As AI agents join SaaS, AWS tells users to expect more pricing puzzles

Cloud giant says choice and flexibility matter more than standardization – for now

Interview As agentic AI solutions flood the market, users will face a complex environment in terms of deployment and commercial models, with standard practices yet to be resolved, says Olawale Oladehin, AWS director, solutions architecture.

Take the example of workflow automation software company Pegasystems, which recently celebrated a financial turnaround on the back of its switch from process automation to so-called agentic AI, causing third-quarter revenue to rise 17 percent from the previous year.

"In applications, Pegasystems is working with us to take legacy applications which may have run on-prem or in a hybrid environment and modernize them in the cloud," he tells The Register.

The vision is that legacy applications can be combined with newer elements and sewn together with LLM agents to present the end user with a coherent experience. Although AWS is not a major provider of enterprise business applications, it is building infrastructure and tooling to support this. Under the hood, Pegasystems uses AWS Bedrock, a managed service offering a choice of foundational models through a single API.

The model sees the creation of a big ecosystem with a broad spectrum of pondlife. Enterprise application providers – both SaaS and legacy – will rub shoulders with LLM providers and companies promising nebulous concepts like automation, workflow, and a seamless experience.

On the one hand, Oladehin sees this simplifying application integration with a set of open agent standards, including the Model Context Protocol (MCP) and Agent-to-Agent (A2A).

"We used to integrate systems in four or five different ways: through transferring files, through API calls, through SOAP and REST," he says. "You're creating this complexity for customers, and what we've done with a few standards is make it understandable and reduce the complexity."

But the same does not apply to commercial arrangements as vendors grapple with new licensing models, with consumption and outcome-based pricing added to per-user or perpetual licensing from SaaS and on-prem models.

"When you think about that evolution of software, you started with completely on-prem, you have a fixed license, you transition to SaaS pay-as-you-go pricing, and now we have this new shift: what does consumption look like? How many API calls do I need to this AI model? And so, how do you help a customer rationalize that? We've really focused on giving customers choice and flexibility in their pricing."

Service desk provider Zendesk is using a value or outcome-based pricing model.

Such examples frame AI as part of users' broader objectives, which might be helpful in ensuring organizations get value from software, but requires more thought up front.

"What value do you see being added? What are you going to do to extend your product? Where are your new opportunities? And in all those cases, I think there will be much more dimensionality and choice for these customers. It is definitely a nuanced conversation, and I would say it's early days," Oladehin tells us.

Looking ahead, AI agents could give users more options in deciding what to do with legacy apps.

For example, ECC, SAP's legacy ERP platform, is still used by some of the world's largest companies. Gartner estimates more than 60 percent of SAP customers remain on ECC 6 on-premises with no decision yet to move. By 2030, the consultancy expects four in ten ECC customers today to still use ECC in key business areas. Mainstream support ends in 2027, while extended support ends in 2030 at a 2 percent premium.

Speaking to The Register in May, Alessandro Galimberti, Gartner senior director analyst and expert in public cloud infrastructures, told us at the time that ECC users have an alternative to SAP's preferred upgrade path of S/4HANA in the cloud.

"You might want to look at this strategy to limit your lock-in. You can use all this extensibility from the other cloud providers as well. Try to use the best tool for the job."

Oracle also supports armies of users on its legacy database and applications, although its support deadline is less pressing.

Oladehin says AWS works with the big application providers to support their recommended upgrade path: it is a partner to the German vendor's RISE with SAP program, for example. But it also says it wants to give users a choice.

"Customers are saying, through AI, I can operate across these different systems. There's so much more untapped potential as we continue to work together across these software companies. It's still very early days as we start to see the rise of agentic work across all these systems," he says.

early days indeed, with much room for improvement. According to research published in the summer, AI agents get office tasks wrong around 70 percent of the time, and two in five pilot projects will be cancelled by 2027 due to unclear business value and insufficient risk controls. Purveyors of agentic AI have a lot of challenges beyond deployment complexity and fine-tuning pricing models. ®