Last week we had David Reid, Samuel Little and Pooja Nair from Teneo on PricingSaaS Office Hours. It was a great session, with tons of practical advice on implementing credit models. Here are 5 things from the session I'd want to know if I was rolling out a credit model right now: 1️⃣ Unpredictability loses more deals than price does. In Teneo's work, unpredictability comes up as a loss reason far more often than total price. When customers push back on credits, it's usually about control, not credits themselves. The fix: t-shirt-size bands, forgiving overage in year one, and enterprise contracts that lock in what credits cost at 1M, 10M and 100M. 2️⃣ Get customers used to the counting before you charge. David recommends a 3-month free pilot with monthly value statements: "You used 1,000 credits and saved 500 hours." Charge a small services fee for onboarding, not ARR. That way it never becomes a big, scary approval. 3️⃣ Don't create a new SKU for every AI feature. If a feature doesn't cost much to run, put it in the base platform with a credit allowance. Size that allowance so customers use it up early in the year. Top-up packs are where expansion comes from. (And no one is paying extra for AI summaries anymore.) 4️⃣ 5 to 8 credit burn categories is the sweet spot. Map out the jobs your AI does and you'll find about 50 ways to charge. Put them into 5 to 8 buckets. Keep the subscription and the credit estimate simple. The rate card itself can be more detailed. 5️⃣ Your margins should improve over the life of the contract. Credit models usually target 70 to 85% margin on COGS. Send each task to the cheapest model that does it well and your cost per task keeps falling. With fixed-price credits you keep those savings. With cost-plus pricing you end up passing them on. Dropping the link to the recording and transcript here: https://drive.google.com/drive/folders/158hWJshwYM3-8pgu2B0bFR4F5a5hxym9?usp=drive_link