← Tim Vasil

AI-native tearsheet

Structuring Product Teams in the Age of GenAI

Small, senior, cross-functional teams use GenAI as elastic capacity, not as a separate function. The company paves the road for safe agentic work; each team decides how to drive on it.

By Tim Vasil •

1Team shape

The unit of delivery is a small, senior, cross-functional team that owns a product area end to end. Agents extend the team’s capacity; they do not replace its judgment.

2The paved road: a company-wide AI platform

Product teams should not assemble their own AI stack. Provide common primitives, guardrails, and defaults centrally, then let teams customize their workflows on top.

Platform constructProvided centrally
Models & interfacesApproved models; chat, IDE, and API access; model routing where it pays; enterprise data and privacy controls
Coding & agent runtimeInteractive coding agents plus asynchronous cloud agents that can run code and tests and open pull requests
Context layerSource, architecture, product docs, tickets, APIs, runbooks, and internal knowledge reachable by agents through repo context, MCP, and other controlled connectors
Instructions & skillsVersion-controlled AGENTS.md, coding standards, architectural rules, reusable agent skills, and shared prompts and workflows
Execution environmentsSandboxed development environments; reproducible builds; scoped network and tool access; ephemeral credentials
Verification & evalsCI, automated tests, type and lint checks, security scanning, agent evals, acceptance criteria, and regression suites
Identity & permissionsLeast-privilege agent identities; explicit tool permissions; protected environments; human approval for consequential actions
Observability & governanceAgent and action logs, traceability to a human owner, usage and cost visibility, quality metrics, and auditability
Memory & feedback loopsDecisions, post-mortems, and review outcomes written back into the context layer, so agents and new hires inherit what the team learned

Standardize the primitives and guardrails. Let teams optimize the workflow.

3Autonomy levels

Autonomy is set per workflow, not per team or per tool. The same team may run its maintenance queue at A4 and its production deploys at A3.

  1. A1AssistAI advises; a human performs the work.
  2. A2PairHuman and AI work interactively, turn by turn.
  3. A3DelegateA human defines the outcome; the agent executes independently and returns a result for review.
  4. A4AutonomousThe agent identifies or receives eligible work, executes it, verifies it, and produces a PR or action. Humans govern exceptions and consequential gates.
  5. A5Pre-authorized actionThe agent may take bounded production actions within explicit policy. Use selectively.

Autonomy is earned through verification.
Bounded scope + tests and evals + observability + reversibility + low blast radius → greater autonomy

4GenAI across the SDLC

Humans keep intent, judgment, and accountability at every stage. Agents take the volume, and the autonomy they are given rises with how well the stage can be verified.

SDLC stageHumans ownRole of GenAI and agentsTypical autonomyPaved-road tools
Discover & understandCustomer empathy, problem selection, judgmentResearch; synthesize interviews and support; find patterns; interrogate analytics; generate hypothesesA1–A2Chat and LLM; transcript tools; analytics; search and knowledge connectors
Define & specifyProduct intent, outcomes, priorities, acceptance criteriaDraft and refine specs; expose ambiguity; enumerate edge cases; convert requirements into structured, agent-executable tasksA1–A2LLM; product docs; ticketing; internal context
Design & prototypeUX judgment, taste, user validationGenerate alternatives; prototype UI and flows; create assets; critique designs; build disposable prototypesA2–A3Figma AI; v0 and Lovable-style tools; coding agents
Architecture & planSystem boundaries, invariants, NFRs, irreversible decisionsInspect the codebase; research alternatives; identify dependencies; draft ADRs; decompose work for humans and agentsA2–A3Coding agents; repo and docs context; architecture skills; MCP
ImplementTechnical direction, hard judgment calls, ownership of correctnessGenerate implementation; refactor; debug; migrate; execute bounded issues independently; run parallel workstreamsA2–A4Cursor, Claude Code, Codex, Copilot; cloud coding agents; sandboxes
Test & verifyRisk model, critical behaviors, the acceptance barGenerate tests; find missing cases; run unit, integration, and E2E suites; fuzz and adversarial testing; diagnose failuresA3–A4CI; Playwright; pytest and Jest; coding agents; eval framework
ReviewArchitecture, maintainability, intent, consequential judgmentAutomatic first-pass review; correctness, security, and style checks; compare the change against the spec; suggest or apply fixesA3–A4AI code review; CodeQL and SAST; agent skills; repository instructions
Integrate & releaseRelease risk and authorizationResolve CI failures; write release notes; assess change risk; verify release criteria; prepare the rolloutA3–A4GitHub Actions and CI; deployment platform; policy gates
DeployHigh-impact production authorizationExecute the validated deployment workflow; monitor the rollout; halt or roll back automatically on predefined conditionsA3–A5CI/CD; feature flags; policy-as-code; deployment automation
Operate & respondIncident command, customer impact, consequential decisionsCorrelate logs, traces, and events; investigate hypotheses in parallel; identify suspect changes; propose remediationA2–A4Datadog and Sentry; OpenTelemetry; runbooks; production-read MCP and tools
Maintain & improvePrioritization and systemic improvementFind and fix flaky tests; dependency upgrades; vulnerability remediation; docs drift; repetitive refactors; tech-debt queuesA4 default where verifiableAutonomous and cloud agents; Renovate and Dependabot; scanners; CI

5The expected engineering workflow

The default used to be

  1. Design
  2. Code
  3. Review
  4. Ship

It is increasingly

  1. Frame
  2. Decompose
  3. Delegate
  4. Parallelize
  5. Verify
  6. Integrate
  7. Ship

A single engineer may simultaneously:

The scarce resource shifts from code production to intent, decomposition, context, judgment, and verification.

6Expectations for every product team