According to the 2026 AI Market Study presented at Data Summit 2026, organizations that move beyond pilots into production tend to have several concrete elements in place:

• Dedicated AI budget: Production organizations typically commit an AI budget of over $100K, signaling that AI is treated as a strategic capability, not an experiment.
• Traceable model outputs: They ensure AI model outputs are fully traceable—from inputs through to decisions—so teams can verify, audit, and trust the results.
• Automated AI-readiness assessment: They use automated assessments to evaluate whether data, processes, and governance are ready for AI, instead of relying on ad-hoc checks.

John O’Brien’s core message is that AI pilots are not meant to be simply “switched” into production. Instead, organizations need to reimagine how they operationalize AI by investing in budget, traceability, and readiness processes that support long-term, reliable use.

Traceability is a recurring theme in the 2026 AI Market Study findings shared at Data Summit 2026. Organizations that are successful with AI in production focus on being able to follow the full path from input to output:

• Verification and trust: Teams need to see which inputs led to which outputs so they can verify that models behave as expected.
• Risk and compliance: Fully traceable outputs make it easier to respond to audits, explain decisions, and manage regulatory or internal compliance requirements.
• Operational reliability: When something goes wrong, traceability helps pinpoint whether the issue is with data, the model, or the surrounding workflow.

As O’Brien puts it, “If you want to do this you have to look at the inputs, you have to look at the outputs.” That level of traceability is what allows organizations to rely on AI in real business contexts, not just in controlled pilot environments.

The 2026 AI Market Study suggests that organizations do not need to wait for a perfect, enterprise-wide governance framework before moving AI into production.

John O’Brien highlights that “a little governance is better than none.” In practice, this means focusing on operational governance that is specific to how AI is actually used:

• Define which AI agent is doing which workflow in which context.
• Put basic controls around data inputs, model outputs, and how decisions are used.
• Ensure there is accountability for monitoring and improving AI behavior over time.

This kind of targeted, operational governance is often enough to get started and can be expanded over time. It helps organizations reshape their AI efforts from isolated pilots into manageable, production-grade capabilities without waiting for a massive governance overhaul.