chief-ai-officer-advisor

Overview of chief-ai-officer-advisor skill

What chief-ai-officer-advisor is for

chief-ai-officer-advisor is a strategic AI leadership skill for founders, startup executives, and CAIO-style operators who need board-level decisions rather than model implementation help. It focuses on four practical decisions: API vs fine-tune vs in-house build, AI regulatory risk classification, API-to-self-hosted cost economics, and AI team hiring sequence.

Use it when your question sounds like “Should we build this model?”, “Is this use case high-risk?”, “When does self-hosting make financial sense?”, or “Who should we hire next for AI?” It is especially useful as chief-ai-officer-advisor for Strategic Planning, because the repository includes decision frameworks plus Python calculators instead of only narrative advice.

Best-fit users and decisions

The best users are startup founders, product leaders, CTOs, AI strategy leads, and consultants preparing an AI roadmap, investment memo, architecture recommendation, or governance review. The skill is strongest when the decision has business constraints: cost ceiling, latency target, token volume, hiring stage, EU/US deployment exposure, or compliance obligations.

It is not meant to replace an ML engineering skill. If you need model training code, embeddings implementation, RAG tuning, GPU deployment, or prompt engineering tactics, use a more technical AI/ML skill first and return to this one for executive tradeoffs.

What makes the skill different

The repository backs the advisory workflow with four reference files and three stdlib-only Python scripts:

• references/model_buildvsbuy_strategy.md
• references/ai_cost_economics.md
• references/ai_risk_governance.md
• references/ai_team_org_evolution.md
• scripts/model_buildvsbuy_calculator.py
• scripts/ai_cost_economics.py
• scripts/ai_risk_classifier.py

That matters because many “AI strategy” prompts produce generic recommendations. This skill pushes the assistant toward structured inputs, 3-year TCO comparisons, regulatory risk tiers, breakeven analysis, and stage-based hiring logic.

How to Use chief-ai-officer-advisor skill

chief-ai-officer-advisor install and repository path

Install from the source repository with:

npx skills add alirezarezvani/claude-skills --skill chief-ai-officer-advisor

The skill lives at:

c-level-advisor/skills/chief-ai-officer-advisor

After install, read SKILL.md first to understand trigger conditions and scope. Then open the reference file that matches your decision. If your question involves numbers, inspect the related script before prompting so you can provide the expected fields instead of asking for vague advice.

Inputs that produce useful recommendations

For build-vs-buy, prepare: use case, expected QPS, monthly query volume, average input/output tokens, latency budget, quality requirement, domain specificity, fine-tuning data availability, ML team capacity, and any self-hosting requirement.

For AI risk, prepare: domain, EU deployment, US states, decision impact, automation level, whether the system is user-facing, biometric processing, and whether children are involved.

For cost economics, prepare: monthly input/output token volume, required model quality tier, self-hosted model size class, target latency, utilization assumption, and whether to include operations cost.

For team planning, prepare: company stage, current team, product maturity, AI roadmap, existing eval infrastructure, and the bottleneck preventing shipment.

Turn a rough goal into a strong prompt

Weak prompt:

Should we fine-tune or use an API?

Stronger prompt:

Use the chief-ai-officer-advisor skill to evaluate API vs fine-tune vs build for a B2B SaaS support-response feature. Peak QPS is 5, monthly volume is 4M queries, average tokens are 800 in and 200 out, latency budget is 2 seconds, required quality is frontier-level, domain specificity is moderate, we have no labeled fine-tuning dataset, one ML-capable engineer, and no hard self-hosting compliance requirement. Give a 3-year TCO comparison, recommendation, failure modes, and what evidence would change the decision.

The stronger version lets the skill apply its calculators and decision thresholds rather than guessing from startup stereotypes.

Suggested workflow for chief-ai-officer-advisor usage

Start with one decision, not an entire AI transformation plan. Ask for a first-pass recommendation, then run or reference the relevant script for deterministic estimates. Challenge the output with changed assumptions: higher token volume, stricter latency, new EU deployment, or a future Series B hiring plan.

For numeric work, create a JSON profile matching the script schema and run:

python scripts/model_buildvsbuy_calculator.py path/to/use_case.json

python scripts/ai_cost_economics.py path/to/workload.json

python scripts/ai_risk_classifier.py path/to/use_case.json

Use the assistant to interpret results, identify missing assumptions, and convert the recommendation into a board memo or operating plan.

chief-ai-officer-advisor skill FAQ

Is chief-ai-officer-advisor only for Chief AI Officers?

No. The name reflects the perspective, not the required job title. Founders, CTOs, product executives, and fractional advisors can use it when they need CAIO-style judgment: capital allocation, risk posture, model sourcing, and organizational sequencing.

How is this better than an ordinary AI strategy prompt?

A generic prompt may say “start with APIs” or “consider compliance.” The chief-ai-officer-advisor skill gives the assistant a more specific operating model: TCO fields, regulatory risk categories, EU AI Act and US state law triggers, breakeven logic, and stage-based hiring patterns. It is still advisory, but it reduces guesswork.

Can beginners use this skill?

Yes, if they can describe the business use case. You do not need ML expertise to use it, but you should know basic constraints such as expected volume, latency tolerance, deployment geography, and whether the AI output affects consequential decisions. If those are unknown, ask the skill to produce a discovery questionnaire first.

When should you not use it?

Do not use it as legal advice, production architecture, model benchmark truth, or a substitute for current vendor pricing. The risk classifier is a governance triage aid, not counsel. The cost references include pricing assumptions that should be verified quarterly. For implementation, pair it with engineering-specific skills.

How to Improve chief-ai-officer-advisor skill

Improve chief-ai-officer-advisor results with evidence

The skill performs best when you provide real operating data instead of aspirational plans. Replace “high volume” with monthly tokens or queries. Replace “low latency” with p95 target milliseconds. Replace “regulated” with countries, states, domain, affected users, and decision consequences. Better inputs make the recommendation auditable.

Common failure modes to watch

The main failure mode is treating strategic advice as deterministic truth. API pricing changes, frontier model quality changes, and regulatory interpretation evolves. Another failure mode is over-indexing on model cost while ignoring engineering overhead, rate limits, eval infrastructure, security review, vendor procurement, and on-call burden.

A third failure mode is hiring too early. The team-org reference is intentionally skeptical of premature ML or research hires before product-market fit, evals, and a clear capability bottleneck exist.

Iterate after the first output

After receiving a recommendation, ask for sensitivity analysis:

• What changes if token volume grows 10x?
• What if EU deployment starts next quarter?
• What if output latency must fall below 500 ms?
• What if we obtain 50k labeled examples?
• What if the API bill exceeds $50k/month?

Then ask for a decision memo with “recommendation, assumptions, risks, reversible decisions, irreversible decisions, and next 30 days.” This turns the chief-ai-officer-advisor guide from abstract strategy into an execution artifact.

Customize the skill for your organization

To improve the skill locally, add your current vendor pricing, approved model providers, security requirements, legal review checklist, cloud GPU rates, hiring bands, and internal risk taxonomy. Keep the original references intact, but add company-specific constraints so the assistant stops recommending options your organization cannot actually approve.