Engineering Manager Vertical Ai Products

Engineering Manager, Vertical AI Products

When to Use

• Build or scale vertical AI product engineering (domain squads, not central platform only)
• Prioritize vertical backlog with PM, GTM, and domain SMEs
• Staff and develop AI engineers, fullstack, and tech leads on vertical lines
• Run launch governance for customer-facing AI (eval, risk tier, kill switch)
• Resolve platform vs vertical build conflicts and shared component roadmaps
• Set team KPIs (ship cadence, eval regression, incidents, cost per vertical)
• Escalate vertical-specific compliance or data-boundary issues

When NOT to Use

• AI solution architecture and multi-tenant ADRs → applied-ai-architect-commercial-enterprise
• Implement RAG, agents, eval code → ai-engineer
• AI production ops, vendor SLOs, org-wide model release process → ai-lead-ops
• AI policy registers and regulatory mapping → ai-risk-governance
• Analytics engineering and dbt marts → analytics-data-engineering-manager-product
• Company-wide non-AI programs → technical-program-manager
• UX research and interaction design → product-designer

Related skills

Core Workflows

1. Vertical squad org design

Hub platform vs vertical pods; domain SME interfaces; ratios.

See `references/vertical_team_org.md`.

2. Roadmap and vertical bets

Platform leverage vs bespoke vertical logic; capacity and bet sizing.

See `references/vertical_roadmap_prioritization.md`.

3. AI feature launch governance

Eval gates, risk tier, rollback, hypercare — coordinate with ops and risk.

See `references/ai_feature_launch_governance.md`.

4. Stakeholder partnerships

PM, sales, solutions, legal, horizontal AI platform.

See `references/stakeholder_vertical_partnerships.md`.

5. Hiring and development

IC/lead levels for vertical AI product engineering.

See `references/hiring_development_vertical.md`.

6. Team metrics and accountability

Delivery, quality, safety, unit economics by vertical.

See `references/team_metrics_vertical_ai.md`.

Output standards

• Roadmap items: vertical outcome, AI capability, platform dependency, owner
• No GA without signed eval/risk checklist for customer-facing AI
• Escalations include trade-offs (scope, date, platform build vs fork)
• Architecture changes route through applied-ai-architect-commercial-enterprise

When to load references

• Org → references/vertical_team_org.md
• Roadmap → references/vertical_roadmap_prioritization.md
• Launch → references/ai_feature_launch_governance.md
• Stakeholders → references/stakeholder_vertical_partnerships.md
• People → references/hiring_development_vertical.md
• KPIs → references/team_metrics_vertical_ai.md

AI Feature Launch Governance

Pre-GA checklist (customer-facing)

Coordinate tiers with applied-ai-architect-commercial-enterprise ADRs.

Launch phases

Rollback triggers

• Safety incident or policy violation
• Eval regression beyond agreed threshold
• Cost overrun >2× forecast without approval
• Critical wrong-answer pattern in production logs

Post-launch

• Retro: eval gaps, platform asks, SME quality
• Feed golden set updates into eval harness
• Ticket horizontal platform improvements with priority label

Internal / low-tier features

Lighter checklist — still require eval smoke and owner sign-off.

Hiring and Development (Vertical AI)

What to hire for

Rarely hire research-only without shipping bar for product vertical teams.

Interview signals

Include ai-engineer or tech lead in loop for AI depth.

Level guide (summary)

Onboarding

Growth

• Deepen domain or platform — both valid paths
• Rotations to horizontal platform for senior ICs
• Managers: exec communication, capacity planning, risk judgment

Retention

• Chronic eval neglect
• Platform team blocking without SLA
• Unclear vertical strategy / pivot whiplash

Stakeholder Partnerships (Vertical AI)

Product management

Go-to-market and sales

• Before GA: Limitations doc (what AI does not do)
• Deal desk: Flag custom vertical requests — cap % engineering capacity
• Solutions: Separate product roadmap from services SOW

Domain subject-matter experts

Contract SME time like a dependency — block launches if absent.

Horizontal AI platform

Risk, legal, security

• Early review for customer data in prompts, cross-tenant, autonomous actions
• ai-risk-governance for tier; counsel for customer contract language
• cybersecurity for traditional appsec; not a substitute for AI eval

Status communication

Weekly vertical sync: shipped / in flight / blocked / decisions

Exec summary: outcome metrics, not model names — avoid overpromising capability

Anti-patterns

• Sales-driven GA without eval sign-off
• Vertical fork of platform without architect review
• Promising domain coverage beyond eval set proof

Team Metrics (Vertical AI)

Delivery

Quality and safety

Economics

Align targets with finance and ai-lead-ops.

Health

Accountability

EM owns predictable vertical delivery and launch discipline, not writing all prompts.

Quarterly review with PM: outcomes, incidents, cost, platform asks.

Vertical Roadmap and Prioritization

Backlog item template

Prioritization axes

Balance vertical wins with horizontal platform debt — negotiate 70/30 or explicit platform OKRs.

Build vs configure

Escalate extend-vs-fork to applied-ai-architect-commercial-enterprise.

Capacity

• Reserve 25% for eval debt, incidents, platform upgrades
• Limit concurrent Tier-1 AI launches per squad
• Front-load data + SME time before engineering sprints

Escalation

Escalate when: deal-driven date conflicts with eval gate; vertical needs exception to platform security; headcount mismatch to committed vertical OKRs.

Vertical Team Org Design

Vertical vs horizontal

EM role: vertical pod delivery while enforcing platform contracts.

Squad composition (typical)

Operating models

Cap bespoke fork of platform — track % of vertical code that bypasses shared SDK.

Interfaces

Anti-patterns

• Vertical team owns production model hosting without platform SRE
• No domain SME → bad eval sets and customer trust loss
• EM committing to model/vendor without architect + procurement path