AI is costly to run and sometimes wrong. How a product charges tells you who absorbs those costs.
- ChatGPT: Plus is $20 a month, and Business is $20 per seat billed annually or $25 monthly, with a two-seat minimum. You pay for access. OpenAI carries the cost risk, since a power user costs far more than a casual one, so each tier has usage caps. The user carries the risk of a wrong answer
- GitHub Copilot: Business costs $19 per user per month. You pay for a person. In June 2026 it moved from premium requests to usage-based AI credits, so heavy agent use now draws down a credit pool. That shifts some cost risk back to the customer. Bad code still costs the developer time
- Intercom Fin: $0.99 per successful resolution, with a 50-resolution monthly minimum. You pay for an outcome. There is no charge when Fin escalates to a human or a customer asks for one, so Intercom carries most of the risk. The catch is that the bot defines the resolution, not the customer
Poyar’s point shows up in all three. None is pure. ChatGPT adds usage caps, Copilot adds a usage paywall to seats, and Fin sits on top of Intercom’s seat fees.
Our product
Our Slack bot answers questions about ongoing projects so PMs don’t need a meeting to share context. Outcome pricing sounds ideal, but “a meeting avoided” is hard to measure. I would charge per seat for PMs who connect projects, give free lite seats to engineers and leaders who only ask questions, and add a fair usage cap. Both ideas come from Poyar. The bot’s value spreads across teams, but the buyer is still the PM.
