Twelve months ago, I watched leaders marvel as AI spun up full authentication systems and built data dashboards in minutes. Fast-forward to today, and some boards I work with share a different mood. The question is no longer “What can AI do?” but rather, “Why isn’t it doing it better, reliably, and at scale?”
If this resonates—if you’ve seen AI deliver brilliance one day and stumble on a simple table layout the next—you’re not alone. The leap from amazement to expectation is reshaping how C-suites and investors view AI investments across every sector.
Here’s where expectations are colliding with reality:
There’s a growing call for procurement to look less like cloud storage (pay for capacity) and more like managed HR or logistics (pay for successful hires, on-time shipments, and realised benefits).
As Microsoft Research puts it,
“As generative AI becomes more capable and widely deployed, familiar questions from the governance of other transformative technologies have resurfaced. Which opportunities, capabilities, risks, and impacts should be evaluated? Who should conduct evaluations, and at what stages of the technology lifecycle? What tests or measurements are required? These are not just academic questions—they impact enterprise procurement, operational standards, and ultimately, the value organizations derive from AI investments.”
Microsoft Research Blog, June 2025
When your AI system can’t complete a task that a median developer would finish before lunch, it’s easy to feel stuck. Yet enterprise-grade procurement isn’t just about cutting a bigger cheque or switching vendors (especially when contracts are locked-in and platforms are proprietary).
The answer? Build a robust feedback-and-response mechanism:
As NIST, the US authority on trustworthy AI, states:
“AI evaluation is not a one-time event but a continuous process. For organizations deploying AI at scale, routine diagnostics, feedback mechanisms, and robust standards are essential to ensure systems perform reliably, ethically, and in line with enterprise requirements.”
NIST AI Portal, July 2025
A global retailer recently upended their AI vendor relationship: they stopped paying for queries and started paying for outputs—inventory reconciled, shipments confirmed. The impact? Issue resolution time shrank by 30%, while user satisfaction soared. Vendors responded by proactively plugging feedback into product sprints, streamlining bottlenecks before they became board-level escalations.
This isn’t just theoretical. Industry leaders are evolving AI from a tool of convenience to a true operational partner. According to Google’s active stance,
“Our AI tools enable your organisation to work smarter and make better decisions. Responsible deployment, continuous feedback, and measurable results are at the core of our enterprise AI platforms because organisations now expect more than innovation—they expect outcomes that matter for business.”
Google AI for Organizations, July 2025
As NVIDIA’s CEO recently summarised:
“AI is transitioning into a ‘reasoning era,’ where models will go beyond instant responses and start thinking through problems. … Reinforcement learning and self-training loops are becoming the real game-changer. AI is no longer just learning from humans—it’s teaching itself. … New models aim to ‘think’ before responding—breaking problems into smaller steps, running multiple solution paths, and selecting the best answer.”
Quantum Information Review, Feb 2025
The time for marvelling is over. Board-level leaders now have a clear imperative: don’t accept friction as normal. Leverage contracts, feedback, and continuous evaluation to turn erratic outcomes into measurable gains. The next wave of AI value doesn’t arrive by accident—it’s won by those who expect more and build systems, teams, and relationships that deliver it.
Links:
Quotes: