Reference Guide · Toolkit
AI Role Design: How to Scope, Staff, and Sequence the AI Function
Built for CAIOs, CIOs, and technology leaders designing the AI function as priorities, structure, and scope are still forming.
Built for CAIOs, CIOs, and technology leaders designing the AI function as priorities, structure, and scope are still forming.
Companion read. The Workforce 2030: Four Forces Reshaping Enterprise Talent Strategy research brief is the thesis behind this toolkit. Both sit in the Intelligence Center.
Want the working toolkit as a Google Sheet? Ten tabs, formula-driven scoring, dropdowns for your inputs, priority ranking that updates live. Click below to get your own editable copy in your Drive.
For CAIOs, CIOs, and technology leaders carrying AI adoption mandates
The problem: AI roles are hired against undefined work. Boards ask for "an AI plan" and hand it to the CIO, CAIO, or CHRO. Requisitions open. 44 days later the seat is still empty. 60% of candidates abandon the process. Most hires that land don't stick.
The cause: role design is downstream of work specification. Without specifying what AI work the organization actually needs done, at what scale, under what governance, every job description is generic and every interview is unfocused.
This toolkit: twelve AI role archetypes across four clusters (strategic, bridge, technical, operational) with comp bands and seniority. A ten-skill matrix mapping every role. A decision tree for which role comes first. A hiring sequence keyed to organizational maturity. Job description templates. Three composite cases.
Use it to structure the conversation with your board, your search firm, and your finance team — before the first requisition is opened.
Working toolkit: this page is the reference. The working instruments — Input Specs worksheet, Priority Scorecard with live formulas, Hiring Sequence planner, JD Builder, Reporting Lines matrix — live in the companion Google Sheet. Copy the Google Sheet to your Drive →
Title-first, scope-second — the failure pattern.
Most AI hiring starts with a title rather than a scope. A board asks for a Chief AI Officer because peer companies have one. A CIO opens a requisition for an AI Product Manager because a vendor suggested the pattern. A head of HR writes a job description by pulling language from LinkedIn listings, few of which reflect the work the candidate will actually do.
The downstream effect is visible in outcomes. 44 days to fill a typical AI role. 60% candidate abandonment during the hiring process. 75% of organizations reporting persistent skill gaps despite active hiring. These are not labor market artifacts. They are symptoms of unclear scope: candidates walk when they cannot tell what they are being hired to do, and organizations cannot tell them because the work has not been defined.
The fix is structural. Before a requisition opens, the role needs an archetype, a skill profile, a seniority band, and a position in a sequence. This toolkit supplies all four.
The upstream work almost no one does.
Before a job description is written, six dimensions of the organization must be specified. Without them, every downstream choice is generic — the archetype, the seniority band, the skill weighting, the hiring sequence, the interview prompts. With them, role design is mechanical.
What is the state of your AI platform, data foundation, and technical stack? Greenfield is different from fragmented legacy, which is different from unified platform with partial data maturity. The architecture you inherit shapes whether MLOps comes before ML engineers, and whether a data product owner is your first hire or your third.
Where is workforce AI fluency today? Check benchmarks by function, training spend as a share of AI budget, usage patterns, and resistance signals. Adoption gaps demand an Enablement Lead before additional engineers. Without absorption capacity, deployed systems sit unused.
Have you mapped where AI actually belongs in daily operations? A formal workflow catalog, AI-to-human handoff design, and friction audits indicate integration readiness. Without workflow specification, AI output is produced and then ignored. Bridge-cluster roles — Business Translator, Product Manager — close this gap.
Regulatory exposure (EU AI Act, GDPR, HIPAA, SOX), policy architecture, named accountability, and incident history. Regulated industries sequence governance before engineering. Without design-time governance, approval queues become six-month bottlenecks and AI systems get retrofitted for compliance they should have been built with.
How many AI initiatives are active (production, pilot, proposed)? What are the elimination criteria? What is the build-versus-buy ratio, and is it explicit or de facto? Sprawling portfolios past fifteen initiatives need a CAIO to consolidate. Build-heavy portfolios need engineers and research scientists; buy-heavy needs product managers and enablement.
Where does AI accountability live? CEO and board, CIO or CTO, business-unit leaders, market pressure, or absent. Named P&L ownership, board reporting cadence, and executive sponsorship all matter. Absent mandate means hiring will not stick. The mandate source determines the reporting line logic for the first AI hire.
Until these six dimensions are specified for your organization, every AI job description is generic, every interview is unfocused, every candidate is a gamble. Most organizations skip the work because nobody owns it — the CEO doesn't define it, the board doesn't define it, the search firm can't, and the CHRO often lacks the framework.
The skill cartography under the taxonomy.
The matrix maps every role archetype against the ten core competencies that determine fit. Cells indicate required depth: foundational awareness, proficient application, or expert-level command. The four clusters — strategic, bridge, technical, operational — appear as row groupings.
| Role Archetype | AI / ML Depth |
Business Strategy |
Product Sense |
Data Engineering |
Ethics & Governance |
Change Management |
Exec Comms |
Systems Thinking |
AI Tooling |
Financial Acumen |
|---|---|---|---|---|---|---|---|---|---|---|
| Strategic Cluster — governs direction and authority | ||||||||||
| Chief AI Officer | F | E | P | F | E | P | E | E | F | E |
| Head of AI Strategy | P | E | P | F | P | P | E | E | P | P |
| AI Ethics & Governance Lead | P | P | F | F | E | P | E | E | F | F |
| Bridge Cluster — translates strategy into execution | ||||||||||
| AI Product Manager | P | P | E | F | P | P | P | P | P | F |
| AI Program Delivery Manager | F | P | P | F | P | E | P | P | P | F |
| AI Business Translator | P | P | P | F | F | P | P | E | P | F |
| Technical Cluster — builds and operates the systems | ||||||||||
| AI / ML Engineer | E | F | F | P | F | F | F | P | E | F |
| AI Research Scientist | E | F | F | P | P | F | F | P | E | F |
| MLOps / AI Platform Engineer | P | F | F | E | F | F | F | P | E | F |
| Data Product Owner | P | P | E | E | P | F | P | P | P | F |
| Operational Cluster — drives adoption and day-to-day use | ||||||||||
| AI Enablement Lead | F | P | P | F | P | E | P | P | P | F |
| Prompt / AI Operations Specialist | P | F | P | F | F | F | F | P | E | F |
Reading the matrix: row patterns reveal role identity (a role defined almost entirely by a single expert skill is likely too narrow; a role with five expert skills is likely too broad). Column patterns reveal which skills are in highest aggregate demand across your future hiring plan.
Four clusters, twelve archetypes, one map.
Each archetype specifies its core responsibilities, required skills, seniority band, reporting line, and representative compensation range. Compensation ranges reflect U.S. metropolitan markets Q2 2026 and are directional — adjust for your geography and industry.
These roles set portfolio priorities, define the organizational operating model around AI, and hold external-facing accountability to boards, regulators, and the capital markets. They are board-visible and typically report to the CEO.
The bridge cluster connects strategic intent to technical delivery. These roles own the translation layer between business problems and AI solutions. When organizations cut this cluster to save cost, technical teams build the wrong things and business leaders interpret AI as underperforming.
The technical cluster builds, deploys, and maintains AI systems. The four roles within it are not interchangeable, and treating them as such—a common error in small AI teams—produces systems that work in a notebook and fail in production, or data products that are technically correct and organizationally unusable.
The operational cluster is where most enterprise AI investments fail. Systems ship, then sit unused or misused. These roles close the absorption gap between capability deployed and capability adopted.
Five questions, one branch per answer.
Answer the questions in sequence. Each branch lands on a specific archetype and a reason. If multiple branches land on the same archetype, your need is well-defined. If the tree produces no answer, the underlying work has not been scoped — return to problem definition before opening a requisition.
The sequence below assumes an organization at mid-market to enterprise scale without an existing AI function. Treat it as a default calibrated to a common starting condition—adapt to your specifics. The sequence derives from the Starkey Model™ applied to role prioritization: score each potential role against organizational value potential and implementation feasibility, then sequence by compounding effect.
Common error: organizations skip the bridge cluster and hire engineers directly after the owner. The engineers build against their own interpretation of the problem, and six months later the organization reports that AI did not work. In virtually every post-mortem, the missing role is bridge, not technical.
The template below is structural, not substantive. Use it as the outline; fill the placeholders with specifics drawn from the archetype and skill matrix above. Generic JDs are a primary driver of the 60% candidate abandonment rate.
Title: [archetype name] — [business unit or function]
Reporting line: reports to [title]. Partners with [two to three adjacent functions].
The work: three to five sentences describing what the person will actually spend time on in their first 90 days. Specific verbs (define, deploy, coordinate, measure) over generic ones (own, drive, lead). Reference the archetype summary and the three highest-weight skills from the matrix.
Outcomes expected in 12 months: three measurable outcomes, each with a quantitative target or observable threshold. Not activities. Outcomes.
Required qualifications:
Preferred qualifications: two to three items. If the list exceeds five, the JD is unfocused.
Compensation range: disclose the range. Per the archetype, calibrated to your geography. Non-disclosure predicts abandonment.
The cases below are composites drawn from recurring patterns in RBD. client engagements and published enterprise AI research. They are not specific organizations. Each illustrates how the archetypes, the matrix, and the decision tree produce a specific hiring sequence for a specific context.
The organization has no AI initiatives in production. The CEO has set a two-year target for AI-driven margin improvement. The CFO is skeptical. IT is a three-person function without data engineering capacity. Operations leaders have surfaced candidate use cases but none have been evaluated against a portfolio framework.
The first hire is a Head of AI Strategy, not a CAIO. The organization is not yet at the scale where a C-suite AI executive makes sense, and the immediate need is portfolio definition rather than board-level representation. The Head of AI Strategy spends the first quarter applying the Starkey Model to the surfaced use cases and selecting three for development. The second hire, late Q1, is an AI Business Translator embedded in operations, the function carrying the most use-case pressure. Q2 hires are an ML Engineer contracted rather than employed (org maturity does not yet justify headcount) and an AI Enablement Lead at director level, reporting to HR with a dotted line to AI Strategy.
The organization has roughly 20 active AI initiatives distributed across business units. None have generated material EBIT. Governance is ad hoc. The CIO is the de facto AI owner but does not have P&L authority for AI outcomes. The board has asked for a plan. The regulator has signaled expectations without issuing formal guidance.
The first hire is a Chief AI Officer with direct P&L and board-facing accountability. The CAIO’s first action is to apply the Starkey Model to the existing 20 initiatives and eliminate 14 of them, redirecting resources to the six highest-scoring. The second hire is an AI Ethics & Governance Lead at VP level, reporting jointly to the CAIO and General Counsel, with a Month-0 governance protocol for the surviving six initiatives. The third hire is a Data Product Owner, because the real bottleneck across the six initiatives is data readiness rather than model capability. Technical hiring follows in Q3.
The company has two parallel AI programs. One is product-facing—AI features embedded in the core SaaS platform, owned by the CTO and VP Product. The other is internal—AI applied to sales enablement, customer success workflows, engineering productivity, owned loosely across functions with no clear lead. The product AI ships regularly. The internal AI is a mess: twelve experiments, three production deployments, no shared infrastructure. The board is asking why the internal side lags the external side.
The first hire is a Head of AI Strategy for the internal program, reporting to the COO with a dotted line to the CTO. The Head of AI Strategy applies the Starkey Model to the twelve internal experiments and consolidates to four. The second hire is an MLOps / AI Platform Engineer—the real bottleneck across both programs is that internal AI has no shared platform, and the engineers reinvent deployment for every use case. The third hire is a Data Product Owner, because both the internal and product AI programs are gated by data readiness rather than model capability.
This toolkit is built to be used inside a single leadership conversation. Bring it to the next board meeting, search firm briefing, or executive team session where AI hiring is on the agenda.
Companion read. The Workforce 2030: Four Forces Reshaping Enterprise Talent Strategy research brief is the thesis behind this toolkit. Both sit in the Intelligence Center.
Want the working toolkit as a Google Sheet? Ten tabs, formula-driven scoring, dropdowns for your inputs, priority ranking that updates live. Click below to get your own editable copy in your Drive.