McKinsey data shows AI pilots fail to boost earnings for most firms
A 2026 survey reveals that while AI boosts productivity, only 37% of companies see improved earnings, highlighting the difficulty of scaling.
Key points
- Only 37% of respondents reported AI improved company earnings in 2026.
- Eight in ten respondents stated AI helped them become more productive at work.
- Scaling AI requires integration with existing software and clear failure processes.
Launching an artificial intelligence pilot is straightforward, but integrating the technology reliably across business operations remains a significant challenge for most organizations. According to a global 2026 survey by McKinsey, only 37% of respondents said AI had improved their company’s earnings, a figure unchanged from the previous year.
Despite the lack of earnings growth, eight in ten respondents reported that AI had helped them become more productive at work. This discrepancy highlights the gap between initial testing and widespread operational adoption. The data suggests that productivity gains do not automatically translate into financial performance without proper scaling strategies.
The Challenge of Scaling AI
An AI tool may perform well during a small test where tasks are simple and data is clean. However, using the system across an entire company introduces complex requirements. The technology must integrate with existing software infrastructure while complying with strict data access and security protocols.
Organizations also need clear processes for handling system failures and designated personnel responsible for performance. Meeting these operational demands requires coordination across IT, operations, finance, and risk management teams. Without this cross-functional alignment, scaling efforts often stall or fail to deliver consistent results.
Cost and Strategic Planning
Costs present another major hurdle for enterprises adopting AI. Companies are accustomed to setting fixed budgets for traditional software projects and estimating staff time. AI expenses are harder to predict because they depend on data volume, model selection, and user frequency.
Zacharie Lahmi, founder of Ellis Partners, recommends assessing AI opportunities one business process at a time. He advises estimating the cost and resources required for each use case before testing. Lahmi emphasizes that return-on-investment discussions should occur before the pilot, not after, to ensure financial viability.
Operational Focus and Success Criteria
Every AI test requires clear goals and a plan for contingencies. Lahmi recommends setting measurable targets for usage, maintenance costs, and error rates. He argues that useful pilots are those intended for production, requiring agreed-upon success criteria beforehand to avoid post-pilot debate about effectiveness.
Lahmi also urges companies to look beyond sales and marketing when selecting AI projects. While those functions offer benefits easier to link to revenue, he believes everyday operational processes offer greater opportunities. The most boring operational tasks and core industry processes are often the best candidates for AI integration.
What should you do?
Organizations planning AI adoption should define clear success metrics and cost estimates before launching pilots. Leaders must ensure cross-departmental coordination for integration, security, and failure management. Focus on core operational processes rather than just revenue-generating functions.
What to do and how to stay safe: McKinsey
- Define measurable success criteria for usage, costs, and errors before starting any AI pilot.
- Estimate total resource requirements and potential costs for each specific use case in advance.
- Ensure cross-functional teams coordinate on integration, security compliance, and failure handling.
- Evaluate core operational processes for AI opportunities, not just sales or marketing functions.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Did AI improve earnings for most companies in the McKinsey survey?
No, only 37% of respondents said AI improved their company’s earnings, unchanged from the previous year.
What percentage of respondents said AI improved productivity?
Eight in ten respondents stated that AI helped them become more productive at work.
Who is Zacharie Lahmi and what does he recommend?
He is the founder of Ellis Partners and recommends assessing AI opportunities one business process at a time with clear cost estimates.
