Enablement, coaching and ramp

Leaders rarely choose the surveillance loop deliberately. They choose a dashboard, and the loop arrives with it. The damage sequence is predictable enough that you can watch for it:

The coaching loop costs one behavioural commitment and gets you the opposite: reps who bring you the wobbling deal on purpose, because they have learned the recording is for them.

Consent is the part everyone agrees with in principle and quietly skips in practice, because saying it out loud feels awkward for about four seconds. Give your reps the words and the awkwardness disappears. This is the whole script:

"Before we get going - I use a notetaker that records and transcribes
this call so I can stay in the conversation instead of typing. It stays
in our system, I'll send you the recap afterwards, and I'm happy to turn
it off if you'd rather. Is that okay with you?"

Then: wait for an actual answer. Then start the recording.

A line in the calendar invite footer: "Meetings may be recorded for
quality purposes." The bot joins, announces itself in the chat, and
nobody says anything.

• Before, not during. Consent captured after the recording starts is not consent, it is a notification. This is the rule the source map carries into leader session a5 as well.
• Check your jurisdictions, not just your head office. Recording consent varies by country and by one-party versus all-party rules. If your team sells across borders the script has to satisfy the strictest jurisdiction on the call.
• Give a real exit. "Happy to turn it off" has to be true. A rep who has never once turned it off is probably not offering it properly.
• This is orientation, not legal advice. Take your script to counsel with a list of the jurisdictions you sell into. Session a5 covers the full compliance rails.

Most automated call scores drift toward personality traits dressed as metrics, and personality is not coachable in a one-to-one. A usable rubric scores things a rep can change on the next call by deciding to.

Two further rails. Scores are inputs to a conversation, never outputs of a judgement - the rubric tells you what to listen for in the call, it does not tell you the answer. And every rubric item must be demonstrable in a real recorded call from your own team, otherwise you are coaching a theory.

Forty calls is the ceiling. Ten curated calls is the thing that actually works, because a new hire with an unfiltered archive will watch three and give up. The curation is manager work and it cannot be delegated to a search box.

Pair each call with a single question to answer in writing. "At what point did the buyer decide this was worth their time?" beats "watch and take notes" by a distance, and it gives you something to talk about in the debrief.

An AI call review will hand a manager fourteen observations about a single conversation, all of them defensible. Deliver all fourteen and the rep changes nothing, because fourteen simultaneous corrections is not feedback, it is a verdict on their competence.

• One behaviour. One week. Chosen together. The rep should be able to say what they are working on without checking their notes. If they cannot, you picked too many things.
• It must be observable in a recording. "Be more consultative" cannot be checked. "Ask for the timeline before talking about price" can be, on Thursday, in call three.
• Measure it on their calls, and show them. This is where the tooling genuinely earns its money: you can count the behaviour across ten calls instead of arguing about impressions.
• Then stop. When the behaviour holds for two weeks, close it and pick the next one. Leaving old items open turns coaching into an ever-growing list of ways the rep is failing.

Four behaviours a month, thirty-six a year, each one actually landed. Compare that with the fourteen-point review that changes nothing, and the arithmetic makes itself.

Three things belong to a human manager permanently, and each one gets quietly outsourced the moment a tool offers to draft it:

• The difficult conversation. Underperformance, a broken commitment, a behaviour that is hurting the team. An AI-drafted version of this is legible as AI-drafted within one sentence, and the rep will remember for years that you could not be bothered to write it yourself.
• The performance plan. It has consequences for someone's livelihood, so a person has to own every line of it. This is the same rail as session 3's committed number: a promise, and a threat, both stay human.
• The praise. Most overlooked of the three. Automated recognition is worth less than none, because it teaches the team that your attention can be faked. Specific praise, said by name, about a thing you actually noticed, is the cheapest management tool that exists and AI makes it cheaper to counterfeit.

Everything upstream of those three is fair game: transcription, summaries, pattern-finding, the objection library, the ramp curriculum, the search across four hundred calls that no human would ever run. Automate the preparation. Never automate the relationship.

Every sales org has an objection-handling document, and almost every one of them was written by marketing, in a room, from imagination. The version built from your own recorded calls is a different artifact entirely, because it contains what buyers actually say and what actually worked.

This is one of the few enablement assets that genuinely gets better with volume, and it is the clearest answer to the question your CFO will eventually ask about what the recording tooling is actually producing.