Practero

Practero turns conflicting requirements, field realities, and technical constraints into a traceable deployment path for forward-deployed teams

Comment

Inspiration

AI capabilities are advancing rapidly, but getting technology to work inside real organizations remains difficult.

Through my experience deploying software across transport, fintech, public-sector systems, operations, and client projects, I repeatedly encountered three different versions of reality:

  1. what leadership and project requirements say should happen;
  2. what users and frontline teams actually experience;
  3. what the existing technology can support.

These realities are usually scattered across meeting notes, requirements, interviews, technical documentation, and project-management tools. The contradictions are often discovered late—during testing, deployment, or adoption.

I built Practero to make those gaps visible before they become expensive implementation failures.

What it does

Practero is an open implementation workspace for forward-deployed engineers, applied AI teams, technical consultants, and deployment specialists.

It structures evidence across three categories:

  • organizational intent;
  • field reality;
  • technical reality.

Practero then identifies Reality Gaps such as contradictions, unsupported assumptions, missing capabilities, operational risks, adoption risks, and unresolved decisions.

Every finding remains connected to its exact supporting evidence. Validated gaps are converted into:

  • prioritized implementation workstreams;
  • engineering and stakeholder actions;
  • dependencies;
  • acceptance criteria;
  • risks and mitigations;
  • pilot-readiness checks;
  • an Executive Brief for decision-makers.

The submission includes MetroMove, a completely fictional urban-mobility pilot that demonstrates the full workflow.

Its central finding is a connectivity contradiction: management requires continuous ride updates, drivers report connectivity outages lasting ten to twenty minutes, and the current architecture has no offline queue or reconnect synchronization.

Practero connects the three sources, classifies the deployment risk, recommends offline event queuing, and creates a measurable acceptance criterion.

How I built it

Practero is built with Next.js, React, TypeScript, Tailwind CSS, and Zod.

The domain layer models engagements, evidence sources, evidence references, Reality Map items, Reality Gaps, workstreams, risks, stakeholder actions, acceptance criteria, and executive briefs.

Severity, readiness, evidence strength, and gap ordering use deterministic policies rather than unexplained model-generated scores.

The MetroMove fixtures are validated against the same Zod schemas planned for future live analysis. Every evidence reference includes its source, chunk, exact excerpt, offsets, category, and content identifier.

The application was developed throughout OpenAI Build Week using Codex with GPT-5.6 Sol. Codex helped with product planning, architecture, implementation, accessibility improvements, deterministic policies, documentation, tests, recovery from an interrupted development session, and production validation.

Challenges

The largest challenge was keeping the product focused.

Practero could easily become a project-management system, document chatbot, or generic consulting workspace. I kept the MVP centred on one critical chain:

evidence → exact reference → validated gap → implementation action.

A second challenge came from slow connectivity interrupting dependency installation and the original Codex task. I preserved the work through Git, resumed it in Codex Desktop, audited the existing implementation, repaired the remaining issues, and completed the production build.

Accomplishments

I am proud that the submission provides:

  • five fictional evidence sources across all three realities;
  • twelve structured and traceable Reality Map items;
  • six prioritized Reality Gaps;
  • four linked implementation workstreams;
  • exact excerpt-level traceability;
  • deterministic severity and readiness policies;
  • twenty-one passing unit tests;
  • a successful production build generating twenty-one pages;
  • a public judge-friendly demonstration requiring no account.

What I learned

I learned that AI-assisted implementation becomes much more dependable when the human defines a clear finish line, strong schemas, explicit constraints, and verifiable quality gates.

I also learned that source traceability is essential. A convincing AI-generated recommendation is not enough for high-stakes deployment work- the team must be able to see exactly why the recommendation exists.

What's next

The current Build Week submission uses clearly labelled fictional sample analysis and does not make runtime model API calls.

The next milestones are:

  • Firebase authentication and persistence;
  • user-created engagements;
  • document and field-note ingestion;
  • GPT-5.6-powered evidence extraction;
  • cross-source Reality Gap analysis;
  • collaborative implementation workflows;
  • GitHub and enterprise-tool integrations.

Practero is open source under the Apache 2.0 licence.

Built With

Share this project:

Updates