BT150 zeitgeist: It’s time to get practical about AI agents

CxOs in Constellation Research's BT150 are getting practical about AI agents, looking to align AI strategies with business outcomes with timelines to structurally reposition operating models with AI in about three years.

The BT150 meetup was held April 3. This Constellation Research CxO call operates under Chatham House rules so the takeaways aren't attributed.

BT150 CxO zeitgeist: AI agent growing pains abound | The 2026 SuperNova Awards Nomination Form is Now Open | BT150 Executives

Key takeaways include:

• Set explicit time horizons for AI projects and bucket them into 90-day experiments, 12-month operating model pilots and three-year structural changes.
• Define AI initiatives by business outcome with clear metrics and owners and unit economics. Contracts with vendors should shift to these metrics and outcomes.
• Beware agent sprawl. One CxO said there's a fear of agent sprawl and creating the equivalent of "a billion Excel macros." Governance is critical.
• Design and agency governance model. For each use case decide whether agents can observe only, act with a human in the loop, or act autonomously with an audit. Enterprises also need to treat agents as first-class actors in risk and control.
• CxOs will need to decide which platform will be the orchestrator of agents across the AI estate. Avoid ad-hoc deployment of AI agents in every SaaS product without central visibility.
• Building applications is cheaper than buy, but you need to decide what you want to own and differentiation. Decide what vendors are only increasing tech debt and rip them out.
• Define the lines between must own and must outsource. Own/build where the system directly shapes your competitive advantage or differentiation and encodes your unique workflows, data, or decision logic. You buy or partner with generic problems like CRM and HR and a vendor can maintain and update faster. For each major system decide whether it’s becoming more or less strategic in an AI operating model and can automate workflows.
• Design roles around orchestration and agent managers as well as outcome owners.
• Define a small set of AI customer journeys with well-defined roles for AI.
• Stand up an AI agency and governance working group that will be cross functional and decide authority models and where humans get involved, standards for logging, audit and kill switches and outcomes.

Related:

• AI Forum 2026: There isn’t an easy button for AI
• Constellation Research's Futures Forum: What CEOs are thinking