OpenAI has published guidance for enterprise leaders on managing AI investment as usage shifts from simple chat interactions to longer-running, multi-step agentic workflows. The piece, posted on 14 July, argues that falling token prices do not by themselves indicate value; leaders should instead track useful work per dollar, including tasks completed and decisions improved.

The company sets out five recommended steps: sharpening visibility into who is using AI and for what; evaluating models on outcome-based ROI rather than raw token cost; governing advanced workflows, including plugin, connector and Computer Use access, before they scale; funding workflows that compound value across the organisation; and matching production capacity, such as Guaranteed Capacity or Scale Tier, to proven demand.

OpenAI points to updated usage analytics and spend controls in its Admin Console as the practical mechanism for the first step, giving admins visibility by user, product and model. For governance, it flags its AI Deployment Engineers service and enterprise privacy controls, including Zero Data Retention, as support for sensitive high-stakes deployments.

Agentic AI’s Hidden Tax: Why $50K Token Bills Are Catching Firms Off Guard
Are you prepared for the AI cost line that doesn’t yet appear on your budget? Rick Doten founded Prescient Cyber Risk earlier this year, and his first months advising boards, VC firms, and startups have surfaced a critical blind spot: agent token cost. Only 2% of the people he speaks with currently run agents in production— yet those who deploy without a token cost strategy are discovering monthly bills of $50,000 before anyone raises the alarm. Meanwhile, 90% of AI security is work organisations already know how to do. The gap is in what they’re not yet planning for. Speakers: Rick Doten, AI & Cybersecurity Advisor, Prescient Cyber Risk; interviewed by Stewart Tinson, Project Director, AI-360 at Newton Media. You’ll learn: • Why 90% of AI security is traditional cybersecurity—asset management, identity, data protection, and vulnerability management • How to manage agent token costs before they become an unplanned CFO problem • The 7-step automated remediation framework replacing manual security patching • Why AI enablement and AI governance are not the same thing—and how to sell the reframe internally • What VCs now require before funding AI security startups, and what it signals for enterprise buyers • How to structure board-level AI strategy around three questions: plan, compliance, and security Key topics: Agentic AI adoption • Agent token cost • AI governance frameworks • Automated remediation • AI enablement vs governance • Board AI strategy • VC investment dynamics • AI security startups • Third-party AI risk • Crowdsourced threat intelligence For CISOs, CIOs, CFOs, and enterprise risk leaders preparing for the shift from AI pilots to production in 2026.
BrightTALK
Share this post
The link has been copied!