A big reason that token costs become an issue in organizations is leadership is not making decisions about how to use AI, the processes that need to change as a result, & how to think about multiplayer use.
Solving it by rationing intelligence first is a pretty blunt strategy
Many users blamed leadership inaction and lack of ownership for driving up AI token costs, arguing that rationing and caps without understanding workflows cause waste and AI rejection.
No Digg Deeper questions have been answered for this story yet.
Token spend is not just about productivity, it also organizational R&D, an investment in figuring how to rebuild around AI that will involve some failure & adaption.
A big reason that token costs become an issue in organizations is leadership is not making decisions about how to use AI, the processes that need to change as a result, & how to think about multiplayer use.
Solving it by rationing intelligence first is a pretty blunt strategy
@emollick Definitely noticing a problem where token cap is low so it forces people to use bad models and they write off AI as a whole
@emollick fr leadership kicks the hard decisions down and then wonders why the token bill is insane
@emollick Very much agree with this. And also the whole "we didn't budget for this" is silly bcs who knew late last year what the world would look like now.
@emollick A lot of variability in how much individuals and groups are using it and for what level of development.
@emollick Skill Issue. If the people at top where technically literate they would have the AI build an ROI calculator and adjust per role/person.
@emollick saw a team get a hard token cap from finance engineers just went back to personal accounts on their phones for anything real
savings on the api bill showed up as new browser tabs
@emollick We are stuck with models that are amazing and workplaces who haven't figured out how to integrate AI.
2027 it'll start to shift imo
Modell Capabilities (We are here) Workflow Integration Institutional Transformation
@emollick Isn't this just about actually thinking about AI use as an expense? I mean, if someone has idea for something that can be built/automated, and AI is the proposed method, then just cost it in, right?
@emollick Reliable execution is becoming a systems problem, not just a model problem.
@emollick Rationing tokens is the equivalent of buying computers but limiting everyone to 2 hours/day — the constraint isn't the hardware, it's the workflow. Until someone decides what decisions AI should own vs. assist, orgs will keep managing cost instead of managing transformation.
@emollick let the mids be your guide
@emollick Rationing tokens before understanding usage patterns is optimizing the wrong thing. Organizations need to know where AI creates value before cutting access.
@emollick What I noticed in teams I've watched is the spend spikes when nobody owns the decision. People retry the same prompt five ways because there's no agreed path, and that shows up as token cost.
@emollick Exactly. If leadership treats AI as a cost problem before a workflow problem, the value never really shows up.
@emollick Rationing tokens optimizes the cheap input. On my invoice app the agent "wasted" tokens re-checking its own PDF output and caught a bug that would've shipped, a support ticket and refund avoided. The costly failure is the one that ships, not the one the loop catches.
@emollick "is not just about X, it also Y"
What happened to you brother?
@emollick The "failure & adaption" part is the actual product. Gave an agent one prompt to build a bilingual invoice app; it shipped, caught its own PDF bug, and fixed it. Token spend that buys a self-correcting loop is the cheapest R&D line you'll run.
@emollick Seen this play out — teams get handed seats, nobody redesigns the workflow, then leadership sees the bill and caps usage. The rationing always lands hardest on the people who actually found a good use for it.