peterbamuhigire SaaS Agent Talent Strategy Skill
Overview
AI-feature SaaS hiring (handled by saas-ai-talent-strategy) covers ML Engineer / Applied-ML / AI PM / Prompt Engineer / MLOps / AI Safety Lead (optional). Agent talent strategy makes the AI Safety Lead mandatory (because of irreversibility risk) and adds agent-specific roles: Agent Architect (designs planner / worker / critic), Tool Engineer (builds + maintains the tool registry and integrations), Eval Engineer (builds + runs the eval-loop on production data), HITL Designer (designs the human-in-the-loop UX and policy), Forward Deployed Engineer (customer-specific agent shipping where vertical depth is the moat).
This skill installs the agent team composition discipline.
Use When
- A SaaS plan ships an agent or multi-agent product
- Section 09 is being built for an agent-product plan
- The plan claims agent autonomy in Class C / Class D actions (must show AI Safety Lead in seat)
- Investor / DFI diligence on team capacity is upcoming
- Talent-cost projections for Section 10 need agent-specific roles
- The plan must pass
meta-agent-bankability-and-investor-readiness
Do Not Use When
- AI is internal-efficiency only — use generic
09-management-team+saas-ai-talent-strategy - The plan is too early (pre-PMF) for full team specification — use directional roles and milestones
Required Inputs
- Agent archetype (single-agent / multi-agent / vertical / platform)
- Action class taxonomy (A/B/C/D)
- Customer count and trajectory
- ARR trajectory
- Geography (where you hire, what compensation market, talent pool reality)
- Build-vs-buy posture for orchestration framework, eval platform, observability
- Customer-deployment model (self-serve / forward-deployed / hybrid)
- Existing team and skills audit
Workflow
Map the mandatory agent roles per
references/saas-agent-talent-and-org-design-template.md:- Agent Architect — designs the multi-step decomposition; planner / worker / critic; tool registry strategy; autonomy ladder
- Tool Engineer — builds and maintains the tool registry; integrations; vendor abstraction; reliability engineering for tools
- Eval Engineer — builds and runs the eval-loop; offline + online; human-correction signal capture; regression detection
- AI Safety Lead — mandatory — irreversibility-class policy; red-team / drill cadence; audit-log; regulator engagement; incident response; kill-switch design
- HITL Designer — human-in-the-loop UX; escalation policy; HITL reviewer training; intervention-cost optimisation
- MLOps / Agent Infra — runtime; observability; tracing; deployment; sandbox / staging; cost engineering at infra layer
- Forward Deployed Engineer (FDE) — customer-specific agent shipping; vertical depth; integration work; required when vertical moat is part of thesis
- Agent Product Manager — agent roadmap; autonomy ladder; customer outcomes; pricing input
- Domain Expert as Trainer — vertical knowledge encoded into eval-set and prompt design (legal, medical, financial, agronomic, etc.)
Stage the hiring plan by ARR / autonomy milestone:
- Pre-seed / pre-PMF: founders cover most; fractional AI Safety Lead acceptable; one FDE
- Seed (USD 0-1M ARR): Agent Architect + Tool Engineer + Eval Engineer + AI Safety Lead (full-time mandatory if Class C/D actions live) + 1 FDE
- Series A (USD 1-5M ARR): above + HITL Designer + Agent PM + MLOps + 2-3 FDEs
- Series B (USD 5-20M ARR): above + Domain Expert team + multiple FDEs + Customer-AI-Ops team
- Growth (USD 20M+ ARR): above + regional safety leads + vertical PM team + AI policy / compliance team
Compensation benchmarks — set salary bands using
references/saas-agent-talent-and-org-design-template.md. African market reality: AI Safety Lead and Eval Engineer are scarce and command premium (often 1.5-2.5x equivalent ML Engineer comp); Agent Architect 1.3-1.8x; Tool Engineer 1.0-1.2x typical senior engineer; FDE 1.1-1.3x senior engineer.Build-vs-buy decisions:
- Orchestration framework: typically buy (LangChain / LangGraph / CrewAI / Semantic Kernel / AutoGen) and build proprietary layer on top
- Eval platform: buy + customise (LangSmith / LangFuse / Arize / Weights & Biases) for general; build proprietary eval sets on customer data
- Observability: typically buy (Helicone / LangSmith / LangFuse)
- Tool registry: build proprietary (this is moat territory)
- HITL workbench: buy or build; depends on vertical depth
- Sandbox / staging environment: build (specific to your tool set)
Retention plan — agent talent is scarce; retention is more important than initial hire:
- Comp at or above market band; review every 6 months
- Equity refresh annually
- AI-specific learning budget (conferences, courses, compute for personal projects)
- Public visibility (talks, papers, open-source contributions) where compatible with IP
- Mission alignment for AI Safety roles
- Avoid "we'll figure it out" leadership; safety / eval people leave fast when leadership is glib about risk
African talent map — where to recruit and the constraints:
- Carnegie Mellon Africa (Kigali) — AI / ML graduates; small cohort
- African Leadership University AI track (Mauritius / Rwanda) — graduates
- Andela AI pool — pan-African; mid-senior
- Deep Learning Indaba alumni network — pan-African
- AIMS network (SA / Cameroon / Senegal / Tanzania / Ghana / Rwanda) — strong fundamentals
- Lelapa AI (SA) / Masakhane (pan-Africa) / Awarri (Nigeria) / EqualyzAI — African AI research community
- InstaDeep alumni (Tunisia/UK; BioNTech acquisition precedent)
- University of Lagos / UNILAG, Covenant University, Ashesi, UCT, Stellenbosch, Strathmore, Makerere, KIST, Addis Ababa University, AAU — CS / ML talent pipelines
- Remote-first or remote-fractional for AI Safety Lead — given scarcity, fractional / advisory / cross-border arrangement often realistic
- South Africa is the densest agent-engineering market in SSA — consider hub model with SA core + distributed engineering
Diversity and inclusion — gender ratio in AI talent in Africa is improving but still skewed; deliberate sourcing matters. Local-language fluency is often a hiring criterion for vertical agents.
Outsource-build-buy posture — fractional AI Safety Lead from external advisors (e.g. Africa AI Safety Consortium, external lawyers with AI specialism) is acceptable at seed; must convert to full-time by A.
Wire to Section 10 — compensation projections feed Section 10 OpEx. Agent team payroll allocates partly to COGS (
saas-agent-unit-economics-and-cogsoverhead allocation, default 60% to COGS).Wire to bankability —
meta-agent-bankability-and-investor-readinesschecks "AI Safety Lead in seat" as a scorecard item.
Quality Bar
- All mandatory roles listed with seniority and full-time / fractional status
- AI Safety Lead role filled (or fractional path explicit if pre-seed)
- Hiring plan staged by ARR / autonomy milestone
- Compensation bands set against market benchmarks
- Build-vs-buy decisions explicit per category
- Retention plan with comp, equity, learning, mission components
- African talent map referenced; sourcing channels named
- Diversity and local-language hiring criteria stated where relevant
- Outsource-build-buy posture for AI Safety + Eval explicit
- Cross-references to Section 10 (compensation costs) and bankability
Anti-Patterns
- AI Safety Lead role optional or "we'll add later" when Class C/D actions are live — bankability impossible
- Single ML Engineer expected to cover Agent Architect + Tool + Eval + Safety
- No HITL Designer despite an active HITL workflow — UX and policy decay
- No Forward Deployed Engineer when vertical depth is the moat — moat does not materialise
- Compensation set at generic-engineer benchmark for scarce AI talent — recruitment fails
- "We'll outsource AI Safety" without specifying provider and cadence
- "We don't need an Eval Engineer because the model is good" — production agents need continuous eval
- All-remote AI Safety with no in-region presence in regulated markets — regulator engagement weakens
- Treating Domain Expert as a side-quest rather than a productised role for vertical agents
Outputs
- Org chart for current stage + next stage
- Mandatory roles inventory with seniority + full-time / fractional / advisor
- Hiring plan by ARR / autonomy milestone
- Compensation bands by role and seniority
- Build-vs-buy posture per category
- Retention plan with concrete levers
- African talent sourcing map (where applicable)
- Outsource-build-buy posture for AI Safety + Eval
- Cross-reference to Section 10 costs and to bankability scorecard
Living-Plan Cadence Defaults
| Element | Cadence | Owner | Variance threshold |
|---|---|---|---|
| AI Safety Lead retention signal | monthly | Head of People + CEO | flight risk signal |
| Agent team attrition | quarterly | Head of People | >15% |
| Hiring plan vs ARR | quarterly | CEO + Head of People | <80% on plan |
| Compensation market scan | semi-annual | Head of People | band drift >10% |
| Build-vs-buy posture review | annual | CTO | structural shift in vendor landscape |
| Domain expert recruitment | quarterly | Head of Product + Head of AI | vertical move |
| Diversity metrics | quarterly | Head of People | structural skew |
References
references/saas-agent-talent-and-org-design-template.md— role specs, comp bands, hiring stagesskills/09-management-team/saas-ai-talent-strategy/SKILL.md— AI talent parentskills/09-management-team/SKILL.md— generic management team sectionskills/saas-sales-org-design-and-capacity-planning/SKILL.md— sales org pairingbook-extractions/cotton-run-a-saas-business-extraction.md— hiring disciplinebook-extractions/agent-products-business-plan-audit-2026.md— agent auditcountry-context/africa-regional/africa-ai-context-extension.md— African AI talent contextcountry-context/africa-regional/africa-agent-context-extension.md— African agent context
Africa / Uganda Application Notes
- AI Safety Lead scarcity in Africa is severe; consider fractional / remote-first / advisory at seed; full-time by A is mandatory for any plan with Class C/D agent actions and regulated-sector customers.
- Eval Engineer scarcity — second-scarcest role. Consider rotation from senior backend / SRE talent with focused training; ALU AI track + Deep Learning Indaba alumni are sourcing channels.
- Agent Architect can often be sourced from senior ML / backend engineers with 6-12 months of agent-product exposure; Andela AI pool is a sourcing channel.
- Tool Engineer is the most-available role; senior backend engineers with API integration experience adapt readily; UG / KE / NG have strong pools.
- HITL Designer — combine product / UX with operations knowledge; African operations-savvy designers can be sourced from BPO / call-centre operations backgrounds (CCI Kenya, Genesys partners, Webhelp / Concentrix Africa).
- Forward Deployed Engineer — vertical-specific; pair domain knowledge with engineering; CMU-Africa graduates are a strong source.
- Compensation realities — AI Safety Lead in Nairobi / Lagos / Cape Town / Kigali commands USD 4,500-9,500 / month loaded; Eval Engineer USD 3,500-7,000; Agent Architect USD 4,000-8,500; FDE USD 3,500-6,500; Tool Engineer USD 2,500-5,000. Remote-USA roles compete at 1.5-3x these levels — retention strategy must address.
- Local-language fluency as a hiring criterion is genuine for vertical agents (Swahili / Hausa / Yoruba / Amharic / Luganda / Lingala / Zulu / Xhosa / Wolof / Tigrinya).
- Sovereign-AI procurement may require local-citizen / local-resident headcount minima; plan for this when targeting public-sector customers.
- Outsource-build-buy posture — initial AI Safety advisory can be from external consultancies (Africa AI Safety Consortium, Lelapa AI partners, EqualyzAI partners) with quarterly engagement; convert to full-time by A.
- Talent-flight risk — pan-African AI talent is heavily recruited by US / EU / UAE remote roles; retention plan must combine comp + mission + equity + learning + visibility.
July 2026 Portable Contract
Required Inputs
| Input artefact | Source/provider | Required | Behaviour when absent |
|---|---|---|---|
| Role requirements, verified biographies, workload, organisation design, compensation assumptions, and hiring evidence for saas agent talent strategy | Founders, HR records, approved payroll model, and reference evidence | Yes | If absent, a biography, role need, workload, or pay assumption is unavailable, mark it unverified and keep the role or hire conditional. |
| Finalised business brief, target reader, country, and stage | Client intake and engagement owner | Yes | Stop section decisions and route the missing context to client intake. |
| Reconciled upstream assumptions that this section consumes | Named pipeline owners | Conditional | Record the dependency, affected claim, owner, and recovery step; do not substitute an invented value. |
Outputs
| Artefact | Consumer | Observable acceptance condition |
|---|---|---|
| Management-team section and staged organisation/hiring plan | Plan author and target decision-maker | The artefact answers the section decision and traces each material conclusion to the supplied evidence. |
| saas agent talent strategy exception and handoff note | Downstream section owners | Every blocked or conditional item names its consequence, owner, evidence request, and restart condition. |
| saas agent talent strategy release record | Reviewer or plan assembler | Records the checks completed, failures, unassessed items, professional review required, and release state. |
Evidence Produced
| Evidence | Format | Acceptance condition |
|---|---|---|
| Credential trace, accountability map, hiring trigger, and payroll reconciliation | Source-linked table, calculation, or annotated prose | The evidence is reproducible from named inputs and distinguishes verified fact, management assumption, and inference. |
| saas agent talent strategy decision record | Decision note | States the selected action, rejected credible alternative, countercase, rationale, and risk accepted or avoided. |
| saas agent talent strategy review trace | Gate entry | Identifies the date, input versions, reviewer role, failed checks, recovery owner, and any check that remains not assessed. |
Capability and Permission Boundaries
For saas agent talent strategy, the controlling focus is agent product, tool, evaluation, safety, operations, and customer-success accountabilities. This skill may analyse roles and draft people plans using authorised records; it may not verify credentials by assertion, disclose sensitive HR data, hire, discipline, or set compensation without authority. Its normal mode is read-only analysis and drafting. Any mutation, external communication, spending, certification, or professional conclusion outside that boundary requires explicit authority and must remain traceable to the approving role.
Degraded Mode
For saas agent talent strategy, loss of evidence about agent product, tool, evaluation, safety, operations, and customer-success accountabilities activates degraded mode. If the controlling saas agent talent strategy evidence is unavailable, the same boundary applies. When a biography, role need, workload, or pay assumption is unavailable, mark it unverified and keep the role or hire conditional. Return the verified subset, label the affected decision qualified or not assessed, explain the downstream consequence, and state the smallest evidence request or authorised action that permits recovery. Do not convert the missing check into a pass.
Decision Rules
| Choice or condition | Action | Failure or risk avoided |
|---|---|---|
| For saas agent talent strategy, a named leader lacks evidenced capacity for a critical accountability | state the gap, assign interim ownership, and define the hire, adviser, or development trigger | Founder-centric organisation claims hide execution and governance gaps |
| For saas agent talent strategy, A current legal, regulatory, tax, accounting, market, or platform claim controls the saas agent talent strategy decision | Verify the controlling source, effective date, jurisdiction, and reviewer status before release | Stale external facts become permanent plan assumptions |
| For saas agent talent strategy, The evidence reconciles with neighbouring sections and the countercase does not overturn the choice | Complete management-team section and staged organisation/hiring plan, attach the evidence and release record, and hand off named dependencies | Premature release and repeated downstream rework |
Workflow
- Define the exact saas agent talent strategy decision, intended reader, jurisdiction, business stage, and permission boundary.
- Collect role requirements, verified biographies, workload, organisation design, compensation assumptions, and hiring evidence and map each material conclusion to its source; stop the affected conclusion when an input could change it.
- Apply the specialist methods and directly linked references already contained in this skill, retaining its domain thresholds, calculations, and Uganda or East Africa context where applicable.
- Compare the credible alternatives, test the countercase and failure path, and apply the decision table rather than selecting a template default.
- Produce management-team section and staged organisation/hiring plan with the evidence, exception, and handoff records; reconcile every shared assumption with its owning section.
- Run the section quality checks, applicable finance or professional review, and anti-slop gate. If a gate fails, correct the evidence or decision and return to the responsible step.
Quality Standards
- Management-team section and staged organisation/hiring plan must answer a real decision for the named bank, investor, DFI, grant, board, or strategic-partner reader.
- Credential trace, accountability map, hiring trigger, and payroll reconciliation must be source-linked, dated where facts can change, and sufficient for another reviewer to reproduce the conclusion.
- The section exposes its countercase, stop condition, recovery action, and effect on neighbouring sections.
- No unavailable source, calculation, tool, or professional review is reported as passed; finance and statutory judgements follow the governing doctrine.
- Language remains specific to saas agent talent strategy, uses British English naturally, and passes the repository anti-slop gate without promotional filler.
Anti-Patterns
- In saas agent talent strategy, treating an unavailable role requirements, verified biographies, workload, organisation design, compensation assumptions, and hiring evidence as confirmed. Correction: qualify the affected conclusion and issue the named evidence request.
- Producing management-team section and staged organisation/hiring plan that restates the brief but makes no choice. Correction: record the choice, rejected alternative, rationale, countercase, and implication.
- Ignoring a conflicting upstream assumption. Correction: return it to its owning section and resume only from a reconciled version.
- Reporting an unavailable check as passed. Correction: mark it not assessed and narrow the release state.
- Claiming compliance, assurance, bankability, or investor readiness from narrative quality. Correction: run the applicable gate and retain its evidence.
- Copying the worked example into a client plan. Correction: use the method only and replace every fact with verified engagement evidence.
Worked Example
The roadmap needs tool integration and evaluation work but not frontier-model research. Hire or contract for tool engineering and eval operations rather than copying a generic ML laboratory structure.
References
- Use the verified project evidence register and the owning upstream pipeline section for saas agent talent strategy; no local deep-dive reference is declared.
- For saas agent talent strategy claims involving money, tax, grants, reserves, revenue, cost, valuation, or financial statements, apply the Chwezi finance doctrine and record the required professional-review state; illustrative figures never become client facts.