Sheet 09
Leadership dashboard
What a Director of Technology would see each month. Every number carries an evidence label and the calculation behind it, because the difference between a measured figure and a self-reported one is the difference between a decision and a guess.
Before reading any number
All figures are fictional demonstration data for a fictional organization. Every metric carries an evidence label because the difference between a measured number and a self-reported one is the difference between a decision and a guess. In a real program these labels would be defined with the leadership audience before the first report.
Every date calculation on this page is measured from a fixed reference date of 02 Sep 2026, so the sample data stays internally consistent instead of quietly ageing.
Evidence key
This comes before the numbers on purpose. A dashboard where every figure looks equally solid invites a leadership team to act on the weakest one.
- Measured
- Counted from records or captured against a documented baseline. Reproducible by someone else.
- Estimate
- Calculated from measured inputs using stated assumptions. The assumptions are shown; change them and the number changes.
- Self-reported
- Reported by participants. Useful for direction and sentiment, not for a business case on its own.
- Pending validation
- Not yet established. Shown so it is visible that it is missing, rather than omitted.
Portfolio
Counted from the registers rather than entered here, so a figure on this page cannot contradict the record it came from.
Registered use cases
10
- How this is calculated
- Count of records in the use-case register, all stages included.
Pilots underway
4 of 5
- How this is calculated
- Pilots in approved, active, measuring, or ready-to-scale stages. One pilot is paused and is counted separately.
- What this does not say
- PIL-005 is paused pending vendor documentation, not cancelled.
Recorded decisions
6
- How this is calculated
- Entries in the decision log across all forums, each with a named decision owner and the alternatives that were rejected.
Tools inventoried
2 of 5 approved
- How this is calculated
- Tool inventory records with a status of approved or approved with conditions, against all tools tracked. The remaining three are in pilot, not approved, or on a paused review.
Exceptions on the log
3
- How this is calculated
- Entries in the exception log. An exception is a documented, time-limited departure from a standard expectation, with compensating controls and an exit plan.
- What this does not say
- 2 are within 45 days of expiry. An exception that lapses without action means the underlying activity has to stop.
Median time from intake to pathway assignment
3 business days
- How this is calculated
- Submission timestamp to recorded pathway assignment, across 64 submissions since March.
- What this does not say
- Assignment is not resolution. Structured and elevated reviews continue from there.
Where the work sits
Two views of the same ten use cases: how far each has travelled, and how much review effort each was assigned.
Show the data as a table
| Category | Use cases |
|---|---|
| Discover | 0 |
| Intake | 1 |
| Triage | 1 |
| Review | 3 |
| Pilot | 2 |
| Measure | 2 |
| Decide | 1 |
Show the data as a table
| Category | Use cases |
|---|---|
| Standard enablement | 2 |
| Structured review | 5 |
| Elevated review | 3 |
Adoption
Monthly active users of approved tools
268
- How this is calculated
- Distinct users with at least one session in the trailing 30 days, from tool administrative reporting.
- Trend
- Up from 48 in January. Growth is flattening as the early-adopter population saturates.
License utilization
64%
- How this is calculated
- 337 active users against 523 licenses held across all tools.
- What this does not say
- 152 assigned assistant licenses show no activity in 30 days. Reclaim or redeploy before renewal — tracked as ACT-047.
Employees who have completed at least one session
389 (31%)
- How this is calculated
- Distinct attendees across six delivered sessions, deduplicated. Attendance is not the same as capability.
- What this does not say
- Show rate is 80% of registrations. Average feedback 4.4 of 5.
Show the data as a table
| Month | Monthly active users |
|---|---|
| Jan | 48 |
| Feb | 96 |
| Mar | 142 |
| Apr | 178 |
| May | 205 |
| Jun | 238 |
| Jul | 259 |
| Aug | 268 |
The line is flattening, and that is the finding. Growth from 48 to 268 came largely from people who wanted these tools before the program existed. The next 200 users are people who do not, and reaching them is an enablement problem rather than an access problem — which is why the barriers table below matters more than this chart does.
- Standard enablement
- Structured review
- Elevated review
Show the data as a table
| Category | Standard enablement | Structured review | Elevated review |
|---|---|---|---|
| Mar | 4 | 2 | 1 |
| Apr | 6 | 3 | 1 |
| May | 9 | 2 | 0 |
| Jun | 7 | 4 | 1 |
| Jul | 11 | 3 | 0 |
| Aug | 8 | 2 | 1 |
Includes questions resolved with guidance, which is why the total exceeds the ten registered use cases. Most requests never need a review — that is the point of the pathway model.
Value and cost
The estimate and the validated figure appear together, always. An estimate on its own becomes a business case by the third time it is repeated, and nobody remembers what it rested on.
Hours returned to participants
~820hours
- How this is calculated
- Measured per-task time reduction in PIL-001 and PIL-003 multiplied by observed volume. PIL-001: 2.6 hours saved across 38 pursuits with 22 writers. PIL-003: 4.7 minutes per lookup across an observed 2,100 lookups.
- What this does not say
- Time returned is not the same as time redeployed to higher-value work. Nothing here claims a cost reduction.
Hours returned, independently validated
~180hours
- How this is calculated
- Subset of the estimate above confirmed by direct observation: timed sampling of 14 pursuits against a pre-pilot baseline.
- What this does not say
- This is the number to put in a business case. The larger figure is directionally useful.
Pilot cost to date
$41,000
- How this is calculated
- Summed cost-to-date across all five pilots: licensing, vendor fees, and third-party support. Excludes participant time.
- What this does not say
- Participant time is the larger cost and is not captured here. A full business case would include it.
Documented value with a named owner
1 of 5 pilots
- How this is calculated
- Pilots where an owner has signed off on measured outcomes against the original success measures. Only PIL-001 has reached that point.
- What this does not say
- This is the honest state of a program eight months in. Most pilots have not finished measuring.
Pilot outcomes
Four dispositions across five pilots. One recommendation to scale after eight months is the honest state of a program this age.
Recommend scale
1 pilot
Measures met, with library maintenance named as a condition of scaling.
Continue measuring
2 pilots
Not yet started
1 pilot
Approved at reduced scope; two conditions outstanding before the September 8 start.
Paused
1 pilot
Vendor documentation gap. Resume criteria written; September 30 checkpoint.
What needs attention
There is no separate risk register in this sample. The program’s live risk is visible in three places instead: exceptions approaching expiry, actions blocked on a party outside the firm’s control, and decisions that have a date on them. 1 action is blocked and 0 are past due as of the reference date.
Open risks and blocked actions
10 open, 1 blocked
- How this is calculated
- Open action register, excluding closed items. Blocked means waiting on a party outside the firm's control.
Decisions pending at the next committee
4
- How this is calculated
- Items on the September 15 agenda requiring a committee decision: PIL-001 scale, PIL-003 scale threshold, PIL-005 resume or retire, and the disclosure practice.
Exceptions nearing expiration
2 of 3
- How this is calculated
- Exceptions expiring within 45 days. EXC-01 expires 2026-09-30, EXC-03 expires 2026-09-11.
- What this does not say
- An exception that lapses without action means the underlying activity has to stop.
Exceptions nearing expiration
2 of 3 entries expire within 45 days of the reference date. Each one has an exit plan; the exit plan is what stops an exception from becoming an undocumented standard.
| ID | Exception granted | Approved by | Expires | Status | Linked records |
|---|---|---|---|---|---|
| EXC-03 | Continue using an unapproved meeting transcription tool for 60 days while migrating to the approved assistant. | Marcus Dell, Information Security Manager | 11 Sep 2026in 9 days | Expiring | TL-01 |
| EXC-01 | Use Draftline for public-facing article drafts before the AI-assisted content disclosure practice is finalized. | Priya Raman, Director of Technology, with Risk concurrence | 30 Sep 2026in 28 days | Expiring | UC-005TL-03 |
| EXC-02 | Proceed with the Cadence Search pilots before the vendor penetration-test summary is received. | Marcus Dell, Information Security Manager | 15 Oct 2026in 43 days | Active | UC-003UC-007PIL-002PIL-003TL-02DEC-2026-03 |
Decisions with a scheduled forum
Drawn from the pilot records. A pilot with no scheduled decision drifts until someone loses interest, which is how programs accumulate work nobody is willing to stop.
| Pilot | Decision required | Business owner | Forum and date |
|---|---|---|---|
| PIL-001 | Scale to all pursuit teams, with library maintenance named as a condition | Dee Okafor, Marketing & Pursuits Director | AI Governance Committee, 2026-09-15 |
| PIL-002 | Continue, modify scope, or pause at the week-16 review | Alan Petrov, QA Lead, Structural Practice Group | Biweekly pilot check-in, then AI Governance Committee 2026-10-06 |
| PIL-003 | Scale, extend the measurement window, or hold pending library cleanup | Tomas Reyes, Data & Reporting Manager | AI Governance Committee, 2026-09-15 |
| PIL-004 | Confirm pilot start once the two remaining conditions are evidenced | Grace Lin, People & Culture Director | Weekly intake triage, then December governance review |
| PIL-005 | Resume if documentation is executed by 2026-09-30; otherwise recommend retiring the vendor evaluation | Sofia Brand, Controller | AI Governance Committee, 2026-09-15 |
Open actions
Every action carries an owner and a date. Blocked means the firm is waiting on an outside party, which is a different problem from an action nobody has started.
| ID | Action | Owner | Raised in | Due | Status |
|---|---|---|---|---|---|
| ACT-044 | Update applicant privacy notice before recruiting pilot start | Elena Ward, Risk, Legal & Privacy | AI Governance Committee, 2026-05-18 | 02 Sep 2026today | In progress |
| ACT-048 | Return two clarifying questions to the Civil deliverable-summarization requester | AI Program Manager | Weekly intake triage, 2026-08-24 | 04 Sep 2026in 2 days | Open |
| ACT-041 | Retire superseded standards revisions blocking citation accuracy | Tomas Reyes, Data & Reporting | Biweekly pilot check-in, 2026-08-04 | 05 Sep 2026in 3 days | In progress |
| ACT-045 | Complete recruiter and hiring-manager training for the reduced-scope pilot | Grace Lin, People & Culture | AI Governance Committee, 2026-05-18 | 05 Sep 2026in 3 days | In progress |
| ACT-042 | Draft AI-assisted content disclosure practice for committee adoption | Dee Okafor with Elena Ward | AI Governance Committee, 2026-07-21 | 10 Sep 2026in 8 days | In progress |
| ACT-046 | Confirm Denver transcription migration and close the vendor account | Transportation Practice Group with IT & Security | Weekly intake triage, 2026-06-15 | 11 Sep 2026in 9 days | In progress |
| ACT-049 | Define resume criteria checkpoint with Ledgerly ahead of the September 30 decision | Sofia Brand with Elena Ward | Risk review, 2026-07-24 | 20 Sep 2026in 18 days | Open |
| ACT-050 | Publish verification-evidence expectation for AI-assisted calculation utilities | Alan Petrov, QA Lead | Weekly intake triage, 2026-08-17 | 25 Sep 2026in 23 days | Open |
| ACT-043 | Obtain vendor penetration-test summary from Cadence Data Systems | Marcus Dell, IT & Security | Security review, 2026-04-02 | 01 Oct 2026in 29 days | Blocked |
| ACT-047 | Reclaim or redeploy 152 inactive assistant licenses before renewal | Marcus Dell with Finance & Procurement | Monthly enablement and adoption review, 2026-08-11 | 31 Oct 2026in 59 days | Open |
Top adoption barriers
Collected from office hours, listening sessions, and pilot check-ins. Mentions are a count of how often something came up, not a measure of how serious it is — the job-security concern is mentioned least and matters most.
| Barrier | What the signal is | Source | Program response | Mentions |
|---|---|---|---|---|
| Uncertainty about client data | Most frequent office-hours question; cited in listening sessions across every practice group | Office hours, listening sessions | One-page reference drafted with Risk; per-agreement answers handled through intake rather than by general rule. | 11 |
| No time to learn something new | Training show rate at 78% of registrations; most no-shows cite deadlines | Training feedback, manager forum | Shorter modules, sessions using participants' own work, and recordings. Hands-on sessions capped at 30 so they stay useful. | 9 |
| Doubt the output is worth reviewing | Reported by experienced engineers who tried it once on a poorly suited task | Practice group listening sessions | Discipline-specific worked examples rather than generic demos. Being explicit about tasks these tools are bad at. | 7 |
| Unclear whether AI-assisted work is acceptable to submit | Raised by proposal writers and by engineers preparing deliverables | Pilot check-ins, office hours | Disclosure practice in development with Marketing and Risk; interim expectation published under EXC-01. | 6 |
| Job security concern | Raised in every listening session, rarely in writing | Listening sessions | Routed to leadership rather than answered by the program. Pilot objectives and measures published so people can see what is actually being pursued. | 5 |
How the program is perceived
Program perceived as helpful rather than blocking
3.9 of 5
- How this is calculated
- Single question in the quarterly program survey, n=142 of 1,240 employees. Prior quarter 3.4.
- What this does not say
- Self-selected respondents skew toward people who have used the program. The 1,098 who did not respond are the interesting population.
Why this sits at the bottom of the page
It is the only self-reported figure here, and it is the one a program is most tempted to lead with. Placing it last is a decision about what leadership should weigh, not an accident of layout.
Pathway key
- Standard enablement
Answered in the weekly triage, typically within five business days
- Structured review
Two to three weeks from complete submission to decision
- Elevated review
Four to six weeks, longer if the evidence a reviewer needs does not exist yet