze feedback. It is which product tasks can be delegated, what evidence the system needs, and who remains accountable for the decision.
- [Faster Prototypes Need Better Evidence](https://alvn.io/blog/faster-prototypes-need-better-evidence): Shipping, 2024-12-16. Generative AI can make a prototype cheap enough to build before a meeting. That speed is useful only when the team is explicit about which assumption the artifact tests and what evidence would change the roadmap.

## Feeds and agent files

- [Sitemap](https://alvn.io/sitemap.xml): Every public page, for crawlers.
- [SKILL.md](https://alvn.io/SKILL.md): How Alvin Omozokpia works and the published evidence for it.

## Connect

- [LinkedIn](https://linkedin.com/in/alvinomozokpia/): alvinomozokpia
- [Email](mailto:hello@alvn.io): hello@alvn.io
- [Book a call](https://calendly.com/hello-alvn/30min): 30-minute intro
- [Resume](https://alvn.io/resume): Experience and skills