suggestions for how to burn tokens as a non-engineer (per company mandate)

Work is suddenly measuring us all on token usage (which tbh I thought everyone knew that was a dumb measure of efficiency but large corporations rarely make sense).

As a not-engineer, I'm not sure what types of requests will burn the most tokens. I'm trying to find ways that model usage would actually improve my actual job but so far our AI tools just make super stupid documents with too many words or dataviz junk that reminds me of early power point days when everyone discovered drop shadows at the same time.

So I need some ideas for things that will seem productive even if they aren't (didn't make the world, just trying to live in it).

EDIT: I'm basically a program manager so it's meetings, small reports, pulling metrics, people stuff. None of which is delivering solid opportunities to "Use AI" (reports are super easy for me and it's taking me longer to edit AI output then write from scratch).

Requests for reports from raw data sets seem to take longer than "turn these notes into a doc" so maybe that's burning more tokens? Translating and formatting requests went super fast so that doesn't seem like a good way to stack the numbers (maybe I'm wrong?) Maybe I should start vibe coding (?) like baby apps that do pointless tracking tools? And then make reports out of that?

(This is legit not me trolling, it's an actual problem for many friends and coworkers now thanks to craziness)

Edit #2: maybe just some simple framework like "metrics formatting is more/less token intensive than document generating". Or, "request everything be a powerpoint".