Every meeting type gets analysed differently, because a discovery call and a one-to-one are trying to establish entirely different things. Setting that up is an administrator's job and it happens once.
Meeting types live in Operations HQ, under Meeting Types, not inside Workspace. Workspace is where the results are read; that screen is where the questions are set.
What a meeting type is
Four things.
A name, which is what appears in the calendar picker.
A department, which decides who sees it. Sales, Engineering, Human Resources, Finance, Marketing, Support, Operations, or General.
A framework, which is the shape of what a meeting of this kind is meant to establish.
Outcomes, the specific things measured, written in plain language.
Nobody writes a prompt
This is worth stating plainly, because it is the part most tools get wrong.
You do not write instructions for a language model. You pick a framework, which fills in a starting list of outcomes in ordinary English, and then you edit that list. The instruction the analysis actually runs is composed behind the scenes from what you chose, and it is regenerated whenever you change something.
The consequence is that the person who knows what a good discovery call looks like can configure it. They do not also need to be the person who knows how to write for a model, and the two are rarely the same person.
Frameworks
Frameworks come from a catalog, organised per department, plus a general set available everywhere. Your organisation can add its own alongside the ones that ship, and those are visible only to you.
There is always a Custom option, which starts from an empty list for the cases where nothing standard applies.
Picking a framework is a starting point and never a constraint. The outcomes it fills in are yours to change, remove or add to immediately, and most organisations end up somewhere between the framework they picked and their own vocabulary.
The five kinds of outcome
Each outcome is a question with a shape. The shape decides how the answer is recorded and how it displays later.
| Kind | Answer | Good for |
|---|---|---|
| Score | A number, out of ten by default | How strongly something came through |
| Yes / No | True or false | Facts that either happened or did not |
| Choice | One from a list you define | Stages, categories, named states |
| Rubric | One to five, with the reasoning | Judgements that need their evidence attached |
| Free text | A short written answer | Anything that genuinely resists a number |
Choose the narrowest kind that fits. Free text is the constant temptation and the reliable trap: it produces something to read for every single meeting and nothing at all to compare across meetings. If a question can honestly be answered yes or no, make it yes or no.
Each outcome also carries a plain-language note about how to judge it. Something like: consider budget qualified when a figure or a range was actually stated, not when it was alluded to.
That note does more for consistency than any amount of clever prompting, because it is the thing that stops the same conversation being scored differently in March and in June.
Designing a set that works
Four to six outcomes. Fewer is not worth the setup effort; more and the analysis is grading a dissertation, at which point the results turn mushy and everybody stops trusting them.
Only things a listener could genuinely tell. Whether budget was discussed is knowable from what was said. Whether the buyer is sincere is not, and asking will produce a confident number about nothing, which is worse than no number because people act on it.
Say what good means. The guidance note is where consistency comes from and it is the field people skip.
Never measure two things in one outcome. "Budget and timeline established" cannot be answered honestly when one was and the other was not, and whichever way it is answered, the record is now wrong.
Use the same vocabulary as your team. If everybody says "champion", do not call the outcome "internal advocate". The scorecard is read by people, and translation is friction.
Who sees which types
People see their own department's types plus the general ones. Administrators, and anybody senior enough, see everything.
That keeps the calendar picker short and relevant. A salesperson does not scroll past nine engineering ceremonies to find a discovery call, and a picker short enough to scan is a picker people actually use.
The general set is for meetings every department has: one-to-ones, reviews, kickoffs, retrospectives.
What comes out
Every recorded meeting of that type carries a scorecard showing each outcome with its measured value, in whatever form you chose.
The value here is cumulative rather than per meeting. One discovery call scored against a framework is mildly interesting. Sixty of them, broken down by person and by quarter, is the answer to which parts of qualification your team is quietly skipping, and that is a question nobody can answer from memory.
Internal only
Outcome results never appear in anything sent to a participant. They are internal business measurements and the separation is enforced rather than remembered.
The practical consequence is liberating: you can measure things you would never put in a customer-facing document, which is precisely what makes them worth measuring in the first place.
Changing them later
Editing a meeting type changes how future meetings are analysed. Past meetings keep whatever results they were given at the time.
That is worth understanding before a large rewrite. Changing an outcome halfway through a quarter means the quarter is not comparable end to end, and somebody will eventually compare it anyway.
Better to change them at a boundary, or to add a new outcome alongside an existing one rather than redefining what an existing one means. Adding is safe; redefining quietly invalidates history.
A reasonable first configuration
Start with two types, not twelve. The one meeting your team has most, and one-to-ones.
Use the shipped framework unchanged for a month. You will learn more from the results than from theorising about the outcomes.
Then edit once, properly, with the team that runs those meetings in the room. Configuring this alone produces a set that measures what one person thinks matters.
Did this answer your question?
No, ask a person