Pega's move to outcome-based pricing reflects a broader shift away from AI tokens

At Pega World earlier this month, the company announced it was ditching tokens in favor of outcome-based pricing, a system where the company only gets paid if an AI agent produces a favorable outcome.

What's driving this change? Don Schuerman, CTO and head of marketing at Pega, who recently joined me on the FastForward on PPN podcast, told me they were hearing that customers were growing increasingly uncomfortable with token pricing with a seemingly endlessly running meter, and were looking for something that gave them more cost certainty.

"We realized the clients aren't going to want to pay an unpredictable [amount] based on how much AI they use," he told me on the podcast. "They want to pay for something that is actually tied closer to the outcome of the work that's actually being done."

Schuerman's boss, company founder and CEO Alan Trefler, said in a separate interview that the whole notion of tokens has been designed to make the cost structure difficult to understand. "Why do they charge by tokens? You know, a token is three quarters of a word? It's mathematical. They could have charged by the word and everybody would have understood what they thought they were paying for, but no, they have tokens," Trefler said.

Pega is not alone, as companies shift away from token-based pricing approaches. Last week Salesforce announced a new general-purpose help agent called Agentforce Help Agent based on a pay-per-resolution process. The customer only pays if the interaction gets resolved satisfactorily by the agent. If an issue gets escalated or the customer reports a bad experience, Salesforce doesn't get paid.

The danger with the outcome-based approach is that you could undercharge and cheat yourself, or you could overcharge and you're back to square one with the customer. Yes, they have cost certainty, and it's certainly high. But Pega in particular, which is based on cases, believes it is well suited to an outcome-based model, and Schuerman says they can do this without shifting the pricing unpredictability from the customer to them as the vendor.

"When we ran the numbers we realized that we could actually charge people a simple uplift based on the number of cases that they do and felt pretty confident that they could deploy as many agents as they want inside that workflow and we would keep the token cost in range," Schuerman said.

Pega believes this approach is possible because of the way customers use its workflows around a specific measurable outcome like an insurance claim or a loan, but it's not always inherently obvious what done is, and that's the tricky part of basing pricing on outcomes.

It depends what done is

Rebecca Wettemann, founder and principal analyst at Valoir, says the outcome-based approach is admirable, but Salesforce's version raises a measurement problem that could apply to both companies. "The challenge is how you measure or count a positive resolution, particularly when a customer may try to interact with a company through multiple channels," Wettemann said.

But talking about Salesforce's approach, IDC analyst Oru Mohiuddin sees it giving businesses a clearer way to measure ROI. "Tying cost to that outcome gives businesses a clear, measurable ROI — the AI either completed the task or it didn't — and gives Salesforce a compelling and honest way to take this product to market," she said.

The ambiguity isn't quite the same for both companies. Salesforce's challenge is determining whether an interaction was successfully resolved by the agent. Pega's is deciding when a case is truly over. In Pega's case, it's not clear, for example, what happens if a claim goes through multiple rounds, or a loan gets rejected and resubmitted.

But Schuerman says Pega is shifting the unpredictability to the design phase, which he believes helps ensure once the workflow is running in production, it's less likely to do unforeseeable things. "There is less chance of the agent spinning off and doing a bunch of stuff I don't want it to do, less chance of it leading to an outcome I don't want and less chance of it leading to a cost I don't want," he said.

The fact is that companies have to come up with more stable approaches to pricing because nobody likes the meter running with no way to measure the outcome until the bill comes. Vendors are experimenting with different approaches trying to figure out the best way to charge customers in this brave new world, and we are still very much in the experimental stage. Whether outcome-based pricing ends up being the metric or not could depend on how both the customer and the vendor feel about the final bill.