Technology executives have chased AI adoption since deep learning networks emerged in the late 2010s. Today, boards expect every business executive to have an AI strategy to generate value with generative and agentic AI. The challenge is that AI adoption without clarity of vision is not a sound strategy; it is a vendor outcome metric.

Context

On May 28, the Financial Times published an article on Amazon scrapping its leaderboard for AI adoption. Unsurprisingly, employees were gaming the system by running useless AI agents to maximize token usage, thus incurring significant cost to the company. And Amazon is not alone. Top-down AI adoption mandates are common across companies. OpenAI even recommends it as a best practice in their ‘Staying ahead in the age of AI’ leadership guide:

Set a company-wide AI adoption goal

Define a measurable goal that connects AI adoption to everyday work. This could be new use cases, frequency of AI tool usage, or setting benchmarks for team experimentation, and incorporate these goals into company planning and KPIs. Communicate this goal through all-hands or company updates to build momentum and signal that AI is part of how work gets done.

Executives are following this guidance with the expectation that productivity gains will eventually emerge, but routinely find that Goodhart’s law - When a measure becomes a target, it ceases to be a good measure. - kicks in. Adoption in isolation was never a good measure in the first place.

Your business strategy needs to answer the hard questions, including defining what is most important and saying no to potentially good initiatives. Technology adoption does play an important role in the implementation of such a strategy, but it is not an end in itself, and adoption does not automatically produce business outcomes. That’s why AI adoption is a success metric for the vendor, not for your business.

What should you measure instead?

Good strategy requires a shared clarity of vision. The first question is not ‘what should we measure?’, it is ‘what is most important?’. Articulate the key business outcomes aligned to your strategy, and own them at the executive level. For example, consider how you can use AI to generate new revenue streams, to improve customer experience, or to improve operational effectiveness. Each goal and strategy requires a different approach to AI adoption.

When you have achieved clarity on your strategic priorities, the next question becomes ‘how do we measure success?’. There is no single proxy metric that can tell you whether a technology initiative actually works. Adoption metrics can be inflated because of a top-down mandate or bad incentives, as in the Amazon leaderboard story. To understand the performance of an initiative, you need a closed feedback loop covering three key areas necessary to guide decision making:

  1. business / quantitative metrics (e.g. adoption, revenue) - use to identify patterns of resistance or high adoption. Allows you to replicate successes and eliminate friction that prevents adoption.
  2. voice of the customer (e.g. customer satisfaction, sentiment, talking to customers) - qualitative data and customer anecdotes provide insights into the customer experience - are people adopting because of a top-down mandate or because the initiative is useful?
  3. operational metrics (e.g. cost, performance, availability, security) give you the confidence that things are running efficiently at the right cost and performance.

Remember: the purpose of metrics is to guide your next decisions, not to feel good about your past decisions.

What about grassroots adoption?

Not every AI adoption story needs to be driven by a strategic top-down initiative. We are in an experimentation phase where teams are still figuring out where AI delivers value. In high-performing organizations, employees feel safe to experiment and follow their curiosity. Leaders who channel this creativity through clear priorities create a competitive advantage. Your role is to create an environment and culture where experimentation leads to insights into what works and what does not, which then inform the broader strategy. Here is how:

  1. Actively listen - talk to your customers and employees to understand what matters to them and what challenges they are facing on a daily basis. Don’t mandate adoption of an AI tool that does not address their needs or adds to their burden.
  2. Communicate strategic priorities - experiments that are not anchored in strategic priorities will rarely get sponsorship, even if the ideas are good. Clarity on what matters the most creates focus.
  3. Create mechanisms to capture experiments and scale the best ones. Create visibility into the creative energy, foster collaboration between teams with similar ideas, and sponsor those experiments with proven outcomes that are most aligned with your priorities.

Conclusion

Adopting powerful new technologies does not automatically lead to good business outcomes. Those outcomes come from teams that embrace a culture of curiosity, purpose, and collaboration. Regardless of whether you are implementing a strategic, top-down initiative or whether you are tapping into grass-roots innovation from within your teams, apply the following 3 principles:

  1. Be a truth-seeker - use a combination of data from metrics and anecdotes to get a deep understanding of what works and what does not. Use this feedback to continuously improve. Be mindful of confirmation bias - the tendency to focus on data that validates your point and discount data that doesn’t.
  2. Focus on what is most important - identify strategic priorities and anchor your investments in these priorities. Say no to distractions.
  3. Embrace organizational models that foster collaboration - cross-functional teams with diverse perspectives deliver better outcomes faster, especially in the age of AI. Avoid organizational silos.

Vendors measure adoption. Leaders measure outcomes. Choose what you optimize for.