I had engineering teams build AI agents on data they were sure was ready, and the answers came back wrong in ways nobody caught until later. The problem wasn’t the data or the model. It was the layer between them that translates raw information into something the model can actually reason about correctly. This piece I wrote for CIO Magazine walks through what I call contextual intelligence and why engineering with AI takes a different approach.
Originally published in Forbes
Not so long ago, I attended three customer calls a week, listening to how leaders articulated their priorities, challenges and ambitions. Now, I read 20 AI-powered call summaries. I worry I’ve traded depth for scale.
That trade-off describes the AI paradox that keeps me up at night: AI is only valuable if it doesn’t come at the expense of quality.
Speed is easy. Maintaining quality is hard. AI-powered work needs frameworks that help employees, leaders and boards understand where AI adds value—and where it introduces risk. The good news is, most organizations already have more of the foundation than they realize.
Take the typical sales pipeline slide, which is essentially a measurement framework. A manager uses it to see which deals are stuck in stage three. I use it to ask why win rates dropped in mid-market. A board member uses it to pressure-test the forecast. Everyone reads the same numbers and knows what those numbers mean.
We do not yet have frameworks for AI that allow us to calibrate work and agree upon cultural norms. There is no widely established way to measure whether AI improves the work, where it creates capacity or where it accelerates risk.
It’s time to introduce expectations for AI-powered work. CEOs who establish shared frameworks now will shape how AI is used, rather than being shaped by it.
Principles To Embed In Your Framework
To build frameworks, start with the basics. Managers direct work, evaluate output and own results. CEOs weigh context, refine direction and align investment to outcomes. Boards oversee risk, track performance against the profit and loss statement (P&L) and hold leadership accountable. Those same disciplines must now be applied to how AI-powered work is communicated and measured across employees, leaders, boards and investors. Here are three principles at the top of my list.
1. Use human judgment to curb risks.
It has always felt good to create more work faster with less effort. With AI, that dopamine rush is amplified. AI companies that use consumption-based pricing make interactions feel good—even if quality suffers. A recent Stanford study found AI models provide nearly 50% more flattering responses in similar situations than humans, even when humans query risky, unethical or even criminal lines of conversation.
Meanwhile, the line between human- and AI-produced work is blurring, which makes accountability murky. Add in fractured focus replacing deep, high-quality work, and the cumulative effect is a loss of quality, exacerbated by overreliance on AI and missing human judgment.
Principle one: Employees must remember they were hired because the employer believes they have good judgment. They should structure their work accordingly.
2. Identify misalignment early and often.
AI can accelerate one function so fast it creates misalignment in the next. Most CEOs won’t see it until it’s a problem. Recently, our chief technology officer showed me the progress his engineering team is making with AI: more pull requests, shorter cycle times. My instinct was to invest even more in engineering. He argued the opposite. Unless we invest in product management, project managers will be unable to create requirements fast enough. Changes in staffing and resourcing due to technological innovation isn’t new, but the speed AI facilitates is.
Principle two: Leaders must recognize when they’ve optimized one area—and shift strategy so the rest of the organization can keep pace.
3. Remember stories are not facts.
In board meetings, leaders often bring stories of one-off AI successes, rather than presenting a continuously measured operational model. AI updates need to shift to outcome- and impact-oriented reporting—a transition McKinsey describes as the move from AI adoption to AI execution.
Principle three: Replace point-in-time AI status updates with structured, continuous reporting that clearly assigns ownership, defines goals and links measurement to business outcomes.
How To Establish An AI-Native Framework
Recently, I introduced a simple accountability framework to our team—one I encourage other leaders to adapt. The message is simple:
Everyone who uses AI is now a manager. Responsible AI use requires the same discipline as managing people.
Here’s your new job description:
- Scrutinize output with the same rigor you would give a direct report.
- Prompt AI like an expert. Give clear direction the way a skilled manager would.
- Take full accountability for what AI produces and sends—whether to a co-worker, CEO or customer.
Healthy skepticism is a leadership discipline at every level. For every output, push AI to challenge itself. Ask where it might be wrong. And treat usage limits as a signal, not a success metric.
For leaders, accountability must be equally explicit. Ask yourself:
- What have I done to establish the right culture around AI?
- How do we drive change so teams have the right tools, skills and training to succeed? Functional leaders own this, working with their HR business partners.
- What is our AI road map and how does it connect to the financial plan?
- How do we measure progress beyond usage, and how is work actually changing?
As a leader, the hardest connection has always been between day-to-day work and the P&L. That doesn’t change with AI. But with the right insight and measurement, leaders can drive EBITDA (earnings before interest, taxes, depreciation and amortization) and growth rather than defaulting to blind cuts.
CEOs, Ask This Question
The best question one CEO can ask another right now is simple:
What is your definition of work productivity in the age of AI?
History shows that while new technologies begin unconstrained, cost and reality quickly impose limitations. The same will happen with AI.
Today, the focus is on innovation and quality. Over time, internal and external constraints will shift the emphasis to efficiency. Measurement discipline will be the competitive advantage.
For now, the priority is building fluency in how work gets done when humans and agents work side by side. I believe leaders who invest in shared measurement frameworks today will shape how work evolves—and set the standard for sustainable productivity and business impact in the era of AI.
