For each AI tool a company pays for: keep it, trim it, merge it, cancel it or look closer. And how sure are we?
The whole story, with what failed: read the case study.
Made by Chris Karageorge as a portfolio proof of concept, not a product.
Start with the six cards under The answers.
This page shows how a company could check the AI tools it pays for. For each tool it gives one answer. Keep it, trim it, merge it, cancel it, or look closer. It also says how sure we are.
Activity is not value. This method can show what is not used. It cannot prove what is worth keeping.
The hours saved are my guesses. You can change them below.
My first checker found every problem I planted. That proved little, because I wrote both the test data and the checker. So I added harder cases, and it missed some at first. The full story is at the bottom of this page.
Each card is one tool. The dollar figures are for one month.
Live seats
116
80 people, 6 tools, as of 2026-06-29. People can hold seats in several tools, so 116 seats is more than 80 people.
Seat cost a month
$4,285
ILLUSTRATIVE prices, 116 seats
Idle seats (certain)
$480 to $520 a month
13 to 14 of 116 seats. Certain because it counts only seats that are paid for and not used, whatever my guesses; the lower figure counts seats that were never used, whose owner left, or that have sat idle for 90 days or more, and the higher figure adds seats idle for 30 to 89 days.
Consolidation savings (a choice)
$0 to $315 a month
Up to 9 seats. Never added to the idle figure.
116 seats are paid for. The seat list near the end of the page has 120 rows, because 4 seats were already cancelled.
How sure we are: High means the answer holds across all my guesses and the data is complete. Low means the answer changes somewhere inside my guesses, or the data is thin in two ways at once. Medium is in between.
Consolidate (keep one of two tools)
The same people use two tools for the same work.
How sure we are: Low
Seats paid for but not used: $120 to $160 a month. Saving by merging tools: $0. Counted on Chat assistant B instead, so this card shows $0.
Consolidate (keep one of two tools)
The same people use two tools for the same work.
How sure we are: Low
Seats paid for but not used: $70 a month. Saving by merging tools: $0 to $315. Counted on this tool: the seats it shares with Chat assistant A would go, because it costs less.
Keep
None of the other four answers applies. It pays for itself at my base guesses and people keep what it makes. It overlaps no other tool, and fewer than one seat in ten is idle.
How sure we are: High
Seats paid for but not used: $0 a month. Saving by merging tools: $0. No other tool overlaps with this one.
Right-size (cancel unused seats)
Many seats are idle, never used, or belong to people who left.
How sure we are: Low
Seats paid for but not used: $90 a month. Saving by merging tools: $0. No other tool overlaps with this one.
Investigate (check before deciding)
People rarely keep what the tool makes, so the time saved cannot be trusted.
How sure we are that it needs a closer look: High
Seats paid for but not used: $50 a month. Saving by merging tools: $0. No other tool overlaps with this one.
Right-size (cancel unused seats)
Many seats are idle, never used, or belong to people who left.
How sure we are: High
Seats paid for but not used: $150 a month. Saving by merging tools: $0. No other tool overlaps with this one.
No seats in these states in this data: Missing usage records, Back from leave, On leave.
| Tool | Active | Light use | Idle | Never used | Left the company, seat still paid | On leave | Back from leave | Missing usage records | New starter | Live seats |
|---|---|---|---|---|---|---|---|---|---|---|
| Chat assistant A | 28 | 5 | 3 | 0 | 1 | 0 | 0 | 0 | 1 | 38 |
| Chat assistant B | 20 | 0 | 2 | 0 | 0 | 0 | 0 | 0 | 1 | 23 |
| Coding assistant | 6 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 6 |
| Image generator | 6 | 1 | 2 | 0 | 1 | 0 | 0 | 0 | 0 | 10 |
| Meeting transcriber | 16 | 7 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | 26 |
| Support-desk AI add-on | 9 | 0 | 1 | 1 | 1 | 0 | 0 | 0 | 1 | 13 |
These numbers are my guesses, not measurements. Usage logs cannot show how much time a tool saves. Change a guess and watch the answers move. Changes stay in this tab. Nothing is saved or sent.
Think of each tool as a counter in a members' café. Every paid seat is a stool. A stamp is a quarter of an hour saved by one active user. A card is one month of one seat. When the stamps fill the card, the seat has paid for itself.
An idle seat is an empty stool. Its rent is already counted in what each active user costs, so it is never charged twice.
Kestrel Supply Co is fictional. The data is synthetic and the prices are illustrative.
Your device asks for less motion, so this picture stays still.
One seat's card, at my base guesses
Key: each small square over a member is one full card earned in a month.
Chat assistant A: 33 active users sit at 38 stools. 4 members walk in and out while the animation plays. The others sit and rest. Each small square over a member is one full card earned in a month. 6 squares are drawn. The card needs 2.83 stamps and has 3 slots. An active user earns 17.32 stamps a month, so every slot is stamped.
Consolidate (keep one of two tools)
Consolidate (keep one of two tools)
Keep
Right-size (cancel unused seats)
Investigate (check before deciding)
Right-size (cancel unused seats)
What this does: it shows how many stamps one active user earns in a month against what the card needs, for Chat assistant A.
Low guess: 0.1 hours a week at $45 an hour. Cut (cancel the tool)
Each active person's saved time covers about 42% of their seat.
Base guess: 1 hour a week at $65 an hour. Consolidate (keep one of two tools)
Each active person's saved time covers their seat about 6 times over.
High guess: 2.5 hours a week at $90 an hour. Consolidate (keep one of two tools)
Each active person's saved time covers their seat about 21 times over.
Where this analogy stops. A stamp is a quarter of an hour of saved time. That is a display choice, not a measurement. Hours saved and break-even are not coffee. The same hours are assumed for every active user.
Sliders and the data switch need JavaScript. Everything else on this page works without it.
The answers at these guesses Tool, answer, and how sure we are.
One worked case, at my base guesses: At 1 hour a week, Chat assistant A saves about $9,288 a month across 33 active users. The seats cost $1,520, so what is left is $7,768.
What has moved: Nothing has moved.
| Counter | Active | Light use | New starter | Idle | Never used | Left the company, seat still paid | All stools |
|---|---|---|---|---|---|---|---|
| Chat assistant A | 28 | 5 | 1 | 3 | 0 | 1 | 38 |
| Chat assistant B | 20 | 0 | 1 | 2 | 0 | 0 | 23 |
| Coding assistant | 6 | 0 | 0 | 0 | 0 | 0 | 6 |
| Image generator | 6 | 1 | 0 | 2 | 0 | 1 | 10 |
| Meeting transcriber | 16 | 7 | 1 | 1 | 1 | 0 | 26 |
| Support-desk AI add-on | 9 | 0 | 1 | 1 | 1 | 1 | 13 |
| Counter | Guess | Hours a week | Hourly cost | Stamps needed | Stamps earned a month | Cards filled | Answer |
|---|---|---|---|---|---|---|---|
| Chat assistant A | Low | 0.1 | $45 | 4.09 | 1.73 | 0.42 | Cut |
| Chat assistant A | Base | 1 | $65 | 2.83 | 17.32 | 6.11 | Consolidate |
| Chat assistant A | High | 2.5 | $90 | 2.05 | 43.3 | 21.15 | Consolidate |
| Chat assistant B | Low | 0.1 | $45 | 3.58 | 1.73 | 0.48 | Cut |
| Chat assistant B | Base | 1 | $65 | 2.48 | 17.32 | 6.99 | Consolidate |
| Chat assistant B | High | 2.5 | $90 | 1.79 | 43.3 | 24.2 | Consolidate |
| Coding assistant | Low | 1 | $45 | 5.33 | 17.32 | 3.25 | Keep |
| Coding assistant | Base | 3 | $65 | 3.69 | 51.96 | 14.07 | Keep |
| Coding assistant | High | 6 | $90 | 2.67 | 103.92 | 38.97 | Keep |
| Image generator | Low | 0.1 | $45 | 3.81 | 1.73 | 0.45 | Cut |
| Image generator | Base | 0.5 | $65 | 2.64 | 8.66 | 3.28 | Right-size |
| Image generator | High | 1.5 | $90 | 1.9 | 25.98 | 13.64 | Right-size |
| Meeting transcriber | Low | 0.1 | $45 | 2.51 | 1.73 | 0.69 | Investigate |
| Meeting transcriber | Base | 0.5 | $65 | 1.74 | 8.66 | 4.98 | Investigate |
| Meeting transcriber | High | 1.5 | $90 | 1.26 | 25.98 | 20.68 | Investigate |
| Support-desk AI add-on | Low | 0.5 | $45 | 6.42 | 8.66 | 1.35 | Right-size |
| Support-desk AI add-on | Base | 2 | $65 | 4.44 | 34.64 | 7.79 | Right-size |
| Support-desk AI add-on | High | 5 | $90 | 3.21 | 86.6 | 26.98 | Right-size |
Chat assistant A has 38 stools. Sitting there: 28 active users, 5 light users, 1 new starter. Needing a look: 3 idle seats, 1 seat whose owner left. Chat assistant B has 23 stools. Sitting there: 20 active users, 1 new starter. Needing a look: 2 idle seats. Coding assistant has 6 stools. Sitting there: 6 active users. Image generator has 10 stools. Sitting there: 6 active users, 1 light user. Needing a look: 2 idle seats, 1 seat whose owner left. Meeting transcriber has 26 stools. Sitting there: 16 active users, 7 light users, 1 new starter. Needing a look: 1 idle seat, 1 seat never used. Support-desk AI add-on has 13 stools. Sitting there: 9 active users, 1 new starter. Needing a look: 1 idle seat, 1 seat never used, 1 seat whose owner left.
Low guess for Chat assistant A: a card needs 4.09 stamps and one active user earns 1.73 stamps a month, which is 0.42 cards. Base guess for Chat assistant A: a card needs 2.83 stamps and one active user earns 17.32 stamps a month, which is 6.1 cards. High guess for Chat assistant A: a card needs 2.05 stamps and one active user earns 43.3 stamps a month, which is 21.2 cards.
These five pictures use real numbers from the data. Each one has a table or a plain-words version under it.
A steady user
Last used 7 days ago.
A flat 30-day rule says: active
This method says: active
A month-end user (Finance)
Last used 35 days ago.
A flat 30-day rule says: idle
This method says: active, it follows the month-end cycle
A sale-season user (Retail)
Last used 189 days ago.
A flat 30-day rule says: idle
This method says: active, it follows the sale season
| Seat | Tool | Days since last use | A flat 30-day rule says | This method says |
|---|---|---|---|---|
| A steady user | Coding assistant | 7 | Active | Active |
| A month-end user (Finance) | Chat assistant A | 35 | Idle | Active |
| A sale-season user (Retail) | Chat assistant B | 189 | Idle | Active |
1 hour saved a week, times 33 active users, times 4.33 weeks a month, times $65 an hour, gives about $9,288 a month. The seats cost $1,520. What is left is $7,768.
The value of time saved goes up in a straight line as each user saves more hours. It reaches the seat cost of $1,520 a month at 0.164 hours (about 10 minutes) a week. Below that, the tool does not pay for itself. My base guess of 1 hour is worth $9,288 a month.
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (0.1 hours) | Cut | Cut | Cut |
| Base hours (1 hour) | Consolidate | Consolidate | Consolidate |
| High hours (2.5 hours) | Consolidate | Consolidate | Consolidate |
The answer for Chat assistant A is not the same in every box, so we are not very sure of it.
Idle seats are certain waste: 13 to 14 seats, $480 to $520 a month. Consolidation savings are a choice: 0 to 9 seats, $0 to $315 a month. The two are never added together.
Open a tool to read the recommendation, the money, and what would change the answer.
Consolidate (keep one of two tools)
The same people use two tools for the same work.
How sure we are: Low
38 seats:
| Case | Value of time saved | Seat cost | What is left |
|---|---|---|---|
| Low | $643 | $1,520 | -$877 |
| Base | $9,288 | $1,520 | $7,768 |
| High | $32,150 | $1,520 | $30,630 |
At my base hourly cost, the answer changes to Cut if each active user saves less than 0.164 hours (about 10 minutes) a week.
| Case | Hours I guess | Hourly cost | Hours needed to break even, a week |
|---|---|---|---|
| Low | 0.1 | $45 an hour | 0.236 hours (about 14 minutes) |
| Base | 1 | $65 an hour | 0.164 hours (about 10 minutes) |
| High | 2.5 | $90 an hour | 0.118 hours (about 7 minutes) |
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (0.1 hours) | Cut | Cut | Cut |
| Base hours (1 hour) | Consolidate | Consolidate | Consolidate |
| High hours (2.5 hours) | Consolidate | Consolidate | Consolidate |
Consolidate (keep one of two tools)
The same people use two tools for the same work.
How sure we are: Low
23 seats:
| Case | Value of time saved | Seat cost | What is left |
|---|---|---|---|
| Low | $390 | $805 | -$415 |
| Base | $5,629 | $805 | $4,824 |
| High | $19,485 | $805 | $18,680 |
At my base hourly cost, the answer changes to Cut if each active user saves less than 0.143 hours (about 9 minutes) a week.
| Case | Hours I guess | Hourly cost | Hours needed to break even, a week |
|---|---|---|---|
| Low | 0.1 | $45 an hour | 0.207 hours (about 12 minutes) |
| Base | 1 | $65 an hour | 0.143 hours (about 9 minutes) |
| High | 2.5 | $90 an hour | 0.103 hours (about 6 minutes) |
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (0.1 hours) | Cut | Cut | Cut |
| Base hours (1 hour) | Consolidate | Consolidate | Consolidate |
| High hours (2.5 hours) | Consolidate | Consolidate | Consolidate |
Keep
None of the other four answers applies. It pays for itself at my base guesses and people keep what it makes. It overlaps no other tool, and fewer than one seat in ten is idle.
How sure we are: High
6 seats:
| Case | Value of time saved | Seat cost | What is left |
|---|---|---|---|
| Low | $1,169 | $360 | $809 |
| Base | $5,066 | $360 | $4,706 |
| High | $14,029 | $360 | $13,669 |
Across all my low, base and high guesses, the answer stays the same.
| Case | Hours I guess | Hourly cost | Hours needed to break even, a week |
|---|---|---|---|
| Low | 1 | $45 an hour | 0.308 hours (about 18 minutes) |
| Base | 3 | $65 an hour | 0.213 hours (about 13 minutes) |
| High | 6 | $90 an hour | 0.154 hours (about 9 minutes) |
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (1 hour) | Keep | Keep | Keep |
| Base hours (3 hours) | Keep | Keep | Keep |
| High hours (6 hours) | Keep | Keep | Keep |
Right-size (cancel unused seats)
Many seats are idle, never used, or belong to people who left.
How sure we are: Low
10 seats:
| Case | Value of time saved | Seat cost | What is left |
|---|---|---|---|
| Low | $136 | $300 | -$164 |
| Base | $985 | $300 | $685 |
| High | $4,092 | $300 | $3,792 |
At my base hourly cost, the answer changes to Cut if each active user saves less than 0.152 hours (about 9 minutes) a week.
| Case | Hours I guess | Hourly cost | Hours needed to break even, a week |
|---|---|---|---|
| Low | 0.1 | $45 an hour | 0.22 hours (about 13 minutes) |
| Base | 0.5 | $65 an hour | 0.152 hours (about 9 minutes) |
| High | 1.5 | $90 an hour | 0.11 hours (about 7 minutes) |
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (0.1 hours) | Cut | Cut | Cut |
| Base hours (0.5 hours) | Right-size | Right-size | Right-size |
| High hours (1.5 hours) | Right-size | Right-size | Right-size |
Investigate (check before deciding)
People rarely keep what the tool makes, so the time saved cannot be trusted.
How sure we are that it needs a closer look: High
26 seats:
| Case | Value of time saved | Seat cost | What is left |
|---|---|---|---|
| Low | $448 | $650 | -$202 |
| Base | $3,237 | $650 | $2,587 |
| High | $13,445 | $650 | $12,795 |
Across all my low, base and high guesses, the answer stays the same.
| Case | Hours I guess | Hourly cost | Hours needed to break even, a week |
|---|---|---|---|
| Low | 0.1 | $45 an hour | 0.145 hours (about 9 minutes) |
| Base | 0.5 | $65 an hour | 0.1 hours (about 6 minutes) |
| High | 1.5 | $90 an hour | 0.073 hours (about 4 minutes) |
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (0.1 hours) | Investigate | Investigate | Investigate |
| Base hours (0.5 hours) | Investigate | Investigate | Investigate |
| High hours (1.5 hours) | Investigate | Investigate | Investigate |
Right-size (cancel unused seats)
Many seats are idle, never used, or belong to people who left.
How sure we are: High
13 seats:
| Case | Value of time saved | Seat cost | What is left |
|---|---|---|---|
| Low | $877 | $650 | $227 |
| Base | $5,066 | $650 | $4,416 |
| High | $17,537 | $650 | $16,887 |
Across all my low, base and high guesses, the answer stays the same.
| Case | Hours I guess | Hourly cost | Hours needed to break even, a week |
|---|---|---|---|
| Low | 0.5 | $45 an hour | 0.371 hours (about 22 minutes) |
| Base | 2 | $65 an hour | 0.257 hours (about 15 minutes) |
| High | 5 | $90 an hour | 0.185 hours (about 11 minutes) |
| Hours saved | Low hourly cost ($45) | Base hourly cost ($65) | High hourly cost ($90) |
|---|---|---|---|
| Low hours (0.5 hours) | Right-size | Right-size | Right-size |
| Base hours (2 hours) | Right-size | Right-size | Right-size |
| High hours (5 hours) | Right-size | Right-size | Right-size |
This is the full list, one row for each paid seat.
Showing 120 of 120 seats. Problems come first, then the longest quiet.
| Seat | Tool | Department | State | Flag | Days since last use | Why |
|---|---|---|---|---|---|---|
| SEAT-IMG-P004 | Image generator | Creative | Left the company, seat still paid | none | 210 | The owner left on 2025-12-01. The seat is still assigned. |
| SEAT-SUP-P041 | Support-desk AI add-on | Customer Support | Left the company, seat still paid | none | 98 | The owner left on 2026-03-26. The seat is still assigned. |
| SEAT-CA-P057 | Chat assistant A | Operations | Left the company, seat still paid | none | 14 | The owner left on 2026-06-17. The seat is still assigned. |
| SEAT-IMG-P015 | Image generator | Marketing | Idle | none | 343 | No use for 343 days, and nothing explains the gap. |
| SEAT-CB-P060 | Chat assistant B | Operations | Idle | none | 336 | No use for 336 days, and nothing explains the gap. |
| SEAT-SUP-P045 | Support-desk AI add-on | Customer Support | Idle | none | 259 | No use for 259 days. Its regular pattern of bursts has stopped. |
| SEAT-CB-P001 | Chat assistant B | Creative | Idle | none | 252 | No use for 252 days, and nothing explains the gap. |
| SEAT-CA-P025 | Chat assistant A | Technology | Idle | none | 231 | No use for 231 days, and nothing explains the gap. |
| SEAT-IMG-P005 | Image generator | Creative | Idle | none | 168 | No use for 168 days, and nothing explains the gap. |
| SEAT-TRN-P062 | Meeting transcriber | Operations | Idle | none | 98 | No use for 98 days, and nothing explains the gap. |
| SEAT-CA-P061 | Chat assistant A | Operations | Idle | none | 91 | No use for 91 days, and nothing explains the gap. |
| SEAT-CA-P037 | Chat assistant A | Customer Support | Idle | none | 35 | No use for 35 days, and nothing explains the gap. |
| SEAT-SUP-P040 | Support-desk AI add-on | Customer Support | Never used | none | never | Never used in the whole year. |
| SEAT-TRN-P072 | Meeting transcriber | People | Never used | none | never | Never used in the whole year. |
| SEAT-CA-P020 | Chat assistant A | Technology | New starter | none | never | The owner started in the last 30 days, so low use is expected. |
| SEAT-CB-P013 | Chat assistant B | Marketing | New starter | none | never | The owner started in the last 30 days, so low use is expected. |
| SEAT-SUP-P042 | Support-desk AI add-on | Customer Support | New starter | none | 14 | The owner started in the last 30 days, so low use is expected. |
| SEAT-TRN-P063 | Meeting transcriber | Operations | New starter | none | 7 | The owner started in the last 30 days, so low use is expected. |
| SEAT-CA-P055 | Chat assistant A | Operations | Light use | none | 28 | Last used 28 days ago, in only a few recent weeks. |
| SEAT-TRN-P047 | Meeting transcriber | Customer Support | Light use | none | 28 | Last used 28 days ago, in only a few recent weeks. |
| SEAT-TRN-P019 | Meeting transcriber | Technology | Light use | none | 21 | Last used 21 days ago, in only a few recent weeks. |
| SEAT-TRN-P027 | Meeting transcriber | Technology | Light use | none | 21 | Last used 21 days ago, in only a few recent weeks. |
| SEAT-CA-P070 | Chat assistant A | People | Light use | none | 14 | Last used 14 days ago, in only a few recent weeks. |
| SEAT-TRN-P067 | Meeting transcriber | Operations | Light use | none | 14 | Last used 14 days ago, in only a few recent weeks. |
| SEAT-CA-P007 | Chat assistant A | Creative | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-CA-P026 | Chat assistant A | Technology | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-CA-P028 | Chat assistant A | Technology | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-IMG-P001 | Image generator | Creative | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-TRN-P028 | Meeting transcriber | Technology | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-TRN-P046 | Meeting transcriber | Customer Support | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-TRN-P058 | Meeting transcriber | Operations | Light use | none | 7 | Last used 7 days ago, in only a few recent weeks. |
| SEAT-CB-P075 | Chat assistant B | Retail | Active | none | 196 | Used in bursts during a known busy season. Quiet now, as expected. |
| SEAT-CB-P074 | Chat assistant B | Retail | Active | none | 189 | Used in bursts during a known busy season. Quiet now, as expected. |
| SEAT-CB-P077 | Chat assistant B | Retail | Active | none | 189 | Used in bursts during a known busy season. Quiet now, as expected. |
| SEAT-CB-P078 | Chat assistant B | Retail | Active | none | 189 | Used in bursts during a known busy season. Quiet now, as expected. |
| SEAT-CB-P080 | Chat assistant B | Retail | Active | none | 189 | Used in bursts during a known busy season. Quiet now, as expected. |
| SEAT-IMG-P080 | Image generator | Retail | Active | none | 189 | Used in bursts during a known busy season. Quiet now, as expected. |
| SEAT-CA-P031 | Chat assistant A | Finance | Active | none | 35 | Used once a month at month-end, as expected for this team. |
| SEAT-CA-P032 | Chat assistant A | Finance | Active | none | 35 | Used once a month at month-end, as expected for this team. |
| SEAT-CA-P034 | Chat assistant A | Finance | Active | none | 35 | Used once a month at month-end, as expected for this team. |
| SEAT-CA-P035 | Chat assistant A | Finance | Active | none | 35 | Used once a month at month-end, as expected for this team. |
| SEAT-CA-P019 | Chat assistant A | Technology | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-CA-P059 | Chat assistant A | Operations | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-CB-P051 | Chat assistant B | Operations | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-COD-P024 | Coding assistant | Technology | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-COD-P030 | Coding assistant | Technology | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-IMG-P003 | Image generator | Creative | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-IMG-P007 | Image generator | Creative | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-IMG-P014 | Image generator | Marketing | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-TRN-P040 | Meeting transcriber | Customer Support | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-TRN-P050 | Meeting transcriber | Customer Support | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-TRN-P064 | Meeting transcriber | Operations | Active | none | 14 | Last used 14 days ago, and used in most recent weeks. |
| SEAT-CA-P002 | Chat assistant A | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P005 | Chat assistant A | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P009 | Chat assistant A | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P014 | Chat assistant A | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P017 | Chat assistant A | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P027 | Chat assistant A | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P029 | Chat assistant A | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P038 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P039 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P040 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P044 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P045 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P049 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P050 | Chat assistant A | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P052 | Chat assistant A | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P056 | Chat assistant A | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P062 | Chat assistant A | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P064 | Chat assistant A | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P065 | Chat assistant A | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P066 | Chat assistant A | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P069 | Chat assistant A | People | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CA-P071 | Chat assistant A | People | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P002 | Chat assistant B | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P005 | Chat assistant B | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P009 | Chat assistant B | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P014 | Chat assistant B | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P017 | Chat assistant B | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P038 | Chat assistant B | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P039 | Chat assistant B | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P044 | Chat assistant B | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P056 | Chat assistant B | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P062 | Chat assistant B | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P064 | Chat assistant B | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P066 | Chat assistant B | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P069 | Chat assistant B | People | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P072 | Chat assistant B | People | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-COD-P019 | Coding assistant | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-COD-P022 | Coding assistant | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-COD-P025 | Coding assistant | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-COD-P026 | Coding assistant | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-IMG-P008 | Image generator | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-IMG-P012 | Image generator | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P037 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P038 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P039 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P043 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P044 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P046 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P047 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P049 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-SUP-P050 | Support-desk AI add-on | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P007 | Meeting transcriber | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P008 | Meeting transcriber | Creative | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P015 | Meeting transcriber | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P017 | Meeting transcriber | Marketing | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P022 | Meeting transcriber | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P029 | Meeting transcriber | Technology | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P036 | Meeting transcriber | Finance | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P038 | Meeting transcriber | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P044 | Meeting transcriber | Customer Support | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P055 | Meeting transcriber | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P066 | Meeting transcriber | Operations | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P069 | Meeting transcriber | People | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-TRN-P070 | Meeting transcriber | People | Active | none | 7 | Last used 7 days ago, and used in most recent weeks. |
| SEAT-CB-P011 | Chat assistant B | Marketing | Closed | none | 301 | The seat was already cancelled. |
| SEAT-TRN-P054 | Meeting transcriber | Operations | Closed | none | 231 | The seat was already cancelled. |
| SEAT-SUP-P048 | Support-desk AI add-on | Customer Support | Closed | none | 147 | The seat was already cancelled. |
| SEAT-CA-P021 | Chat assistant A | Technology | Closed | none | 70 | The seat was already cancelled. |
The first version found every problem I planted. That was expected. I wrote both the data and the checker, and I set the limits with the data in front of me. It tests the code. It does not test the world.
All the hard cases were suggested by Claude in chat and approved by me. I ran the checker on them first without changing it. Then I logged each miss. Then I fixed each one on its own.
Some fixes are real improvements. Others only fit these cases, and the limits they use are my judgement. To see these cases in the seat table, pick a hard-case data set under The answers.
The tables below show what the checker found and missed, before and after each round of fixes. The full log of what failed, what changed and why is in the technical log, word for word.
| Planted problem | Planted | Found | Precision | Recall |
|---|---|---|---|---|
| Idle seats | 9 | 9 | 1 | 1 |
| Never-used seats | 2 | 2 | 1 | 1 |
| Seats of people who left | 3 | 3 | 1 | 1 |
| New starters | 4 | 4 | 1 | 1 |
| People using two tools for the same work | 9 | 9 | 1 | 1 |
| The pair of overlapping tools | 1 | 1 | 1 | 1 |
| The tool whose output is rarely kept | 1 | 1 | 1 | 1 |
| The tool with heavy, well-kept use | 1 | 1 | 1 | 1 |
| Case | Before any fix | After the fixes |
|---|---|---|
| Case 1: Parental leave | 1 of 4 | 4 of 4 |
| Case 2: Shared login | 3 of 4 | 4 of 4 |
| Case 3: Quarterly user in Marketing | 2 of 4 | 4 of 4 |
| Case 4: Department transfer | 3 of 5 | 5 of 5 |
| Case 5: Missing usage records | 0 of 3 | 3 of 3 |
| Controls: Controls | 8 of 8 | 8 of 8 |
| All checks | 17 of 28 | 28 of 28 |
| Case | Before any fix | After the fixes |
|---|---|---|
| Case 6: Leave plus a quarterly rhythm | 1 of 5 | 5 of 5 |
| Case 7: Transfer plus a business cycle | 2 of 4 | 4 of 4 |
| Case 8: Leave ended, no return to use | 2 of 4 | 4 of 4 |
| Controls: Controls | 3 of 3 | 3 of 3 |
| All checks | 8 of 16 | 16 of 16 |