A successful AI pilot doesn't mean your AI strategy is working
The Next Web published an interview-based analysis with Zacharie Lahmi, founder of Ellis Partners, arguing that launching an AI pilot is relatively straightforward while integrating it reliably and at scale into business operations is far harder. The article cites McKinsey's global 2026 survey, in which only 37% of respondents said AI had improved their company's earnings, unchanged from the previous year, while eight in ten said AI had helped them become more productive at work. Lahmi recommends evaluating AI opportunities one business process at a time, estimating cost and resources and agreeing measurable success criteria before a pilot, and looking beyond sales and marketing to core operational processes.
The piece argues the test for enterprise AI should shift from a successful demo to a dependable, cost-controlled and auditable tool in production, offering a framework for judging enterprise AI returns.
Launching an AI pilot is relatively straightforward. But integrating it reliably and at scale into business operations is a far greater challenge. McKinsey’s global 2026 survey found that only 37% of respondents said AI had improved their company’s earnings, unchanged from the previous year. Yet eight in ten said AI had helped them become more productive at work.
An AI tool may work well during a small test, when the task is simple and the data is clean. But using it across a company brings new challenges. It must integrate with existing software, comply with data access and security requirements, and have a clear process for handling failures. Someone must also take responsibility for its performance. Meeting these requirements means coordinating across IT, operations, finance, and risk teams.