ElevenLabs keeps the commission in the agent business case
Paying the account owner when an agent closes revenue can buy cooperation. It also means the human commission belongs in the agent's operating cost.
ElevenLabs' account owners still earn commission when an AI agent brings in revenue on their accounts. In Jason Lemkin's September 9 recap of a conversation with early ElevenLabs executive Carles Reina, that policy is presented as a way to make the sales team welcome the agents. The person whose account gets more productive shares the upside. Read the interview recap.
That is a more useful detail than another revenue-per-employee headline. The compensation plan tells us which human expense survives the automation. An agent can perform the activity while the account owner continues to receive the variable payment.
My rule for this business case: commission follows the contract, not the hands that did the work. If the contract still pays the human, a spreadsheet cannot remove that expense because a model now performs the task.
Cooperation has a price
The interview's rationale makes sense. A salesperson who loses income every time an agent succeeds has a reason to keep useful context away from it. Paying for the account's result can align the person with the new workflow. That is an incentive design described by the speaker, not evidence that sabotage actually occurred.
But an adoption expense and a labour saving are different entries. A company might rationally spend more per account for a while because it can reach customers it previously neglected. It might also free experienced sellers to handle complex deals without reducing anybody's compensation. Both can be good outcomes. Neither is automatically a reduction in payroll.
I would separate the decision into three questions: what new contribution arrived, which cash expenses disappeared, and where the recovered time went. Combining those into one impressive productivity percentage makes it difficult to tell whether the organisation bought growth or bought a cheaper process.
Price the accepted outcome
Consider an illustrative renewal workflow. An agent identifies an opportunity, prepares an offer and gets an acceptance; the account owner receives the usual commission. I would put model usage, workflow infrastructure, human exception handling and that commission against the incremental contribution from the renewal. I would not charge the account owner's entire salary to each transaction, but I would not erase a payment that still happens either.
There is another awkward entry: revenue that would have renewed anyway. If every agent-touched renewal is credited as new agent revenue, the project receives credit for the installed base. A comparison needs a defensible baseline, such as similar accounts handled through the previous process, with obvious differences in account quality made visible.
This is my proposed evaluation method, not a test I have run on ElevenLabs' accounts. The public interview does not provide the transaction data needed to calculate its incremental return. Its growth figures should not be read as a controlled experiment on the compensation policy.
Before Friday, take one agent business case and label every claimed benefit as added contribution, removed cash expense or redeployed capacity. Then ask the compensation owner to confirm which variable payments actually change. A surprisingly productive meeting can start with a column the first spreadsheet omitted.
The policy may be exactly right. Buying cooperation can be worth more than immediately capturing every saving. The decision becomes clearer when cooperation is an explicit cost with an expected payoff, rather than an invisible deduction from the automation story.
Keep the human commission in the agent calculation until the compensation contract actually removes it.
