Enterprise AI agents shift SaaS to service-as-software – News

Everyone’s talking about enterprise AI agents, but few are ready for what it really takes to make them work.

Thousands of startups are building narrow-purpose agents, especially for customer-facing tasks. But real enterprise transformation demands more than just spinning up chatbots — it requires the integration of systems of intelligence to guide actions, systems of record to execute them and systems of engagement to deliver a usable experience. Salesforce Inc. is the only major incumbent building this full stack. 

In a recent interview, I joined theCUBE’s Dave Vellante to talk with Marc Benioff (pictured), co-founder, chairman and chief executive officer of Salesforce, about how the company is rearchitecting enterprise software for the agent era. Benioff laid out why this next evolution — from software as a service (SaaS) to service as software (SaSo) — is more disruptive than the cloud and mobile wave that began 15 years ago. Cloud computing made applications cheaper and more accessible; enterprise AI agents, paired with human oversight, completely reimagine how those applications function.

Below are highlights from our conversation.

How enterprise AI agents amplify human productivity

Benioff compares today’s excitement to Salesforce’s early startup days. He says current CEOs represent “the last generation of executives leading exclusively human workforces.” He is aggressively implementing this vision within Salesforce itself, targeting ambitious 50% productivity increases across engineering, services and support functions through “agentic layers.” 

He expects such gains to continue compounding year after year — a key promise of enterprise AI agents.

Digital labor market

While Salesforce will generate about $40.9 billion in revenue this year as part of its SaaS business, within a total enterprise software industry of roughly $500 billion, the digital labor market could reach $3-12 trillion. Enterprise AI agents and agent-based digital labor convert SaaS into SaSo.

This represents a fundamental shift, as illustrated by customer examples such as OpenTable Inc., which show productivity improvements that were “just not possible a year or two years ago.” Customers and partners will be able to exchange these services via an agent store, similar to Apple’s App Store.

Salesforce integration vs DIY

The architectural foundation of Salesforce’s AI strategy consists of three deeply integrated layers. Salesforce is doing the “heavy lifting” to rewrite and deeply integrate core applications within the Data Cloud. 

Data Cloud integrates data from external and Salesforce sources to provide a rich 4D map of the state of the business. Tableau is an example, which was rebuilt on the Data Cloud, creating fluidity between Sales Cloud, Service Cloud and other apps. This integration allows Tableau to appear within other apps and even inside Slack. Adding Agentforce (the third layer) completes the picture. 

The alternative is DIY integration efforts that only more sophisticated organizations can attempt. And, even then, they are challenged to get the same reliability — especially when building and deploying enterprise AI agents at scale.

Becoming a software-only hyperscaler

Salesforce is becoming a “software-only hyperscaler,” providing cloud-scale application and platform capabilities without building physical data centers. Salesforce provides “data fluidity,” federating data with other data platforms and siloed applications. But when in Data Cloud, data network effects enrich the data so that it’s more useful than when previously trapped in silos. 

One customer, Disney, now has “agent fluidity,” where enterprise AI agents for thousands of theme park attendees can simultaneously access customer preferences, ride availability data and other information across multiple systems to provide recommendations that human staff would struggle to coordinate.

Answering Nadella’s contention that SaaS goes away with agents

Microsoft CEO Satya Nadella went viral several months ago when he suggested that SaaS would disappear with agents directly accessing database schemas. That sounded more like trolling. “Just dumping all your data through some kind of legal discovery API into some big repository and then letting all your employees jostle through that data” would cause significant problems. 

Requirements to govern the data and define it with metadata don’t disappear with AI agents. There needs to be a deterministic software layer that mediates access while layering on non-deterministic agent capabilities. We pushed back on his characterization of Microsoft 365 Copilot as Clippy, but he said Salesforce will be able to federate that data into Data Cloud as well. As for reaching a Klarna moment, Benioff didn’t want to offer a timeframe.

Here’s theCUBE’s complete interview with Marc Benioff:

And make sure to check out our “Road to Service as Software” podcast playlist:

https://www.youtube.com/watch?v=videoseries

Photo: News