TwinLadder
Score your organisation

The judgment register, as an app

A subscription web app that keeps the register from Judgment Capital: six modes, three tiers, the drills, and a verified board reading each quarter.

Runs in your browser. Nothing to install.

Score your organisation

Free, the ten-question ladder from chapter 12. See plans

Two lines, two owners

Succession belongs to the CHRO. Decision quality belongs to the CRO. The chief executive answers for both.

A register is the ordinary way a board watches a serious risk: an owner, tiers of severity, a quarterly clock, and conditions set in advance for escalating each entry and bringing it back down. The app gives judgment the same form, so your board reads it the way it reads credit or cyber.

A woman with a microphone stands at a whiteboard and speaks to colleagues seated in front of her
The succession line: the CHRO.
A grey-haired woman in glasses reads a document on a clipboard, pen in hand, at a table by a window
The decision-quality line: the CRO.
“A board should not have to learn a new instrument to watch a new risk.”
Judgment Capital, chapter 10

Six modes, three tiers

Every reading lands in a tier, and each tier tells the owner what to do next.

Watch
The reading is moving the wrong way but sits inside the range the owner accepts. Most entries live here, most quarters. It is where the risk is meant to be seen early.
Concern
A threshold set in advance has been crossed, or two readings on the same line move together. The owner brings the board “a plan, with its own clock.” (Chapter 10)
Live loss event
The failure the register predicted has happened. The book calls it “a bill that has arrived.” An owner or the verifier declares it; the app never computes it.

The app proposes each tier. The owner and the board decide.

Each entry’s breach response is agreed by the board in advance; the book’s defaults are to pause further delegation in the affected process and to sample the unseen volume harder.

When a threshold is crossed or a rung is lost, the app opens the event, names the owner and starts the clock. The pause is your decision, made in your systems; the app records it and the owner’s confirmation.

What every entry carries

Five fields, as Appendix B sets them out.

  • Owner

    A named executive.

  • Threshold

    The reading at which the entry moves up a tier, set in advance.

  • Escalation

    What happens when it does.

  • The condition for coming down

    The evidence the owner must show, held for four consecutive quarters before the board will move the entry down. The practice has to keep running; “Nothing on this sheet is ever cured.” (Chapter 11)

  • Review cycle

    Read at every meeting of the owning committee, and quarterly at the board.

Entries are escalated or downgraded, never closed. Floors and pass rates are set before the first window opens and cannot be back-dated.

Drills, run from the app

The tests the book prescribes, scheduled, logged and reported by desk.

Seeded cases

A cell outside the desk takes known-bad cases from your own files. It can plant them in the live queue of your own systems and log each one in the app, or run them as a drill in the app, mixed with clean cases, for the desk to flag or clear. The app computes the detection rate, the miss rate and the time to detection, by desk, never by person.

Drill readings are marked as taken under exercise conditions, and a desk with fewer than twenty seeded cases in the quarter gets a reading marked indicative. A desk of fewer than five people reports through its function. Each reviewer can see their own result, privately; the views your organisation opens show desks only.

Missed cases become teaching cases for the whole desk. (Appendix A, A.4.)

Defect review

Each month, ten failures read by eye against outcomes. Missed cases become teaching cases and new seeds. (A.2.)

Blind re-judgment

A share of decisions, for example one in eight, re-judged without seeing the machine’s answer. (Chapter 11.)

Tools-down rebuild

At least once a year, a defined set of real tasks done with the tools off, scored and kept. The app labels it as the book does: it detects, it does not cure. (A.5.)

The board reading

Each quarter ends in one reading the board can table: the register, the rung, and the answers in writing.

The reading opens on the six modes and their tiers, the ladder score with every zero named, and what changed since last quarter. It carries the five questions for Monday with the written answers:

  1. Who overrode the machine?
  2. Who decided what to protect?
  3. Tools down, who catches it?
  4. What did nobody see?
  5. Where is it written?

It is your first task as a subscriber, before any method has data: the motion from chapter 13. Management answers the five questions in writing within sixty days, a named person verifies the answers, and they return to the committee as an agenda item.

Board members read it in the app through a private link, free, or as a PDF of about four pages. Once verified, a reading is locked; corrections go into the next one.

Board reading · Q1 2027Sample data
Six modes2 at Concern · 4 at Watch · 0 live loss events
  • Pipeline collapse · CHROConcern
  • Competence drift · CROWatch
  • Inquiry groupthink · CROWatch
  • Detection blindness · CRO, CHROConcern
  • Authorship erosion · CROWatch
  • Agency erosion · CROWatch
Decisions asked of the board2
  • Approve the claims-desk plan, re-read 15 Apr 2027Concern
  • Hold the junior intake floor at 6 for 2027Concern

13 of 20 · no zeros · Controlled needs questions 3–5 at 2

  1. The line and the name2
  2. The signals on demand2
  3. The marked map1
  4. The threshold that fired1
  5. The mark that survived money1
  6. The work without the tool1
  7. The stage that cannot be skipped1
  8. The pair in the calendars2
  9. The override that taught1
  10. The rebuild, tools down1
  1. Who overrode the machine?36 overrides on the claims desk in Q4; none in payments, recorded as a question to answer.CRO · 12 Jan · verified
  2. Who decided what to protect?Process owners signed the map for 5 of 12 processes; the rest are due by 31 Mar.CHRO · 12 Jan · verified
  3. Tools down, who catches it?The claims desk completed 6 of 10 tasks unaided in November.CRO · 14 Jan · verified
  4. What did nobody see?9 of 40 seeded cases passed the claims desk; all 9 are now teaching cases.CRO · 15 Jan · verified
  5. Where is it written?In this register, the committee minutes of 21 Jan, and the evidence appendix.Company secretary · 15 Jan · verified
Since Q4 20264 changes
  • Detection blindness, claims deskWatch→Concern
  • Pipeline collapse, junior intake 4 against floor 6Watch→Concern
  • Ladder score 11 → 13; the zero on question 10 cleared by the November rebuildAware
  • Pairing logged both ways in every team, two quarters runningQuestion 8 at 2
Verified by Head of Internal Audit (sample) · 18 Jan 2027 · lockedPDF, 4 pages + appendix

Your verifier signs it

Each reading is verified by a competent person you name, independent of the officers whose lines are read.

By default that is your internal audit function. It can be an outside person. The app will not accept a verifier who owns a line in the reading, and the verifier signs a declaration of independence. The verifier opens the evidence behind the numbers, raises queries, and signs the reading or declines it. A declined reading still goes to the board, marked as declined.

If you want TwinLadder to verify, it is a paid add-on, from €490 a reading. TwinLadder never verifies a register that its own workshop set up in the same year.

“The self-score locates you; it does not certify you.”
Judgment Capital, chapter 12
A man in glasses signs a document at a table while the person opposite holds a printed contract
Verify the Q1 2027 readingSample data
Verifier
Head of Internal Audit (sample)
Evidence sampled
24 items across 6 entries
Queries raised
3 · all answered by the owners

I own no line in this reading and have no part in the work it reads. Declared 15 Jan 2027.

Your people enter what happened

Override counts, pairing hours, intake and drill results go in by form or spreadsheet upload.

Most of these numbers already exist somewhere: in the case system, the HR system, a team lead’s spreadsheet. The app asks each named person for them on a schedule, records who entered each result and when, computes the readings, and tells the owner when something moves.

It works beside your AI systems and does not connect to them, which is why every number has evidence attached and a verifier who samples it.

The calendar, Q1 2027

Each instrument with its date, who enters it, and where it stands. Sample data.

January 2027

6
A.2 Defect review, DecemberClaims · ten failures read by eye
Head of ClaimsEntered 6 Jan
8
Junior intake, Q4All functions · against the floor of 6
HR business partnerUploaded 8 Jan
10
A.6 Overrides exercised, Q4Payments · 0 entered, shown as a question
Payments operations leadEntered 9 Jan
12
A.1 Pairing hours, DecemberCustomer operations · both directions
Customer operations leadLate 4 days · CHRO told
13
A.4 Seeded cases, resultsClaims · 40 planted by the quality cell
Quality cellEntered 13 Jan
18
Verification of the readingThe register, the ladder, the five answers
Head of Internal AuditSigned 18 Jan

February 2027

3
A.2 Defect review, JanuaryClaims · reminder sent 27 Jan
Head of ClaimsDue in 7 days
3
A.1 Pairing hours, JanuaryUnderwriting · spreadsheet upload
Underwriting team leadDue in 7 days
4
Plan for the claims deskDetection blindness · Concern, with its clock
CRODue in 8 days
11
A.3 Divergence methodBoard decision on the 2027 automation case
Company secretaryScheduled
24
Blind re-judgment, 1 in 8Claims · sample drawn outside the desk
Quality cellScheduled

March 2027

3
A.2 Defect review, FebruaryClaims · ten failures read by eye
Head of ClaimsScheduled
9
A.6 Halt-point drillPayments · one named halt point, tested
Payments operations leadScheduled
15
A.4 Seeded cases, Q2 drill opensClaims · 80 seeds, doubled under the plan
Quality cellScheduled
31
Protect and hand-over map7 processes still to sign
Process ownersScheduled
31
Ladder re-score, ten questionsEvidence attached to each answer
CRO and CHROScheduled
A board meeting by tall windows: a grey-haired director speaks while the others at the table listen
The board reads the quarter; each officer reads their line.

Who reads what

  • Board directors

    Opens to the reading

    “What is our adoption programme doing to our ability to judge what we have adopted?” You read the quarterly reading and the rung.

  • Risk officers

    Opens to the decision-quality line

    You own the decision-quality line: drift, groupthink, detection, authorship, agency. You set the thresholds and present the reading.

  • Compliance officers

    Opens to the record

    You keep the record: what was measured, who verified it, when. The readiness check sits beside it for the AI Act floor.

  • HR

    Opens to the succession line

    You own the succession line: intake, pairing, the bench a decade out, and the formation half of detection blindness.

  • Legal

    Opens to halt points and answers

    You read the halt points, the override channel and the written answers, dated and verified.

Where it runs

In your browser. Nothing to install.

Each person sees what their role allows: the board, line owners, process owners, the verifier, the desk. Board seats are free.

Drill results are reported by desk, never by person. A reviewer sees their own result and nobody else does. Desks of fewer than five people report through their function. Personal data is limited to name, work email and role, plus the two logs where the book asks for a name: pairing and overrides.

You are the controller of your data and we are the processor, under a data processing agreement you accept at checkout.

Questions

No. It is a register for the people who check your AI, and it works beside whatever systems you run. Your people enter what happened. The book’s rule applies: “We specify the instrument, not the vendor.”

No. When a threshold is breached, it opens the response your board agreed in advance and records what the owner did.

No. Drill results are reported by desk. A reviewer can see their own result, privately; managers, owners and the board see desks only. The book is firm: “A monthly handful reads the desk, not the person.”

Small is self-serve: a checklist takes you from your first processes to named owners and a verifier. On Standard and above, a workshop sets it up with you, working in your own app.

It helps. Every screen links to the chapter it comes from.

The readiness check covers the Article 4 floor on seven pillars, and a result can be attached to your first reading. The first check is free, and every plan includes it after that.

Yes, as a PDF of about four pages with an appendix.

Start with ten questions

The ladder from chapter 12, free. You see your rung, your zeros, and the five questions to take to your next board meeting.

Ready for the app? See plans.

The ladder self-score
You answer from

Free. No account needed. Your result shows on screen. Leave an email only to save it, get the one-page PDF and a reminder to re-score in 13 weeks.