Skip to main content
You do not need to know any tool names, because your assistant works out which part of Exec to reach for. Treat the prompts below as starting points rather than syntax and reword them freely. Anyone can use any of them: what comes back is scoped to what you can already see in Exec, so a rep gets their own practice and a manager gets their team.

​
Your first five prompts

If you only try five things, try these. Together they walk the whole loop: read, diagnose, chase, coach, create.
1

Get the lay of the land

2

Find the gap

3

See what is outstanding

4

Understand a weak session

5

Build something new

Prompts 2 and 4 carry an extra clause on purpose, and it is not padding. Ask for the real thing, further down this page, explains why, and it is the most useful five minutes of reading here.
If you have other tools connected alongside Exec, start with “Using the Exec MCP, …” so your assistant knows which one you mean.

​
If you run enablement

​
If you manage a team

Ask about one person at a time when you want their transcripts or detailed feedback. Ask about the whole team when you want scores and themes. Full transcripts are large, so a team-wide request for them will get trimmed.

​
If you are practicing

You see your own practice. You do not need a manager or an admin to run any of these.

​
One tip that saves the most time

Ask for feedback across several sessions in a single question rather than one at a time:
That is one request, not ten. Your assistant can pull the feedback for many sessions at once, and it is much faster than working through them individually. Save the deep dive, with the full transcript, for the one session you actually want to examine.

​
Making your workspace easier to ask about

Your assistant finds things by name, so the names you use in Exec decide how well it can answer.
  • Descriptive scenario names beat numbered ones. “Discovery call with a skeptical CFO” is findable. “Scenario 4” is not.
  • Groups make team questions possible. Without groups, “how is my team doing” has no team to refer to.
  • Consistent skill names let you compare across scenarios and over time.
This is worth ten minutes before you roll the connection out widely, and it improves the product for everyone regardless.

​
Getting reliable answers

Everything above is what to ask. This is how to make sure the answer is one you can act on. Your assistant answers the question you asked. That sounds obvious, and it is the single thing most worth understanding about working this way. Ask for the worst performer and you get whichever row is numerically worst. On a real workspace that is very often a row that is technically true and practically meaningless. The rest of this page is how to avoid that, and how to check an answer before you act on it.

​
Ask for the real thing

Every workspace accumulates rows that are not really results: skills nobody has practiced, sessions somebody opened and abandoned, people who were assigned something last week. Those rows hold the extreme values, so they are exactly what a superlative question surfaces first.

​
Weakest skill

Asked the first way, you get the skill with the lowest average score. In a workspace tracking 64 skills, that turned out to be a skill scoring 13 out of 100, where only 21 people had been scored out of 498 who had it assigned. Elsewhere in the same workspace, a skill read as 100 percent proficient on the strength of a single session. Neither number means anything, and nothing in the answer warns you, because the answer was right. It gave you the lowest average, which is what the question asked for. Adding the floor is what turns that into the thing you actually meant.

​
Lowest scoring session

Asked the first way, you get sessions scoring zero, unranked, with no duration at all. Nobody performed badly in those. Nobody performed. There is no conversation to review and nothing to coach. Asking for a ranked session is the reliable fix, because gold, silver, and bronze are graded attempts and unranked is where abandoned ones collect.

​
The general pattern

Turn this around and it becomes one of the most useful questions you can ask: Which skills do we not have enough practice data to judge yet? That finds your coverage gaps, which is usually more actionable than any score.

​
Check the answer

You are going to act on some of these answers, and a few of them will end up in a review conversation with one of your reps. So it is worth knowing how to see the evidence. Your assistant will show its work if you ask. Three questions, in increasing depth:
1

Ask what it counted

The fastest sanity check there is. A confident claim resting on four sessions is a different thing from the same claim resting on four hundred.
2

Ask which sessions

Now you can open any of them in Exec and see the same thing your assistant saw.
3

Ask for the moment

This is the level you want before giving anybody feedback. Quoting the actual moment is far more useful in a coaching conversation than repeating a summary, and it lets you judge for yourself whether the grade was fair.

​
Things worth double-checking

  • Any number you are about to forward. Ask what it counted before you paste it into a deck.
  • Comparisons between teams. Check that both groups actually have comparable amounts of practice. Two averages are not comparable if one rests on ten times the data.
  • Anything phrased as a cause. Your assistant will happily say practice improved performance if you ask it to look. Practice data and outcome data moving together is not proof that one caused the other, and a claim like that will get challenged.
  • A grade you find surprising. Pull the criteria and the quoted moment. If it still looks wrong, that is worth raising with us rather than passing along.

​
If an answer looks wrong

Work through it in this order:
  1. Ask what it counted. Most surprising answers are a small-sample artifact.
  2. Ask it to list the underlying sessions, then check one in Exec. If Exec disagrees with your assistant, tell us.
  3. Check you are seeing everything you should. If you are a manager and results look thin, you may only be seeing your own groups. See What the AI can see.
  4. Check the answer is not empty for a connection reason. An unfinished sign-in returns nothing rather than an error, and an assistant will describe an empty workspace without suspecting itself. See Troubleshooting.

​
Next

Practicing on your own

Your own practice, with a video walkthrough

Managing your team

The manager walkthrough, end to end

Managing the org

Reading the whole workspace

What the AI can see

Permissions, in full