The Validation Approach
AI drafts. You decide. Every time.
Judgement is one of the three things that never switch off in the SIGNAL loop. In practice, judgement means one habit, repeated at every stage: the AI prepares, you interrogate, verify and decide. This page collects that discipline in one place, because it is easy to read about once in the Foundation chapter and then quietly stop applying it once the outputs start looking polished.
AI drafts, you decide. Never let an assumption become part of your permanent knowledge, your account plan, or your customer conversation without confirming it yourself.
Validate before you save
This is where the discipline starts: building your Foundation. When the Setup Skill reads your website or searches the web, it will occasionally infer something wrong or repeat an outdated claim. The skill is built to force a validation step before anything becomes permanent.
Before writing the final document, the Setup Skill shows you a short summary of its key conclusions and every assumed item, and asks you to confirm or correct each one. Only after you confirm does it produce the single clean “SIGNAL Permanent Knowledge” document, flagging any claim you should double-check before using it with a customer.
The AI reading your website and searching the web saves you real time, but it will occasionally infer something wrong or repeat an outdated claim. The validation phase is where you catch that. Never let an assumption become part of your permanent knowledge without confirming it. This is the same discipline the whole method rests on: AI drafts, you decide.
Verify, even here
The same discipline applies to every output the teammate produces on a live account, not only the one-off Foundation build. It applies just as much to the worked examples in this method as it will to your own real accounts.
Every output an AI teammate produces reads confidently. That is exactly the habit to watch for. In a real run, the AI will occasionally state a title, a date, a figure or a market claim with the same fluent confidence, whether or not it is actually grounded in your source material. Evidence and inference tags are the teammate telling you where it is guessing. It will not always tag itself so honestly, and it can be wrong even where it sounds sure. Treat every AI-stated fact, above all names, dates, titles and numbers, as unverified until you have checked it against the source.
Notice the shape of that habit: separate evidence from inference, and check anything before you act on it or say it to a customer. It is the first of the eight reasoning standards you built into the Foundation, and it is the one most likely to erode quietly once a teammate has been reliable for a while. Fluency is not accuracy. A confident, well-formatted answer is not the same thing as a correct one.
Save it once, on Claude
A practical note once your Foundation is validated and saved. You do not have to paste the Companion Skill every time you start a new account conversation.
On Claude, paste the Companion Skill once and ask Claude to save it as a Skill. From then on it is available on demand, no pasting required. Claude is also the stronger choice for build-to-format outputs such as the SIGNAL Account Plan. The paste-in method remains the fallback for any platform without a saved-skills feature.
The habit, summarised
Validate before you save. Confirm or correct every assumed item before it becomes part of your Permanent Knowledge document.
Verify, even here. Treat every AI-stated fact, above all names, dates, titles and numbers, as unverified until checked against the source.
AI drafts, you decide. Judgement, the thing that never switches off, is what turns a fluent draft into something you can act on with a customer.
Want the whole method as one document?The complete field guide has the full walkthrough, all six build exercises, and the appendix of ready-to-paste prompts, including the Setup and Companion Skills.
Download the Field Guide