Introduction
A Director of AI is measured on whether AI investments deliver business value. The role should center on strategy, capital, risk, talent and adoption, while architects own the design.
The title Director of AI covers several different roles, depending on the size of the organization and where its AI work sits. This post addresses the enterprise leader who runs AI enablement across several business portfolios.
|
Type of Director of AI |
Typical setting |
Main focus |
How this post applies |
|---|---|---|---|
|
Head of AI at a startup or scale-up |
One product, small teams |
Builds models and AI features hands-on |
Partly: the nine areas apply, the tracker is more than needed |
|
Business-unit Director of AI |
One product line or division |
AI features and adoption within that unit |
Partly: one portfolio, fewer epics |
|
Enterprise Director of AI or SVP Technology |
Several portfolios, many organizations, a shared AI platform |
Platform, portfolio, budget, risk and adoption |
Fully: this is the role the post is written for |
|
AI transformation leader |
Consulting firms or investors with many client companies |
Business value per client, delivery under deadlines |
Partly: portfolios become client companies, objectives become financial impact |
|
Director of AI research |
Research labs and R&D groups |
Research agenda, publications and research talent |
Little: the tracking model is built for delivery, not research |
Contents
- Where AI Leaders Lose Focus
- Who Owns What
- The Nine Areas a Director of AI Owns
- From Strategy to Execution
- Budget Tracking
- AI Application Registry
- The Director’s Daily View
- Leading Several Portfolios
- Strategic Velocity
- Portfolio Roadmap
- A Weekly Check
- Conclusion
The Director’s daily view is calculated, not written. Teams update stories as part of their normal work, progress, spend and application status roll up automatically, and the leader records only three moves a day.
Where AI Leaders Lose Focus
Talks such as this session on governed AI agents cover gateways, tool servers, memory and token optimization, which is essential material for architects.
When senior leaders adopt that material as their agenda, they drift into system design, and decisions on investment, risk and adoption are left without an owner.
Who Owns What
|
Role |
Owns |
Example question |
|---|---|---|
|
SVP / Director of AI |
Why and whether |
Should we fund one shared agent platform? |
|
AI Architect / Platform lead |
How |
How do agents authenticate to internal tools? |
|
Engineering teams |
What |
Build the claims agent on the platform |
The Nine Areas a Director of AI Owns
Each area below is a decision the leader is accountable for, paired with the evidence that shows whether the decision is working.
|
Area |
What the leader decides |
Evidence it is working |
|---|---|---|
|
Strategy and portfolio |
Which business problems justify agents, and which pilots move to production or stop |
Each initiative is tied to a business KPI, and the number of stalled pilots is falling |
|
Platform investment |
Whether to fund one shared platform, and whether to build, buy or partner |
Teams build on the shared platform rather than creating their own stacks |
|
Autonomy and risk appetite |
Which actions agents may take alone, which need approval, and who in the business signs off |
Every production agent has an approved autonomy level and a named escalation path |
|
ROI |
How value is measured and reported to the CEO, CFO and board |
A quarterly value report exists, and low-value agents are retired |
|
Cost |
The AI budget, chargeback to business units, and cost per task against the human baseline |
Spend stays within budget, and cost per task declines quarter over quarter |
|
Organization and talent |
Team structure, hiring versus upskilling, and ownership of each agent |
Every agent has a business owner and a technical owner, and critical roles are filled |
|
Adoption |
How workflows are redesigned around agents and how employees are trained |
Active usage grows, and redesigned workflows show measurable cycle-time gains |
|
Executive alignment |
The steering forum, the funding model and the review cadence |
Legal, finance, security and business leaders review the roadmap on a fixed schedule |
|
Vendors and ecosystem |
Which model providers and partners to commit to, and when to switch |
Contracts avoid single-vendor lock-in, and vendor reviews happen at least twice a year |
From Strategy to Execution
At enterprise scale, the leader needs one connected system of record in which every piece of work traces back to a strategic objective. Each item carries its portfolio, its organization and one of the nine categories.
|
Level |
Owned by |
Typical count |
Time span |
Tagged with |
|---|---|---|---|---|
|
Strategic objective |
Director or SVP |
5 to 7 |
Year |
None |
|
Epic |
Portfolio lead |
5 to 10 per portfolio |
Quarter |
Portfolio and category |
|
Feature |
Organization lead |
3 to 8 per epic |
Month |
Organization and category |
|
Story |
Delivery team |
5 to 15 per feature |
1 to 2 weeks |
Organization and category |
Progress is calculated from completed work rather than entered by hand, which keeps status reporting consistent across portfolios.
|
Level |
How percentage progress is calculated |
|---|---|
|
Story |
Done or not done, weighted by its effort estimate |
|
Feature |
Completed story effort divided by total story effort |
|
Epic |
Average of its features, weighted by effort |
|
Objective |
Average of its epics, weighted by budget or priority |
Budget Tracking
Budget is tracked at the epic level and rolled up to each portfolio, so spend can be compared directly with progress.
|
Field |
Meaning |
|---|---|
|
Approved budget |
Funding allocated for the quarter or year |
|
Committed |
Contracts and headcount already signed |
|
Actual spend |
Split into people, cloud, model usage and vendors |
|
Forecast at completion |
Expected total spend when the epic closes |
|
Variance |
Forecast minus approved budget, flagged above an agreed threshold |
|
Cost per point of progress |
Actual spend divided by percentage complete, which shows when spend is outpacing delivery |
AI Application Registry
Every AI application enabled on the enterprise AI platform is listed in one registry, with a named owner, a lifecycle stage and a link to the epic that funds it.
|
Field |
Example |
|---|---|
|
Application ID and name |
AIA-042, Quote assistant |
|
Portfolio and organization |
Commerce, Quotes |
|
Business owner and technical owner |
One named person for each |
|
Lifecycle stage |
Proposed, Approved, Pilot, Production or Retired |
|
Autonomy level |
Recommend only, Act with approval, or Act alone |
|
Platform services enabled |
Gateway, memory, evaluation, sensitive-data blocking, context optimization |
|
Models used |
Provider and model version |
|
Monthly cost |
Taken from the budget tracker |
|
Quality score |
Most recent evaluation result |
|
Linked epic |
The epic that funds the application |
|
Last review date |
Date of the most recent governance review |
The Director’s Daily View
The daily view is calculated from the work hierarchy, the budget tracker and the registry, so it does not grow with the number of organizations. The only part the leader writes is a short list of three moves for the day.
|
Panel |
What it shows |
|---|---|
|
Objective progress |
Percentage complete for each objective, with the change since the previous day |
|
Epics at risk |
Epics rated Amber or Red, with the owner and the blocker |
|
Budget alerts |
Epics over their variance threshold, or spending faster than they deliver |
|
Registry changes |
Applications that changed stage, entered production or reported an incident |
|
Three moves today |
The leader’s own actions for the day, each linked to an epic |
Any action that cannot be linked to an epic does not belong on the list, which keeps the daily view focused at any organizational size.
Leading Several Portfolios
A leader with four portfolios and about eighty organizations cannot track every organization daily. The journal therefore stays at portfolio level, and organizations report upward on a fixed cadence or when something needs a decision.
|
Level |
Who records |
Cadence |
What is recorded |
|---|---|---|---|
|
Director |
You |
Daily |
Reviews the daily view and records up to three moves linked to epics |
|
Portfolio |
Four portfolio leads |
Weekly |
Epic progress, budget and registry updates, with escalations |
|
Organization |
About eighty organization leads |
Continuous |
Feature and story progress, updated as part of normal delivery work |
The grouping below is illustrative and should follow your actual reporting lines.
|
Portfolio |
Organizations |
Categories that need the most attention |
|---|---|---|
|
Commerce (quote to cash) |
Product catalog, deals, quotes, orders, billing, subscriptions |
ROI, Adoption, Autonomy and risk appetite |
|
Customer and partner |
Customer onboarding, partner integrations, reports |
Adoption, Executive alignment |
|
Platforms |
Cloud platforms, data platforms, agentic platforms, core platform |
Platform investment, Cost, Vendors and ecosystem |
|
Reliability, risk and research |
SLOs, SLIs and SRE, regulation and compliance, security, research |
Autonomy and risk appetite, Organization and talent |
Strategic Velocity
At this level, velocity measures how quickly decisions are made and value reaches production. Story points and sprint burn-down stay with the delivery teams.
|
Measure |
How it is calculated |
Source |
Cadence |
|---|---|---|---|
|
Decision velocity |
Decisions made per week, by portfolio |
Director’s daily view |
Weekly |
|
Decision latency |
Average days from an escalation to a decision |
Director’s daily view |
Weekly |
|
Overdue follow-ups |
Follow-ups past their date |
Director’s daily view |
Weekly |
|
Pilot-to-production time |
Average days from pilot start to production |
Epic and feature records |
Monthly |
|
Agents in production |
Net agents added, by portfolio |
Monthly scorecard |
Quarterly |
|
Value realized |
Value delivered against plan |
ROI report |
Quarterly |
Portfolio Roadmap
The roadmap belongs in the monthly review. The daily view shows whether epics are progressing, and the roadmap shows whether that progress is keeping milestones on schedule.
Platform general availability in July gates the production rollouts in the Commerce and Customer and Partner portfolios. The dates and initiatives are illustrative and should be replaced with each portfolio’s actual plan.

The dates and initiatives are illustrative and should be replaced with each portfolio’s actual plan.
A Weekly Check
Each Friday, review the week’s three-move lists and ask what share of your time went to why and whether, and what share went to how.
If most of it went to how, delegate those items to the architects and bring your attention back to the nine areas above.
Conclusion
A Director of AI is accountable for whether AI delivers value across the portfolios, not for how each system is built. That accountability needs a tracking model that stays the same size as the organization grows.
The model in this post links every story to an epic and every epic to an objective, so progress is calculated rather than reported. Budget and the AI application registry are tied to the same epics, so spend, progress and application status can be compared in one place.
With this in place, the leader’s daily effort is limited to reviewing the calculated view and recording three moves. The remaining time goes to the decisions only the leader can make: where to invest, how much risk to accept and how to drive adoption.