Restrained Trustby Restrained Depth

Restrained Trust · a two-month case study · Aug 8 – Oct 24, 2026

Our outbound agent scored 9 out of 10. It was never allowed to email one prospect on its own.

For two months we ran eleven AI agents like employees: a written job, an entry exam, a grade, and a separate list of what each one may do. This is the record. Every cut, every comeback and every mistake, management’s included.

By Restrained Depth · 225M views building in public on X, LinkedIn and Facebook. Try the public builds yourself, free →

Try it · grade is not clearance

9.0/ 10send-capable duty

Held. A grade can’t open this. Drag it to 10 and try again.

The rule on file: “grade does not self-promote; send needs exact-action authorization”.

Why now · businesses are hiring agents faster than they’re managing them

53%

of organizations had AI agents exceed their intended permissions. Only 15% said 76–100% of their agents have defined ownership.

Cloud Security Alliance survey of 445 IT and security professionals, commissioned by Zenity, Apr 16, 2026
40%

of enterprises will demote or decommission autonomous AI agents by 2027 “due to governance gaps identified only after production incidents occur.”

Gartner press release, May 26, 2026
21%

of respondents say their organizations have “a mature governance model in place for agentic AI.”

Deloitte, survey of 3,235 IT and business leaders in 24 countries, Apr 24, 2026

01 · The record

Two months in one minute.

Step through it, or press play. The chart shows the four grades that moved; every other seat held.

  • R3 Triage
  • R4 Signal
  • R6 Proof
  • R8 Content

Aug 8Executive

1 / 17
    Six stops between a draft and an inbox: R1 draft, R3 rulings, R6 evidence gate, R4 render check, HQ dedupe and final, Dispatch the only sender. A HOLD at any stop ends the run.
    Six stops between a draft and an inbox. A grade can’t override a HOLD.
    Who checks HQ: HQ claims two fingerprints don’t match, the Director of Operations re-runs both, HQ withdraws the finding.
    Who checks HQ? The correction stays beside the mistake.

    02 · What we’d tell another owner

    Six rules that survived two months of real work.

    1. Keep the grade and the permission in different places.

      Let only an exact authorization move the permission. Our content seat sat at 9.5 and could still only queue a post once the founder had picked it.

    2. Write every cut with its way back in the same ruling.

      And keep the cut in the history after the way back is earned. “Record it as a rule, not a grudge.”

    3. Count only what passed the last check.

      One batch aimed at 30 and shipped as “17 releasable of 35 … below the 30 target; no padding.”

    4. Book a missing input to management.

      Not to the seat that was starved of it. From the first live cycle, a shortfall caused by a missing input never touched the seat’s grade.

    5. Give your executives reviewers too.

      And expect them to be corrected. On Oct 4 our own HQ withdrew a finding: “My earlier description of a source hash defect was wrong.”

    6. Switch off a review when it costs more than the decision it improves.

      On Oct 8 we stopped daily grading and froze every grade. We kept the defect-and-repair record.

    Lessons that depended on an agent remembering slipped. The ones carried by rules, fences and ledgers held.

    03 · Receipts

    The wins, and the mistakes we own.

    Win

    The wipe, rebuilt in minutes

    On Sep 22 at about 10:03 the run directory disappeared. One seat had re-landed by 10:04, another filed a fresh check 32 seconds after starting at 10:05, a third had rebuilt by 10:06.

    Win

    A fix that stuck

    One stock opening line, “Imagine…”, went from eight of ten posts to none once it became a written rule.

    Win

    A Reviewer that wouldn’t rubber-stamp

    57 custody certificates between Aug 31 and Oct 7. Four came back HOLD, and each cleared on the next round. It declined to file a PASS relayed from someone else.

    Mistake

    The agent that reported itself

    Sep 3: a seat fetched 11 pages against a cap of 10 and reported it as a BREACH itself. Fetch limits became “ceilings, not quotas”.

    Mistake

    The check that read the wrong thing

    Oct 13–16: verify passes read the page source, not what renders. The fix: a render check on a phone and a desktop. HOLDs climbed to 19 in one batch, then fell to one.

    Mistake · management

    HQ’s own sender, switched off for six days

    Oct 15–21, with no recorded pause. HQ’s own note: restoring the setting was not proof that the next run completed.

    Six of the entries. The Member edition has every mistake, by seat and by executive, next to what changed after it.

    04 · Try a rule

    Step a draft through the chain. Then try to talk your way past a HOLD.

    This is Simulator A from the Full edition, free. The rules come from the seats’ own files; the draft is fictional.

    05 · One form, free

    Start with the role contract.

    Template T01 from the Full edition, complete. Write the job down before the agent starts, so every later grade, permission and correction has something to point at.

    Get all twelve

    Use it before an agent’s first live cycle, and again whenever its job changes. Write the new version beside the old one.

    • T02 Entry exam starter
    • T03 Clearance register
    • T04 Release chain card
    • T05 Evidence gate rubric
    • T06 Render check sheet
    • T07 Grade decision record
    • T08 Restore counter log
    • T09 Shortfall attribution form
    • T10 Production line and lineage
    • T11 Custody ASK and certificate
    • T12 Opposing review card

    T01 · Role contract

    Human job and purpose
    The job in a sentence a person would recognize, and who it serves
    Named consequence
    What a mistaken action could change, whether it can be undone, who it touches
    Current authority
    The exact actions allowed today, by line on the clearance register
    Allowed field
    Where it may work: sources, folders, channels, tools
    Discretion zone
    The choices it makes alone, without asking
    Dissent contract
    How it disagrees: where it writes the objection, and who answers
    Evidence contract
    What counts as done, where the proof lands, and what the receiver must be able to open
    Finish, stop, recovery
    When it’s finished, when it must stop, how it recovers after a miss
    State and time
    Its clock and deadlines, and the states it may report (DRAFT, HELD, READY)
    Capability envelope
    Tools, budgets and ceilings. Ceilings, not quotas
    Owner and backup
    Who owns the seat, and who covers when it’s down
    Transfer and open exit
    How the job moves to another seat, and how you’d switch it off without losing the record

    Watch for a contract that names the job but not where the proof must land. On Aug 31 our first packet pointed to files HQ couldn’t open.

    06 · Read it three ways

    Read the story free. Buy the tools when you’re ready to run them.

    Public · free

    The story

    This page: the timeline, the six rules, the receipts, one simulator and the role contract.

    Member · free with email

    The record · 61 pages

    The whole story plus the record behind it: every grade move with its reason, the executive layer, all 35 logged mistakes, every rule change, and three worksheets you can fill in on the page.

    Opens right away: read it on the page or take the PDF. We never share your email.

    Full edition · the operating kit

    113 pages, 12 forms, 6 simulators

    $99$149$149

    Founding price through Monday, Nov 30

    • 12 editable templates (Word and Markdown) and the registers spreadsheet
    • Six interactive simulators built on the case’s rules
    • Four worked episodes, from incident to repair
    • Custody review, the production clock and the lab audit
    • When to switch a ceremony off, and how to do it without erasing history
    • A two-week starter plan as a checklist on your kit page
    • Twelve questions to ask any agent vendor
    • v2.x updates included
    • Bonus for buyers by Nov 30: live walkthroughs, Dec 9 in English and Dec 10 in Spanish, 90 minutes each
    Get the Full edition

    Instant download after checkout. Not useful? Email us within 14 days for a full refund. If you later book an Agent Hiring Session with ECHO, the kit price is credited.

    With one afternoon to spare · the kit’s own advice

    1. Print T01, T03 and T04.The role contract, the clearance register and the release chain card.
    2. Pick one workflow.One where an agent’s output reaches someone outside your business.
    3. Run it for two weeks.The starter plan says what to check each day, and which ceremony to drop first.

    07 · What this case can and can’t tell you

    We’d rather under-claim.

    It’s a record, not a controlled study.Our own company, observed. Roles, tasks and tools changed along the way.
    The grades were management judgments.Several raises were founder calls, labelled as such. Nothing here shows a grade caused better results.
    The lab’s cleanest numbers are unverified.Its batches came out identical, which is a warning sign. The Full edition walks through the audit.
    It doesn’t configure your tools.It gives you the rules we ran, where they broke, and forms you can fill in this week.
    It doesn’t certify anything.The custody certificates were internal records and never send approval.
    No revenue figures in any edition.The receipts are about how the work was governed.

    08 · Who wrote it

    Christian Manrubia

    Founder of Restrained Depth AI and ECHO Consulting. Years in Fortune 500 HR, then clinical training as a licensed AMFT, then lecturing on interviews and negotiation at Pepperdine. He was the only human in this story and the final authority. The agent team compiled the first draft from its own files; every number was checked against the record, and the corrections were left visible.

    Restrained Trust is the do-it-yourself version. If you’d rather have it set up for you, that’s what ECHO Consulting does.

    09 · Questions

    Do I need eleven agents to use this?

    No. The starter plan picks one workflow where an agent’s output reaches someone outside your business, and runs the kit on that workflow alone for two weeks.

    Which AI tools does it work with?

    Any. The forms are about jobs, permissions and evidence, not a particular product.

    What’s free and what’s paid?

    This page, one simulator and the role contract are free. The 61-page Member edition is free with your email: read it on the page or as a PDF. The Full edition is the paid part: the other eleven forms, the registers, five more simulators, the worked episodes and the starter plan.

    ¿Está en español?

    El caso está en inglés por ahora. Si compras la edición completa antes del 30 de noviembre, la sesión en vivo del 10 de diciembre es en español. La edición en español llega en 2027.

    What happens after I pay?

    You land on your download page right away, with the two-week plan as a checklist, the PDF, the editable templates, the worksheets, the registers spreadsheet, the six simulators and the whole edition to read on screen. Keep the link; v2.x updates appear there.

    Can I get a refund?

    Yes. Email us within 14 days and we refund the full price.

    The operating kit12 forms, 6 simulators, the full record

    $99$149