BT150 zeitgeist: In search of AI agent ROI

Executives in Constellation Research's BT150 agree that AI has reached an inflection point in operations, but still are looking for returns and productivity gains.

The BT150 meetup was held May 28. This Constellation Research CxO call operates under Chatham House rules so the takeaways aren't attributed.

Meet the 2026-2027 BT150: Top Digital Transformation Leaders Driving Disruption and Innovation (full list) | BT150 CxO zeitgeist: AI agent growing pains abound | BT150 zeitgeist: It’s time to get practical about AI agents

Here's a look at the takeaways:

Reality on AI budgets is setting in. The BT150 CxOs all agreed that AI and token budgets have the potential to run amok. The issue isn't "token maxxing," which is a terrible term, but whether the AI spend drives business value.

Several participants said token consumption is hard to explain to stakeholders and doesn't translate well for CFOs. The more meaningful framing is time and effort saved on low‑value work, but that’s also hard to measure rigorously.

The shift to outcomes, productivity and transformation will take time and there's a massive gap between vendor messaging and buyer expectations.

• CIOs see AI agents as journey, token budgets a concern, says Box CEO Levie
• AI inference costs are going to be a big concern: What's the fix?
• AI Forum 2026: There isn’t an easy button for AI
• Constellation Research's Futures Forum: What CEOs are thinking
• Where’s tokenomics for the rest of us?

AI-driven development and productivity. CxOs had concerns about pushing vibe coding and just asking a model anything. The best practice should be structured AI development flows based on analysis, model selection, execution and evaluation instead of using expensive frontier models.

Teams are actively testing multiple models from the same vendor to optimize for costs. You don't need the best model for every use case. One executive said he's hiring more architects to build high-level specification files for models. The hard part right now is developing a system architecture that's coherent, auditable and explainable.

Finding these architects isn't easy. If AI does a lot of the junior work how do you grow senior engineers?

Bottom line: Find the right models, possibly open source models, and architecture not tokens.

AI may mean you need more people. CxOs argued that the narrative that AI will replace headcount is eliminating the generalists who know how the business runs. That shortsighted move breaks the apprentice to journeyman to master pipeline.

Here's why CxOs in the BT150 are coming around to the idea that you may need more smart people to make judgments not fewer.

• You get more “at-bats” because AI accelerates work, but that means you must staff enough capable people to swing at those extra pitches.
• Advanced workflows with no human in the loop are seen as dangerous, especially in regulated contexts.
• At the end of every value chain, someone must be accountable. “The AI decided” is not a defensible stance.

Where AI reality resides. Among executives, AI is no longer a yes/no question; it’s assumed and baked into future decisions and architectures. However, there's a backlash coming. The "we spent money but didn't get value" narrative will surge and force more realistic conversations about prioritization, problem selection and business alignment.