Fractional
CAIO Partner
Most organizations are stuck in pilot mode. They pay for tools nobody uses, there is no AI policy, and no one owns the work.
You may not need a full-time AI executive. You need the function. Here is the full function, in five phases.
Discover and Assess
Find out what is real today before you plan anything.
Shadow AI discovery
Find every AI tool staff use, and map it back to what is approved.
Shadow AI discovery
Find every AI tool staff use, and map it back to what is approved.
- Get the list of currently approved tools
- Send an AI use survey to all employees
- Map the difference between approved tools and actual use
- Review the features inside each unapproved tool
- Map each feature back to an approved tool, or flag it as a gap
Workflow audit
Map where hours go each week and find the repetitive work.
Workflow audit
Map where hours go each week and find the repetitive work.
- Interview team leads about their weekly tasks
- Log time on repetitive work for one week
- Rank tasks by hours spent and frequency
- Flag tasks that touch sensitive data
Data classification
Decide what data can go into which tool.
Data classification
Decide what data can go into which tool.
- List the main data types and where they live
- Assign a level to each: public, internal, confidential, regulated
- Match each level to the tools that can hold it
- Write the rules in a one-page data guide
Risk and regulatory review
Check exposure in hiring, HR, contracts, and insurance.
Risk and regulatory review
Check exposure in hiring, HR, contracts, and insurance.
- Find where AI touches hiring and HR decisions
- Check client contracts for AI and data clauses
- Identify the state and industry rules that apply, such as the Colorado AI Act
- Review the cyber insurance policy for AI requirements
- Rate each risk by likelihood and impact
Readiness baseline
Measure where people start so you can show progress later.
Readiness baseline
Measure where people start so you can show progress later.
- Hold a strategy call with leadership to align on goals
- Survey staff skills and attitudes toward AI
- Record the baseline scores
- Identify early adopters and people who resist
Strategy
Choose the few use cases that matter and set a clear plan.
Use case inventory
Collect every AI idea in one list.
Use case inventory
Collect every AI idea in one list.
- Run an idea session with each team
- Add ideas from the survey and the workflow audit
- Write each use case in one sentence with an owner
- Remove duplicates
Impact and effort scoring
Start with high impact, low effort work.
Impact and effort scoring
Start with high impact, low effort work.
- Score impact: time, cost, and quality
- Score effort: data, tools, and skills needed
- Plot each use case on an impact and effort grid
- Review the top picks with leadership
90-day roadmap
Sequence the top use cases into three 30-day sprints.
90-day roadmap
Sequence the top use cases into three 30-day sprints.
- Select three to five priority use cases
- Assign each to a 30-day sprint
- Set a milestone for each sprint
- Get sign-off from the owner
Success metrics
Define numbers the owner can see.
Success metrics
Define numbers the owner can see.
- Capture the before numbers for each use case
- Pick one or two metrics per use case
- Set a target for each metric
- Agree on how often you report
Ownership and budget
Name who owns AI and what it costs.
Ownership and budget
Name who owns AI and what it costs.
- Name an executive sponsor
- Name an internal AI lead
- Set the budget for tools and training
- Define who approves new tools and use cases
Governance
Make the safe path the fast path.
AI acceptable use policy
A short policy staff can read in ten minutes.
AI acceptable use policy
A short policy staff can read in ten minutes.
- Draft the policy from the data classification rules
- Review it with leadership, HR, and legal
- Publish it and collect staff acknowledgment
- Set a date for the next review
Approved tools list
Show which tools are approved, for which data.
Approved tools list
Show which tools are approved, for which data.
- List each approved tool and the data levels it can hold
- Create a request form for new tools
- Define review criteria: security, privacy, and cost
- Publish the list where staff can find it
Human review rules
Define which outputs a person must check.
Human review rules
Define which outputs a person must check.
- List outputs that go to clients, candidates, or the public
- Assign a review level to each output type
- Name the reviewers
- Add the review step to workflow checklists
Logging and evidence
Keep the records auditors and cyber insurers ask for.
Logging and evidence
Keep the records auditors and cyber insurers ask for.
- Turn on admin audit logs in each approved tool
- Set how long you keep the logs
- Keep records of policy acknowledgment and training
- Collect the evidence your insurer asks for
Agent approval path
Review AI agents before they act on company systems.
Agent approval path
Review AI agents before they act on company systems.
- Define what counts as an agent
- Create an intake form: purpose, data, and permissions
- Set risk tiers with an approval level for each
- Schedule a review of each approved agent
AI incident response
Add AI events to your incident response plan.
AI incident response
Add AI events to your incident response plan.
- Define what an AI incident is, such as a data leak into a tool
- Add AI response steps to the incident response plan
- Name who to contact
- Run a tabletop exercise
Tech Stack
Pick the right platform and set it up securely.
Platform selection
Compare options against your use cases and data rules.
Platform selection
Compare options against your use cases and data rules.
- List requirements from the roadmap and data rules
- Shortlist two or three platforms
- Run a short pilot with champions
- Write a recommendation for the owner
Licensing tier
Choose the plan with admin controls and data protection.
Licensing tier
Choose the plan with admin controls and data protection.
- Compare admin controls and data terms by plan
- Count seats by role
- Confirm the contract terms on data use
- Purchase the licenses
Security configuration
Set up the platform before staff use it.
Security configuration
Set up the platform before staff use it.
- Turn on single sign-on (SSO)
- Set data retention
- Confirm model training is off for your data
- Set user roles and permissions
- Document the configuration
Integrations and connectors
Connect only what the roadmap needs.
Integrations and connectors
Connect only what the roadmap needs.
- Pick the connectors the roadmap needs
- Test each one with the least access it needs
- Get security approval
- Document each connection
Subscription cleanup
Cancel what you do not need.
Subscription cleanup
Cancel what you do not need.
- Pull AI spend from expense and card statements
- Match the spend to the discovery results
- Cancel redundant subscriptions
- Move those users to the approved tool
Implement and Train
Get people using AI well, and prove the value.
Role-based workshops
Train each team on its own work.
Role-based workshops
Train each team on its own work.
- Build content from each team's use cases
- Deliver hands-on sessions
- Share a prompt library for each team
- Collect feedback after each session
Champion network
One or two people per team who help others.
Champion network
One or two people per team who help others.
- Nominate champions from each team
- Give champions advanced training
- Hold a monthly champion meeting
- Share wins across the company
Use case coaching
Work 1:1 until each workflow runs without help.
Use case coaching
Work 1:1 until each workflow runs without help.
- Schedule 1:1 sessions with use case owners
- Build and test the workflow together
- Document the workflow as a standard procedure
- Hand it off to the owner
Adoption tracking
See who uses the tools and for what.
Adoption tracking
See who uses the tools and for what.
- Pull usage data from the admin dashboard
- Compare usage against the baseline
- Follow up with teams that have low adoption
ROI reporting
Report results against the phase two metrics.
ROI reporting
Report results against the phase two metrics.
- Measure the after numbers
- Compare them against the baseline
- Write a one-page report
- Present it to the owner
Quarterly review
Keep the program current.
Quarterly review
Keep the program current.
- Review metrics and incidents
- Update the roadmap
- Update the policy and the tools list
- Survey staff again
Who this is for
Start with phase one
A discovery engagement shows what your people already use, where the risk is, and what to do first. It stands on its own, and it is the input to everything after it.