Welcome to this week’s edition of Twimbit X Insights.

Software is starting to get paid for the work it finishes, which raises a question every team will soon answer: what counts as finished?

The Invoice With No Seats

Picture a support leader opening this month’s invoice and finding a line for every conversation an agent resolved overnight.

That bill is already arriving, with CRM makers HubSpot and Salesforce and service platforms Intercom and Zendesk now charging for part of their agents by outcome. HubSpot’s Prospecting Agent, for one, costs $1 for each qualified lead it recommends to a sales team.

Intercom shows the idea best: its agent, Fin, answers a company’s customer questions and handles tasks like refunds, and Intercom bills only when a conversation is resolved, meaning the customer asks for no further help after Fin’s last answer.

Fin’s average resolution rate has climbed from 30% at launch to 76%. Intercom puts its own money on the line: a high-volume customer in its success program receives $1M from Intercom if Fin resolves fewer than 65% of that customer’s conversations.

Why Intercom Switched

Intercom’s customer research pointed the way: buyers disliked paying per conversation because the bill arrives even when the agent is no help, and they preferred to pay when a problem got solved.

Charging for resolutions also changed every team’s job at once, says Intercom’s president, Archana Agrawal. Sales moved from counting seats to asking how much volume a customer wants to automate, success shifted from watching logins to tuning the agent, and finance began forecasting resolutions.

Customer service was a natural fit because it already had a clear definition of done, and buyers who arrive knowing exactly what counts can shape the terms.

Which Jobs Deserve an Agent?

Paying for results works only when someone defines a result, and for the work your team owns, that someone is you.

Before handing anything over, ask three questions: is the result one countable unit, does the job repeat often enough to absorb an occasional miss, and can someone check it quickly?

That last question matters most, because agents can take on more of the doing while the judgment stays with people. Someone has to decide what good looks like, notice when a task is technically finished but still misses the mark, and answer for the result when it reaches a customer.

It also protects your budget, since the trustworthiness of a bill that counts finished tasks rests on the standard your team sets.

Which three jobs on your team could an agent finish today, and who would judge whether it finished them well?