Use this when Kate is scoping or running a fractional Chief AI Officer engagement: a senior, part-time mandate covering AI strategy, governance, and portfolio prioritization, not hands-on building. This skill designs the mandate, the first 90 days, the governance scaffold, the opportunity portfolio, the operating cadence, pricing, success metrics, and the capability handoff. If the client wants generic fractional product leadership rather than a CAIO-specific mandate, use /fractional-engagement-design instead.
Related skills: Generic structure comes from
/fractional-engagement-designand the run-the-engagement rhythm from/fractional-operating-system. The portfolio and 90-day plan feed/ai-transformation-roadmap. Package the result with/engagement-proposal.
The hard part most teams miss
A fractional CAIO sells judgment, not hours. The value is the decisions made and the messes avoided, which is exactly the part that is hard to see on a timesheet.
- A CAIO sells judgment and governance, not throughput. The mandate is deciding what to build, what to stop, and how to govern it safely. The moment a client expects the fractional CAIO to ship features, the engagement has been miscast. Name this in the first conversation. If they want hands-on build, that is a delivery engagement with a different shape, different rate, and different person doing the typing.
- The first win must be visible and fast or the mandate erodes. A CAIO arrives without a team that already trusts them and often without a clear reporting line. Authority is granted, not assumed. If the first 60 to 90 days produce only strategy decks, the org concludes the role is overhead and quietly routes around it. Land one concrete, visible win the business can feel, then use that credibility to do the slower governance work.
- The engagement must build the org's own capability, with a planned handoff. A fractional CAIO that becomes permanent is not a success, it is a dependency, even though it keeps paying. The goal is an org that can prioritize, govern, and run its own AI portfolio without the fractional. Design the handoff on day one: name who inherits the mandate, build the scaffold they will operate, and let the retainer decline as capability transfers.
Process
Step 1: Gather inputs
Ask the client:
- Why a CAIO, and why fractional? (The trigger: board pressure, a stalled initiative, a near-miss on risk, competitors moving. This tells you what "working" means to them.)
- Who does this role report to, and who grants its authority? (CEO, board, a sponsoring exec. A CAIO with no clear line has no mandate to spend.)
- What is the current state of AI in the org? (Shadow usage, pilots in flight, tools already bought, prior failed attempts, existing data and security posture.)
- What does the sponsor expect in the first 90 days? (Their stated near-term win. If they cannot name one, that is the first thing to surface.)
- What is the risk and regulatory context? (Regulated industry, customer data, contractual AI restrictions, board-level risk appetite.)
- What internal capability exists to inherit this mandate? (A data lead, a product head, an eventual full-time hire. This sets the handoff target.)
- What is the budget envelope and term? (Monthly retainer range and how many months the sponsor is prepared to commit.)
Step 2: Define the mandate
Write the mandate as three to five sentences the sponsor can repeat verbatim. It must state:
- What the CAIO owns: AI strategy, governance, risk posture, and portfolio prioritization across the org.
- What the CAIO does not own: hands-on build, line management of delivery teams, or being the single approver on every model call. Draw this line explicitly. The sharpest version names the failure mode: "This role decides what we build and how we govern it, not the person who writes the code."
- Decision rights: what the CAIO can approve, what they recommend, and where the sponsor or board holds the final call.
If the client cannot agree to a mandate that centers judgment and governance over throughput, stop and reclassify the engagement. Say so plainly.
Step 3: Design the first 90 days
The first 90 days do three things in order: assess, win, scaffold. Do not let assessment run the full quarter.
| Phase | Window | Focus | Output |
|---|---|---|---|
| Assess | Days 1-30 | Map current AI state, shadow usage, tools, data, risk exposure, and stakeholder appetite | Current-state assessment and a ranked opportunity longlist |
| Land a visible win | Days 20-60 | Pick one high-confidence, low-friction opportunity the business can feel; drive it to a result | One shipped or de-risked outcome the sponsor can point to |
| Stand up governance | Days 45-90 | Build the lightweight governance scaffold: an AI use policy, a risk register, a working group, and a decision path | A governance scaffold the org operates without the CAIO present |
The visible win and the governance scaffold overlap on purpose. The win buys the credibility that makes the governance work land instead of reading as bureaucracy.
Step 4: Stand up governance and risk
Build the minimum governance that makes AI use safe and auditable, not a compliance theater that no one follows. The scaffold has four parts:
- An AI use policy: what is allowed, what is gated, what is forbidden, and where the line sits for customer data and sensitive decisions.
- A risk register: known and emerging AI risks, owners, and mitigations, reviewed on a standing cadence. (Conceptually this is the artifact
/ai-risk-registerproduces.) - A governance framework: who reviews AI initiatives, against what criteria, and how a decision gets made and recorded. (This is the structure
/ai-governance-frameworkdefines.) - A working group: the cross-functional group (product, data, legal, security) that operates the framework so governance survives the CAIO's exit.
Gate the irreversible and the regulated. Keep the reversible and low-stakes fast. Governance that slows every experiment to a crawl gets routed around, which is worse than no governance.
Step 5: Prioritize the AI opportunity portfolio
Score every candidate AI initiative on value and feasibility, then sequence by both. This is the core of the mandate: deciding what to build and, just as important, what to stop.
For each opportunity, capture:
- Value: the business outcome and a rough size (revenue, cost, risk reduction, time saved).
- Feasibility: data readiness, technical lift, organizational change required, and risk profile.
- Verdict: do now, queue, watch, or stop.
| Quadrant | Value x Feasibility | Action |
|---|---|---|
| Quick wins | High value, high feasibility | Do now. These fund credibility, including the first 90-day win. |
| Big bets | High value, low feasibility | Sequence deliberately; build the data and capability prerequisites first. |
| Fill-ins | Low value, high feasibility | Delegate or batch; do not let them crowd the calendar. |
| Money pits | Low value, low feasibility | Stop. Naming what to kill is a CAIO's highest-leverage move. |
Step 6: Set the operating cadence
Define the rhythm that keeps the mandate visible and the org moving.
| Activity | Frequency | Duration | Purpose |
|---|---|---|---|
| Exec / sponsor sync | {{weekly/biweekly}} | 30-45 min | Decisions, escalations, portfolio shifts |
| AI working group | {{biweekly/monthly}} | 60 min | Govern initiatives, review the risk register, unblock |
| Portfolio review | Monthly | 60 min | Re-score opportunities, confirm what to start and stop |
| Board / leadership readout | Quarterly | 60-90 min | Progress, risk posture, capability transfer status |
| Async update | Weekly | Written | Decisions made, risks surfaced, asks of the sponsor |
The async weekly update is non-negotiable for a fractional. It is how a part-time leader stays present in an org they are not in every day, and how the mandate stays visible between syncs.
Step 7: Output the engagement plan
# Fractional CAIO Engagement: {{client name}}
## Mandate
{{3-5 sentences the sponsor can repeat. Owns: strategy, governance, portfolio. Does not own: hands-on build, delivery line management. Reports to: {{sponsor}}. Decision rights: {{approves X, recommends Y, escalates Z}}.}}
## First 90 days
- Days 1-30 (Assess): {{current-state assessment + opportunity longlist}}
- Days 20-60 (Visible win): {{the one win and how it gets driven}}
- Days 45-90 (Governance scaffold): {{policy, risk register, working group, decision path}}
## Governance and risk
- AI use policy: {{scope, gates, forbidden zones}}
- Risk register: {{owner, review cadence}}
- Governance framework: {{review path, criteria, decision record}}
- Working group: {{members, who chairs after exit}}
## Opportunity portfolio
| Opportunity | Value | Feasibility | Verdict |
|---|---|---|---|
| {{name}} | {{H/M/L + size}} | {{H/M/L + constraint}} | {{do now / queue / watch / stop}} |
## Operating cadence
- Exec sync: {{frequency}}
- Working group: {{frequency}}
- Portfolio review: monthly
- Board readout: quarterly
- Async update: weekly written
## Deliverables
- Month 1: {{current-state assessment, opportunity portfolio}}
- Month 2-3: {{visible win result, governance scaffold}}
- Ongoing: {{monthly portfolio review, quarterly board readout}}
- Exit: {{handoff doc, capability transfer plan}}
## Pricing / retainer
- Model: monthly retainer, {{6-12}} month term
- Rate: {{$X/mo for ~Y days/month}}
- Structure: {{flat retainer, or declining retainer as capability transfers}}
- Out of scope (costs extra): {{hands-on build, additional workshops, tooling spend}}
## Success metrics
- {{Visible win shipped by day 90}}
- {{Governance scaffold operating without the CAIO present}}
- {{Portfolio decisions made and recorded; initiatives killed counts}}
- {{Capability transfer milestone reached by month {{N}}}}
## Exit / handoff
- Handoff target: {{named internal owner or full-time hire}}
- Capability signals: {{org runs its own portfolio review; working group self-chairs; risk register maintained internally}}
- Transition: {{30 days documenting, 15 days training the successor, final handoff, optional advisory tail}}
Step 8: Review
Ask before finalizing:
- Can the sponsor repeat the mandate in their own words, including what the CAIO does not do?
- Is there a concrete, visible win scheduled inside the first 90 days, not just analysis?
- Does the plan name who inherits this mandate, and does the retainer or scope plan for that handoff?
- Is the governance scaffold light enough that teams will use it instead of routing around it?
- If the client actually wants hands-on build, has that been reclassified out of this engagement?
Anti-patterns
| Anti-pattern | Why it fails | Do instead |
|---|---|---|
| CAIO as hands-on builder | Sells judgment but gets measured on shipped code; the mandate gets miscast and underused | Draw the build line in the mandate; reclassify build work as a separate delivery engagement |
| A quarter of assessment, no win | The org concludes the role is overhead and routes around it before governance lands | Schedule one visible win inside 90 days; let it buy credibility for the slower work |
| Governance theater | Heavy review on every experiment gets ignored; teams build in the shadows | Gate the irreversible and regulated; keep reversible, low-stakes work fast |
| No handoff plan | The engagement quietly becomes permanent dependence, which pays but fails the mandate | Name the inheriting owner on day one; build the scaffold they will operate; decline the retainer as capability transfers |
| Mandate with no decision rights | A CAIO who can only recommend cannot kill initiatives or set policy; nothing moves | Secure explicit decision rights and a clear reporting line before signing |
| Portfolio that only adds | Everything is a "do"; the calendar fills and focus dissolves | Force a verdict per opportunity, including stop; naming what to kill is the highest-leverage move |
Output location
Present the engagement plan as formatted text in the conversation for Kate to refine, then feed into /engagement-proposal.