If an AI can coach your methodology, your methodology is real.

After an engagement that had gone well, I sat down with the client's president and told him he had paid me to enable the wrong audience. The managers were coaching. The field was running the conversation. The client was happy with what they had bought. And the most capable learner in that building had not been in the room.

I was only beginning to absorb what had changed. AI was the first technology in history that could understand and execute conversational dynamics. Everything before it sat alongside the conversation and counted things: calls made, minutes talked, keywords hit, stages moved. None of it could hear what was actually being established between two people. This can, which makes it the first technology that could legitimately become sales technology living in the reality of a conversation rather than beside it.

There is a joke here at my expense. One of my reasons for moving into sales in 2001 was watching the internet take apart whole categories of work and deciding I never wanted to be in a field that could be disrupted. Sales, I was certain, could never be that field. So much for my vision.

The test nobody wants to run

Here is the useful part, and it is the reason to keep reading if you have no interest in my career. Try to encode your methodology and you will find out inside two weeks whether you have one.

Most companies asking for AI coaching have a methodology that lives in three places: a framework name on a slide, a two-day rollout three years ago, and the instincts of two or three good sellers who could not tell you what they do. That is not a methodology. That is folklore with an acronym on it.

A machine cannot work from folklore. To encode a conversation you have to write it down move by move: what a seller is establishing, at which point, and what that sounds like in the buyer's own words once it has been established. The moment you try, the gaps surface. Two of your leaders disagree about what qualified means. Nobody can say what evidence of an urgent problem sounds like. The framework tells a rep what to do and says nothing about how the conversation goes.

None of that is an AI problem. All of it was true before anyone tried to encode anything. The encoding only made it visible, and that is the most valuable thing it does.

Evidence, not vocabulary

If you do get that far, the next decision is the one that separates a coaching system from a compliance checker.

A rep typing they have budget into the CRM is an assumption. The buyer saying what the budget is, on the call, is evidence. A system that scores whether a rep said the right words will be gamed inside a quarter, and it will deserve to be, because what it is measuring is recitation. A system that scores whether the buyer confirmed something has to point back at a line somebody actually said. A score built any other way is an opinion with a number attached.

Sellers can tell the difference immediately, and so can managers. The first kind of system gets quietly discounted. The second gets argued with, which is what you want.

A name on every judgment

Nobody should be judged by a machine, and the fix for that is not a policy. It is an order of operations.

Build the verification step before you tune the scoring, not after. The system proposes an assessment. A named manager confirms it or overrides it. Every assessment that reaches a seller carries the name of the person who confirmed it. The overrides are not a complaints channel, they are the tuning mechanism: they are how the encoding learns what your managers actually mean by good.

That has a second effect worth more than the first. Your managers converge. Two managers who scored the same call two different ways, both able to defend it, now have to reconcile against one rubric. The standard a seller is coached against stops depending on who happened to do the coaching.

What it actually buys

Coaching stops being rationed. Which calls a seller gets developed on has always depended on who had a free hour that week, which is why a seller quietly repeating the same mistake gets seen the fifth time instead of the first. That constraint is arithmetic rather than effort, and arithmetic is the kind of constraint this lifts.

The conversation your company sells with also stops living inside two or three good sellers. Written down move by move, it becomes something you can teach to a new hire, debate as a leadership team, and change on purpose when the market gives you a reason to.

And the number that gets a CFO's attention is not the coaching one. Emulating field behaviors drives a lower cost of sale while increasing conversion rates. That should free capital to invest in higher leverage assets and raise operating income out of a lower operating budget. That is a different conversation from buying a coaching tool.

Working to be irrelevant

There is a version of this that people in my line of work are quietly afraid of, so I will say it plainly. Encoding a methodology into a client's own systems reduces how much they need me. That is the point of it, and it is what I say in the first meeting.

I have made myself irrelevant before, several times, inside companies. Every time, the same thing happened: bigger tasks were waiting. Most firms hold AI back from the coaching loop because it threatens the coaching revenue. That is bad advice and it is a bad trade.

Where to start, if you want to know

Do not start by evaluating tools. Start with a page.

Write your conversation down move by move, with what evidence of each move looks like in a transcript, in a room with the people who actually run it. If you can, you have a methodology, and the software becomes a question of budget and timing. If you cannot, you have found the thing that was going to break the next rollout anyway, and you found it for the price of an afternoon.

Either way you keep the more valuable half.