credit burn

Share
1 views9/18/2026
When AI adoption moves faster than enterprise decision-making

0 Comments

Be the first to leave a comment.

Show captions (49)
00:00 - 00:03September credit usage is high again.
00:04 - 00:05But we have 75,000 credits.
00:05 - 00:07Usage should stabilize.
00:08 - 00:12Management has repeatedly discussed the situation.
00:12 - 00:15Surely we can make it through September.
00:17 - 00:19And the latest burn forecast?
00:19 - 00:21Everything will be fine?
00:24 - 00:26My Führer...
00:27 - 00:28The model...
00:31 - 00:33At the current burn rate...
00:34 - 00:36...we run out September 23.
00:53 - 00:58Everyone leave except IT, Architecture, and whoever owns the AI strategy.
01:13 - 01:15THIS IS THE THIRD TIME!
01:15 - 01:17WE RAN OUT IN JUNE!
01:18 - 01:23I WROTE AN INCIDENT REVIEW! NO MONITORING! NO OWNER! NO OVERAGE PLAN!
01:25 - 01:28THEN JULY CAME!
01:29 - 01:31WE PREDICTED DEPLETION!
01:31 - 01:34AND WE STILL RAN OUT!
01:34 - 01:37SO I BUILT MONTE CARLO SIMULATIONS!
01:37 - 01:40BUSINESS DAYS! WEEKENDS! HOLIDAYS!
01:40 - 01:42August actually had credits left—
01:42 - 01:46BECAUSE I BEGGED EVERYONE TO PUMP THE BRAKES!
01:46 - 01:48Users are choosing cheaper modes now—
01:48 - 01:52I WROTE A 14-PAGE GUIDE ABOUT THAT!
01:53 - 01:54INSTANT: ZERO!
01:56 - 01:57THINKING: TEN!
01:57 - 02:00PRO: FIFTY CREDITS PER MESSAGE!
02:00 - 02:03I EVEN PUT A LUNCH TOKEN ON THE LINE!
02:04 - 02:08AND AFTER ALL THAT, SEPTEMBER IS BURNING EVEN FASTER!
02:08 - 02:13WE'RE ON TRACK TO RUN OUT A FULL WEEK BEFORE MONTH-END!
02:14 - 02:16AND THE ENGINEERS WANT CODEX!
02:17 - 02:21"CODEX AND THEN CODEX AND THEN MAYBE CODEX."
02:27 - 02:29But the plugin is still "in progress."
02:30 - 02:34We have usage data for a tool we haven't approved.
02:34 - 02:36Fine. Build the AI gateway.
02:41 - 02:42Authentication.
02:43 - 02:47Logging. Cost controls. Multiple models. One controlled API layer.
02:48 - 02:53Something we can actually integrate into our products!
02:54 - 02:56How far have we gotten?
02:56 - 02:59We discussed it again at Technology Meeting.
03:00 - 03:02OF COURSE WE DID.
03:04 - 03:07Maybe they'll ask him for another dashboard.
03:14 - 03:16What more evidence do they need?
03:19 - 03:23Usage reports. Heat maps. Gini coefficients. Burn forecasts.
03:25 - 03:26Monte Carlo.
03:31 - 03:33Employee productivity surveys.
03:40 - 03:46I'm using AI to analyze AI so I can explain why we need to make decisions about AI.
03:46 - 03:49And we still don't have Codex.
03:53 - 03:56Fine, we'll do a performance review