February 24, 2026 · The Pengo Team
Why we stopped billing per resolution
Per-resolution AI billing sounded fair on paper. In practice it created disputes, unpredictable invoices, and a product story we did not want to sell. Here is what we changed and why.
#ai #pricing #product
When we first shipped Pengo Intelligence, we billed AI only when the bot closed a conversation without a human stepping in. We called it “pay for outcomes, not replies.” It sounded principled.
It was also the wrong model for the product we are building.
What went wrong
Unpredictable invoices. A good month for AI meant a bigger bill. Customers could not forecast spend, and every invoice review turned into a transcript audit.
Disputes over heuristics. Did the visitor confirm resolution? Did they reopen within 72 hours? Was confidence high enough to auto-close? Each edge case became a support ticket about billing, not about the conversation they missed.
Wrong story. Our customers are service businesses, agencies, and SMBs that need reliable customer conversations: fast answers, clear handoffs, predictable cost. Per-resolution pricing sells cost savings; flat credits sell a tool you can budget.
Wrong fit for how teams buy. Per-resolution math only makes sense when your buyer thinks in deflected tickets. Pengo customers think in answered questions and resolved chats. They need a monthly number they can plan around, not a post-mortem on every closed conversation.
What we do now
Each plan includes a pooled monthly allowance of weighted AI message credits:
- 1 credit for a standard AI reply (FAQ answer, product question, policy lookup)
- 10 credits for an advanced action (book a meeting, send a follow-up, run a connector)
- 0 credits for human agent replies and live chat handoffs
When you hit your cap, AI pauses until the next cycle or you add a credit pack. Conversations never get cut off. Your team keeps chatting for free.
No grace periods. No resolution-confidence thresholds. No “did the visitor reopen within 72 hours?” math.
What stayed the same
We still track resolution outcomes in analytics. They are useful for measuring AI quality. We just stopped using them as a billing trigger.
Human handoff on low confidence or negative CSAT is still free. The widget is still under 5KB gzipped. Performance claims are still verifiable on our live benchmark.
The takeaway
Billing should match how customers think about value. Our customers think in answered questions and happy visitors, not resolutions confirmed. Flat credits let them forecast, scale their team without seat math, and focus on conversations instead of invoice archaeology.
See the current tiers on the pricing page.