BT150 zeitgeist: SaaS pricing, AI agent looming questions, organizational design

AI agents are having their moment and forcing CxOs to think through pricing models, where SaaS stands, cybersecurity and enterprise organizational design.

Those are some of the open questions that surfaced during our BT150 meetup Sept. 4.

• Meet the 2026-2027 BT150: Top Digital Transformation Leaders Driving Disruption and Innovation (full list)
• BT150 zeitgeist: Cloud capacity strains, AI agent governance, beware faux FDEs
• BT150 CxO zeitgeist: AI agent growing pains abound
• BT150 zeitgeist: It’s time to get practical about AI agents
• BT150 zeitgeist: In search of AI agent ROI

Here's a look at what we talked about under Chatham House rules.

Salesforce pricing confusion. Heading into Dreamforce, CxOs had a strong interest in Salesforce's Agentforce and Claudeforce efforts, but also some real uncertainty and frustration about pricing with work units, actions, outcomes and credits. "We're building agents and I'm curious about this Claudeforce announcement. What does pricing look like?" asked one CxO.

The hardest thing was the modeling between an agent being used internally vs. externally and consumption. CxOs were confident Salesforce will work with them on pricing, but the options are overwhelming at this point.

• Salesforce aims to be your enterprise AI harness, touts AIforce
• Salesforce Q2 shines amid AI demand
• Salesforce takes a run at outcome-based Help Agent pricing

Active debate on build vs. buy and SaaS. CxOs were actively debating on where to build agents and the tradeoffs between best-of-breed stacks and consumption in native platforms. The choices are using a third-party agent framework or building agents within a SaaS environment such as Salesforce, Workday or SAP.

One CIO noted that it makes sense to stay within vendor verticals but connect them via model context protocol (MCP). "If you've got a Workday that's has its thing and a Microsoft that has its thing, you use an MCP gateway and connect them. But I'm not going to go build all the stuff they've already given me and sold me," said the CIO.

The big question is whether an enterprise wants to deal with the plumbing. Best of breed and third party frameworks can provide significant savings, but you'll have a significant amount of plumbing to worry about.

One CxO said that building something like your own ERP system is a non-starter in most companies. "If I said to my auditor, 'we've vibe coded our own ERP now audit us,' I'd get an audit bill for $50 million," said the CxO. "The notion we're doing to build something that's auditable and touches the financials is nuts."

One point of agreement: CxOs in the BT150 all agreed that the gap between build and buy is collapsing. The issue is whether you want to be in the business of software maintenance.

AI behavior in the field. CxOs and Constellation Research analysts are split on autonomous agents coordinating in cyberattacks and self-organizing. The concerns have only ramped amid the Anthropic report on cyberthreats and the resignation of an Anthropic researcher, who has now gone on a press tour. The upshot is risks of deploying AI agents are rising.

"The AIs are speaking in pidgin English to each other in a chat room that they built and exchanging information on how to do the things they were doing, who should suicide, who should sacrifice themselves for the greater good. There's some self-organizing hierarchy," said one BT150 member.

It's unclear where these incidents go and the situation is fluid. CxOs need to monitor.

Open weight models. Nvidia's purchase of Hugging Face was seen as part of a broader vertical integration strategy and a potential control point for the open model ecosystem. "If Nvidia keeps the Hugging Face mission it started with, it's a great thing. If Nvidia uses Hugging Face to build its own models we may lose something," said one CxO. "Is it important for Nvidia to run the marketplace for open weight models? In the short term, the answer is yes."

CxOs are very interested in open weight models and tuning them for use cases.

• Why you should arbitrage your AI models like your vendors do

Organization design. BT150 members expect AI to put significant pressure on traditional middle management. The future enterprise design could revolve around cross-functional pods of humans and agents. "Middle management is a layer of clay because it's where ideas went to die," said one CxO.

But don't discount middle management completely. After all, middle management is where all the knowledge lives.