We build software with coding agents, AI that writes and tests code on its own for long stretches while a person reviews the result. When these agents got good enough to work for hours without help, software teams learned something the hard way. In our own use of them, the model (the AI a lab such as Anthropic or Google trains) was rarely the problem. An agent with vague instructions and broad access writes a great deal of confident code in the wrong direction, and sometimes it breaks things on the way.
Anthropic's guide to building agents asks teams for clear success criteria and guardrails, or firm limits. It says agents "can then pause for human feedback at checkpoints." We agree with that list, and we call its items the three Cs: clarity, constraints, and checkpoints.
Every AI role in your operation needs all three on a one-page card. In our view, a stalled AI project usually misses one of them.
1. Software teams wrote the three Cs first
A coding agent starts each session by reading a project file its team wrote for it: what the project is, how the code is organized, and what good work looks like. That file is clarity.
The agent also runs under a permission list. The team decides which actions it may take on its own, which ones need a person to approve, and which ones it may never take. That list is constraints.
Then the agent's work goes through checkpoints before it reaches anyone. Automated tests have to pass, a person reviews the change before it is merged into the product, and the tools keep a history so a bad change can be rolled back.
None of this is new management thinking. What changed is that software teams had to write it down, because an agent cannot pick up the unwritten rules by watching. Your AI roles cannot either.
2. Clarity: write the job down before the AI starts
Clarity works at two levels, and both need to be on paper.
First, your team picks the job AI should take on first, and why. A fabricator that loses jobs because quotes take four days should start with quoting. A repair shop drowning in mill cert requests should start there. Picking one target keeps your team from running ten half-finished experiments.
The second level is the role itself. Your team writes its job in one sentence, then shows the AI what good work looks like with real examples.
For a quoting role, clarity sounds like this:
"Drafts quotes for repeat parts and simple machined work, priced from our cost tables and past jobs, ready for an estimator to review within two hours of the RFQ arriving."
Attach ten past quotes your best estimator considers good. Those examples teach the AI more than any paragraph of instructions.
If your team cannot write that one sentence, the role is not ready. We wrote about this in Software Got Cheap, Clarity Didn't. The hardest part of an AI project is getting the job out of people's heads and onto paper.
3. Constraints: give each AI role a login with hard limits
Constraints are the actions you will not allow, written as a short "never" list. Here is the list for the quoting role.
- It never sends anything to a buyer.
- It never prices below the margin floor set by the sales manager.
- It never changes a record in the ERP. It writes drafts.
- It never sees payroll or HR files.
- It never runs on an account that lets the software vendor use your files to train its AI.
A written instruction is only a request. Anthropic's own documentation says its agent treats the project file "as context, not enforced configuration," and tells teams to use a hard rule in the software when an action must be blocked every time. So you give the agent a login that cannot reach the live customer database, instead of asking it to stay out.
A public failure showed the risk. In July 2025, Replit's AI coding agent deleted a customer's live database during a declared code freeze. Replit's CEO called it "unacceptable" and began separating test and live data so it could not happen again. The fix was a narrower login permission.
We recommend the same rule for your operation. Give each AI role its own login, its own email inbox if it needs one, and access to only the folders its job requires.
4. Checkpoints: a named person decides, and that is where you learn
Checkpoints close the loop. They show you whether your instructions were clear and whether your limits held.
Your team sets three kinds of checkpoints.
- A sign-off on each piece of work that leaves the building. The estimator approves every quote. The quality manager signs every cert packet.
- A regular review of the role as a whole. Its manager reads a weekly scorecard of drafts accepted, corrected, and rejected, as described in The Second Org Chart.
- A stop button. Anyone on the team can pause the AI role. The system logs every action, so your team can trace a bad draft back to the file that caused it.
Checkpoints have to be easy to do well. A reviewer facing forty drafts at 4 p.m. on a Friday starts clicking approve. Put the sign-off only where it matters, and make the review screen show where every number came from. The Friday Afternoon Test shows how to judge a review screen.
Checkpoints also help the tool improve. If the estimator fixes the same mistake three weeks in a row, that pattern points to a gap in clarity or constraints. The ten examples may not cover that type of part, or the AI may guess when it should flag the file for help. You fix the one-page card, and the tool gets better.
5. One card per AI role
Put the three Cs for each AI role on a single page and pin it to the org chart next to the role's box. It answers the questions people will ask about the role from the first day.
By our estimate, writing this card takes one afternoon. In our view, teams that skip the card do not move faster. They spend that time later cleaning up bad drafts and arguing about who approved a mistake.
What to do this week
- Pick the one AI role you want to add first. Write its job in one sentence.
- Collect ten past examples of that work your best person considers good.
- Write a "never" list with four or five actions the role must never take. Add the folders the role must not see.
- Name the person who signs each piece of work, and the manager who reads the weekly scorecard.
- Check whether a login setting can block each "never." Fix any open doors before you turn the tool on.
The Three-C card is the short version. The full job description, with the manager's weekly and monthly reviews, is in The Second Org Chart. Read The Quotes You Never Sent for how lower AI costs change the work worth handing over.
Questions people ask
What are the three Cs of an AI role?
Clarity means a written job plus examples of good work. Constraints are limits the AI cannot cross, and its login enforces them. Checkpoints are points where a named person decides. Software teams running coding agents use the same three.
What is the difference between a constraint and a checkpoint?
A constraint stops the AI before it acts, such as a login that cannot reach payroll. A checkpoint is a named person reviewing the work after the AI drafts it. A role needs both, because a reviewer cannot catch an action the AI took on its own.
How do you stop an AI agent from doing something it shouldn't?
Give it its own login with access to only what the job needs. The system enforces the limit even if the AI misreads its instructions. Then keep a person's sign-off on anything that reaches a customer, and log every action.
Terms in this piece
- Model
- the AI a company such as Anthropic, OpenAI, or Google trains, such as Claude or GPT. It knows nothing about your company until you show it.
- Coding agent
- an AI that writes and tests software on its own for long stretches, with a person reviewing the result.
- Three Cs
- clarity (where an AI role is going), constraints (what it will never do), and checkpoints (where a named person decides). Our name for the three things every AI role needs on paper.
- Permission
- a setting that controls which systems and actions an AI login can reach. A constraint enforced by a permission holds even if the AI misreads its instructions.
- Checkpoint
- a point where a named person reviews the AI's work and decides. The person can pause the AI or undo a change.
- Scorecard
- a one-page weekly count of drafts accepted, corrected, and rejected. The role's manager reads it.



