Companies at the leading edge of automated customer service delivery would say that resolution—or outcome based pricing—is more aligned with the value an enterprise client derives from that product. And it’s a timely proposition as the hype around AI abates, and many businesses are starting to scrutinise their return on investment (ROI) for AI transformation projects.

Cloud customer service platform Zendesk’s CEO Tom Eggemeier proffers that the company is the only one applying resolution pricing at scale. But much depends on the definition. The company works with a standard definition which its own customers are comfortable with, but may not be consistent industry-wide. “There’s some nuance in what a resolution actually is,” says Eggemeier.

Eggemeier explains how he was on an airplane in the US about to take off, when he realised that he would not make his connecting flight. He accessed the airline’s app and a chatbot completed around 90% of rearranging his schedule before the flight attendant asked him to turn his phone off.

On landing, Eggemeier couldn’t access the chatbot and ended up engaging with a human agent who didn’t have any context for the query but ended up solving it nevertheless. “The airline was charged $2 for the chatbot interaction, even though it wasn’t solved. So, the airline actually paid more money to the to the software company, by paying $2 for AI agent interaction, and they were also paying for a human seat,” explains Eggemeier.

With Zendesk’s pricing model, the fact that the query was not resolved by the AI agent would not generate a charge. This alignment of outcome-based pricing with AI, is something that is fairly unique in the industry, according to Eggemeier. “There are some really small players that are on this outcome-based pricing, but we’re pretty much the only bigger player that’s on a true resolution based pricing,” he adds.

Resolution pricing with flexibility is critical

Zendesk’s proposition is a hybrid traditional SAS based seat model with levels of resolution pricing built in to capture the uncertainty of how many times a query may require human intervention.

“We want to give people the flexibility because it’s important to know what your costs are up front,” says Eggemeier. “Usually for the first year, if you exceed predicted resolutions, because we want to drive more automation, and achieve ROI of the labour arbitrage there, we kind of cap it so that people have a little more predictability,” he adds.

More resolutions means less seats, and Eggemeier sees this as an equation of value between the two, with flexibility built in as critical. After all, businesses are only at the start of their automation journey where flexibility to transfer between a high-cost human interaction and a cheaper automated one will help the overall path towards greater automation.

The benefit of this march towards automation is not just ROI. “We see higher customer satisfaction on the automated resolution by AI agents. So that’s the other part of it,” says Eggemeier.

Some 80% of interactions between customers and company brands will be automated within five years, Eggemeier predicts, though this will vary by industry. “Business to consumer will be probably higher, business to business will be probably lower. Complex interactions will be human based. Non-complex will be AI based,” he adds.

But is the technology ready? The reputational damage from an AI hallucination or malfunction can have serious downstream consequences. “An AI agent makes a hallucination. A human agent makes a mistake. A hallucination sounds worse than a mistake, quite honestly, even though it could be the same thing, right?” notes Eggemeier.

In addition, the kind of rote task more likely to be automated carries more human risk as humans are less able to provide consistent service over long periods of time. The translation functionality of AI agents is also another factor in the AI agent versus human labour arbitrage.

Another challenge is customer resistance to automated agents based on early interactions with less sophisticated agents launched when the technology was still developing. “Firstly, you should always disclose it is an AI agent,” says Eggemeier who also recommends that each company should have AI agents consistent with their brand in terms of tone and levels of empathy, for example.

No evidence of job displacement (so far)

But does this mean agents will eventually replace humans? According to the World Economic Forum’s 2025 Future of Jobs Report, 41% of employers worldwide intend to reduce their workforce in the next five years due to AI automation. But this does not align with what Eggemeier is seeing in the trenches. He notes that the 16 million people globally that are working in a customer service or helpdesk role has remained “flattish” in terms of headcount for the last couple of years.

“And the clients that have heavily adopted AI have, generally, kept headcount flat or up 3% in hiring humans, which is interesting, it’s counterintuitive,” he says.

“Companies have a choice. It’s almost like an automation dividend, and they decide what they want to do with it,” he says.

“A lot of companies we’ve seen right now are taking some of the savings that they’ve had from automation and reinvesting it into humans to do more sophisticated tasks, to make sure that they’re addressing interactions at the root cause of problems,” he adds.

As a rule, Eggemeier has seen that as automation in customer services increases, overall volumes of queries increase. On a macro level, Eggemeier views this process simply as barriers coming down within the process of delivering customer services rather than job displacement, though he is careful to caveat this by noting that this could change over time.

For the moment, whatever the level of automation, companies are starting to look for meaningful value in their AI transformation projects and innovative and evolving pricing models will surely play an increasing role in this equation.