Blog

This is the third post in a series on AI strategy for business leaders. Read about [organizational AI transformation] and [individual AI skills] in previous posts.

A troubling pattern has emerged from my conversations with business professionals: employees are hiding their AI usage at work. In a meeting with a new Tuck MBA student, I learned the company they worked for prior to coming to business school still had a generative AI ban in place, forcing this student to either hide their use of AI or take significantly more time to compile and complete analysis in hours that previously took days, or spend significantly more time on the analysis than necessary.  The student noted they were afraid their manager would see the use of AI as “cheating.”

This secrecy isn’t just counterproductive. It’s a symptom of a leadership crisis. As Stanford research shows 22-25 year-olds facing 13% employment declines in AI-exposed roles, managers are leading a workforce transition using performance systems that actively discourage the adaptation their organizations need to survive.

The Performance Crisis AI Exposed

The difficulty of evaluating knowledge work predates ChatGPT. Unable to measure thinking quality or insight generation, managers defaulted to proxy metrics: hours worked, process adherence, deliverable volume. These input-based metrics were always inadequate, but they were tolerable when analytical work was time-intensive.

AI shattered that tolerance. When analysis that required days now takes hours, what are managers actually evaluating? An employee who uses AI to generate insights quickly and then builds prototypes appears unproductive by traditional metrics while delivering more value.

Meanwhile, workplace training spending has remained flat at around $1,280 per employee for over a decade, right when skill requirements are changing fastest.

From Analysis to Insight: The Real Work Begins

Recently, a Tuck alumnus at an AI-native startup shared his CEO’s approach to requests: “No rush, tomorrow’s ok.” The message is clear—with AI, data processing happens quickly. The valuable work is generating insights and making decisions based on analysis.

This reframes knowledge work value. Analysis was never the point; it was a step toward decision-making. But when analysis required significant manual effort, many employees and managers conflated analytical difficulty with importance. AI eliminates that confusion by making analysis commodity work.

The hard work now: synthesizing meaning from data, identifying non-obvious patterns, making judgment calls with incomplete information, translating insights into action. These require human creativity, experience, and wisdom—capabilities AI augments rather than replaces.

Case Study: Airtable’s Structural Solution

Howie Liu, Airtable’s co-founder and CEO, recognized that AI transformation required restructuring work itself. He reorganized the company into two team structures:

Fast Teams: Ship AI features weekly, optimize for iteration speed, embrace experimentation. Success metrics focus on speed of learning and adaptation.

Slow Teams: Focus on infrastructure, deliberate planning, systematic progress. Traditional metrics around stability and reliability apply.

The insight: different work requires different management. Rather than forcing one performance framework, Liu acknowledged some people need fast iteration space while others maintain operational stability.

Critically, Liu models the behavior he expects; he codes daily and is Airtable’s top global AI user. He implements “mandatory AI experimentation time,” telling employees to cancel meetings for a week to explore tools. This creates organizational permission for exploration that traditional productivity metrics discourage. Listen to his recent interview on Lenny’s Podcast to learn more from Howie Liu.

Managing Hybrid Teams: Humans and AI Together

The analyst-to-executor transition is just the beginning. Forward-thinking managers are learning to manage teams that include both humans and AI agents as collaborative members, not just humans using AI tools.

This requires rethinking team composition, workflow design, and task delegation. When AI agents handle analytical tasks as effectively as humans, managers need new frameworks for optimal task allocation, managing interdependencies between human and AI work, and ensuring effective communication across diverse team members.

Workflow orchestration becomes critical. Traditional project management assumes human-paced handoffs. Hybrid teams operate at different speeds because AI processes information instantly while humans need time for synthesis and judgment. Effective managers design workflows leveraging these different capabilities rather than forcing everything into human-paced processes.

A Manager’s Playbook

1. Shift to Outcome-Based Metrics

Evaluate the quality of insights and actions taken, not time spent. Ask: “What decisions were enabled?” rather than “How much time did this require?” Create space for employees to explain their process, including AI usage.

2. Create Psychological Safety

Model curiosity about AI tools. Share your learning process, including failures. When employees mention using AI, respond with interest not suspicion. Establish forums for sharing experiments, both successes and failures.

3. Emphasize Insight Generation

When employees work faster using AI, dig into the quality of their insights. Help them understand time saved should be reinvested in synthesis and strategy. As the AI-native CEO puts it: if analysis can be done by tomorrow, the real work is generating insights and making decisions.

Actively reward insight quality and decision-making over analytical thoroughness. Ask: “What does this tell us we didn’t know?” and “What should we do differently?”

4. Support Experiential Learning

Create low-cost learning opportunities: internal tool-sharing sessions, conference attendance, cross-functional AI projects, dedicated experimentation time. Focus on learning through real work problems, not formal training.

5. Address the Transition Directly

Help analytical employees understand their value lies in what they do with insights. Provide opportunities to develop execution skills: prototype building, stakeholder communication, project management. Recognize employees who successfully transition.

6. Develop Hybrid Team Skills

Think about AI agents as team members with specific capabilities and limitations, not just tools. Practice workflow design combining human judgment with AI processing power. Learn effective AI communication through prompt engineering. Experiment with task allocation considering both human strengths (synthesis, judgment) and AI strengths (processing, pattern recognition).

The Bottom Line

Leading AI transformation requires managers to evolve as much as their employees. Move from supervision to facilitation, from process management to outcome enablement, from individual evaluation to team learning optimization.

The organizations that master this, like Airtable, will develop sustainable advantages through continuous human-AI collaboration improvement. Those measuring AI-age work with pre-AI metrics will manage increasingly frustrated employees who leave for progressive organizations or hide their innovative work.

The choice: evolve your leadership approach to match AI-transformed work, or manage by metrics that discourage the adaptation your organization needs to thrive.

Discussion Question: What’s your biggest challenge in evaluating knowledge work outcomes versus traditional productivity metrics? How are you creating space for AI experimentation on your team?