AI Level 1
Some Guardrails, Some Users
What This Looks Like
- You've provisioned Copilot / Claude / ChatGPT seats for a few curious teams.
- You have basic logging (e.g. Purview defaults).
- You don't distinguish between Internal and External models.
- Or you've banned AI outright, leading to increased shadow AI.
AI Level 2
Good Governance, Limited Adoption
What This Looks Like
- You have all appropriate data guardrails in place.
- Many of your staff have AI access, but only a few use it regularly.
- Your leadership team wants to start demonstrating real AI progress.
- You've begun to map out your AI adoption plan but haven't implemented it yet.
AI Level 3
Discovering Real AI Use Cases
What This Looks Like
- Your users are building repeatable workflows that move beyond general purpose chat.
- You have specific, high-value use cases connected to curated data.
- You've doubled the number of power users and weekly users quarter-over-quarter.
- Your wins are real but still not widespread.
AI Level 4
Department-Level AI Fluency
What This Looks Like
- Every employee in a leading department uses AI for key day-to-day tasks.
- You understand ROI per use case and have granular cost controls in place.
- Your board recognizes your significant process improvements directly attributable to AI.
- You want to replicate your AI success across all departments.
AI Level 5
Industry Leadership in AI
What This Looks Like
- You've driven significant, org-wide improvements through AI use.
- Your entire org has moved “beyond the hype” and balances AI tradeoffs and benefits at scale.
- Your leadership team and users are regularly asked to contribute to AI forums and conferences.
- AI is a significant competitive advantage leading directly to growth.
Level 1 → Level 2
Governance Improvements
- AI Use Policy Three to five pages, non-legalese, with approved vendors listed by name.
- Data Loss Prevention DLP and access rules established separately for Internal and External models.
- Data Management A hardened data perimeter where inference, logs, and application data remain inside your network.
- Logging Word-for-word auditability across every AI interaction.
- Automated Risk Scoring Every prompt scored on a fixed scale, with real-time filtering and alerting.
Where AI Initiatives Stall
The Zone of Failure
Five Rollout Pitfalls
- Power Users Only A few enthusiasts adopt and everyone else watches. When power users move on, so does the value.
- Prompting like Google Your casual users enter basic queries and get generic results. They never become power users.
- The Empty Textbox Problem Your non-users see a blank textbox and don't know where to start. “Figuring out AI” becomes something they'll do later.
- Cost Surprises Fear of surprise bills causes you to restrict power users while casual user quotas go unused.
- Over-Automation Improper AI automations give you brittle workflows and months of wasted effort.
