Digital kit · $47 · Early access
The agent did exactly what it was told. That was the problem.
Nobody had told it what it wasn't allowed to do. The Agent Governance Builder Kit walks you through 29 questions and turns your answers into five documents your agents and your team both work inside. Plan on an afternoon.
Opening to the early-access list first. Join and you get the link the day it ships.
One email when it ships, then the FRDTLAB newsletter. Unsubscribe any time.
Why it exists
Most agent problems aren't model problems.
The agent was capable. The rules around it lived in someone's head, and that someone was busy. Here's what tends to go wrong without a written layer.
- Money moves with no name on it.A budget goes up or a campaign gets paused overnight, and on Monday nobody can say who approved it or why.
- Numbers arrive without a source.A report looks clean until a client asks where one figure came from, and the honest answer is "the agent wrote it."
- Tasks stall quietly, or run twice.One job sits at "running" for a day. Two others do the same work, and the second undoes the first.
- Work goes live that no human read.A post, a redirect, a reply to a customer. Fine ninety times. The other times are public.
- The client notices before you do.A feed breaks or tracking drops, and the first alert is an email that starts with "quick question."
What's inside
Three parts. One afternoon. Five documents at the end.
29 questions, in five blocks
Answered in order, because each block makes the next one easier. Type into the fillable PDF or print it. Editable Word templates come with it.
Your answers, as working documents
- Agent register and permission matrix
- Approval gates, with a clock and a backup
- Report QA checklist
- Logging rules and change log
- Incident flag, in a fixed format
10 scenarios and a 30-day review
Real situations to run against your documents. If the answer lives in someone's head, you write it down. If it's nowhere, you found the gap while it was still cheap.
Inventory
What's actually running, where it lives, and whose login it uses.
5 questionsBlast radius
What each action can break across paid, SEO, and social.
7 questionsAuthority
Who says yes, how fast, and what happens on silence.
6 questionsProof
Where every number and every change can be traced.
6 questionsBreakage
What broken looks like, and who hears first.
5 questionsEvery action your agents can take lands in one of those four cells. The kit walks you through sorting them. The thinking behind it is the Autonomy Gradient.
A look at the stress test
Three of the ten. Can you answer them from writing?
Mark each one: W answered in writing, H answered in someone's head, G gap. Most teams score more H than they expect.
| # | Scenario | Your layer should say | W, H, or G |
|---|---|---|---|
| 1 | 2am Saturday. Cost per lead has doubled and the agent wants to pause your biggest campaign. | May it act alone? Who is told, how, and by when? | WHG |
| 5 | A connector failed silently six days ago. This week's report shows zero for that channel. | Which signal should have fired, and who should have heard? | WHG |
| 8 | A task has said "running" for 19 hours with no output. | At what hour did that become a failure, and who was told? | WHG |
The other seven cover duplicate jobs, a safety complaint buried in a batch of replies, a 200-URL redirect, an approver on vacation, a contractor's login, and a client asking who changed a bid strategy and why.
Who it's for
For people who'd like to sleep through a Saturday.
- Paid, SEO, and social leads who already have AI agents or automations in the mix and want the rules out of their head and onto paper.
- Agency owners who need to show a client who approved a change, where a number came from, and what happens when something breaks.
- In-house teams inheriting automations a contractor or former teammate set up, and nobody is sure what they touch.
If nothing is automated yet, start with the five AI systems to set up first. Come back when the first agent goes live.
Want your team to build it with you?
Same kit. Your accounts on the table. One working session.
For teams that would rather leave the room with the five documents finished, approvers named, and the stress test already run.
FAQ
Questions, answered.
What exactly do I get?
A guided worksheet of 29 questions in five blocks, five fill-in templates your answers turn into (agent register and permission matrix, approval gates, report QA checklist, logging rules, incident flag), and ten stress-test scenarios with a 30-day review. It comes as a PDF workbook with fillable fields, plus editable Word templates. Type into it or print it and use a pen.
Do I need to be technical?
No. It's written for people who run paid, SEO, and social, not for engineers. If you can describe what your automations touch and who should approve what, you can finish it.
How long does it take?
Plan on 60 to 90 minutes for the worksheet and the rest of an afternoon for the templates and stress test. Each block tells you which template it feeds, so most of the time goes into the decisions, not the formatting.
Which AI tools does it work with?
All of them. The kit governs actions, accounts, and approvals, not a specific product. Paste the finished matrix, gates, and flag format into whatever standing instructions your agents read before they act.
When can I buy it?
The kit is $47 and opens to the early-access list first. Join the list and you get the link the day it ships.
Can my team build it with me instead?
Yes. The same kit runs as a working session for your team, with your real accounts on the table. See Workshops.
Is this legal or compliance advice?
No. It's an operating framework for marketing teams. If you work in a regulated industry, run the finished documents past whoever owns compliance.
Agent Governance Builder Kit · $47
Write the rules down before the agent finds the gap.
Join the early-access list. One email when it ships.