What changed
On November 11, Google released the Gemini Decisions API with a new model, Gemini Decision 1. It answers three types of questions: yes/no, choose one from a list, or score something on a scale you define. You can ask up to 1,000 questions in one call. It reads text, images, audio, video, and PDFs.
The price is $0.10 per million input tokens. Output is free because the model returns only probabilities, not text. Google says that makes it nearly five times cheaper than using a standard chat model for the same work.
What you can do with it this week
Hand it the checks you do by hand today. Does this expense photo show a receipt? Is this customer inquiry urgent or routine? Does this product review mention a defect? Score incoming sales leads from 1 to 10 based on budget and timeline.
You write the question once, then paste in a batch of invoices, photos, emails, or form responses. The model scores all of them in seconds. One team running 50,000 checks a day would pay about $9 if each check uses 1,800 tokens.
Start with a free Google AI Studio account to test it on 10 real examples from last week. If it gets 9 out of 10 right, build it into your workflow.
What to keep in your own hands
Don't upload customer financial records, medical data, or anything you can't afford to send to Google's servers. The model scores what you give it; it doesn't explain why. So keep the judgment call — the borderline expense, the angry customer email that needs a personal reply — for a person. Use AI to clear the obvious nos and the clear yeses. You handle everything in between.