Most business owners understand that AI tools are getting better. Teams report that tasks are completed faster and outputs improve. Yet revenue per employee and overall output look much the same. That gap between capability and measurable results is highlighted by the Forecasting Research Institute’s survey.

In what claims to be the most comprehensive study of its kind, economists expect meaningful progress in AI by 2030. But they do not expect a step change in economic growth. The median forecast holds close to current trends.

US GDP growth is forecast at around 2.5 percent through to 2050, only slightly above today’s level. Labour force participation edges down from 62.6 percent to 58 percent. The expectation is that we adopt the technology without transforming the headline numbers.

This view reflects a world where AI improves steadily and becomes more capable, but where adoption spreads slowly and unevenly. Economists assign a 61 percent combined probability to moderate or rapid progress in AI capabilities by 2030. Even with that expectation, their core economic forecasts barely move. That tells you that the constraint sits with the system, not the technology.

Let’s call this diffusion drag. It is when new capability enters the economy faster than businesses can reorganise around it. While the tools work, the structure of an organisation does not change at the same pace. That gap delays any visible impact on productivity and growth.

There is a lot of press about how AI is transforming the nature of warfare. The reality is a lot more familiar to businesses struggling to isolate gains from AI tools. The US Army’s VictorBot is being built to give soldiers real-time answers based on lessons from past missions. It draws on hundreds of internal data sources to help with tasks such as configuring complex equipment in the field. This should reduce repeated mistakes and speed up decision making under pressure.

Yet the focus of the project is on knowledge management, data integration and system reliability. Before a soldier can trust a response, the data has to be cleaned, structured and connected across siloed units. That is a back office problem. The Army is effectively rebuilding how information flows before the tool can scale.

Even in a high urgency environment with generous funding, adoption depends on slow groundwork. That is the pattern economists expect to play out across the wider economy. Some industries adopt quickly, others lag for years.

AI at scale needs energy, chips and data centre capacity. Infrastructure constraints slow deployment even when the software is ready. On top of that there are demographic pressures and geopolitical friction, both of which weigh on growth regardless of technology.

There is also a timing issue. General purpose technologies take years to show up in productivity statistics. The benefits arrive in layers. Early gains improve existing tasks, while larger gains require changes to processes, pricing and business models. Most firms stop at the first layer.

The survey makes this clear in another way. Most of the disagreement between economists is not about how powerful AI will become. It comes from uncertainty about what businesses will do with it. Behaviour rather than capability is the limiting factor.

The message for SME owners is that waiting for AI to lift the market will not work. Gains will stay local and incremental unless you force them through the business.

The first step is to track where time is saved. Measure the hours released across people’s roles, rather than just on specific tasks. If an estimator saves ten hours a week, that is capacity you can redeploy or remove. If you do nothing, it becomes idle time.

The second step is to link that capacity to revenue. More quotes, faster turnaround and tighter follow up. The aim is to increase throughput, rather than just reduce effort. If volume does not increase, the efficiency gain has no commercial value.

The third step is pricing. Faster and more consistent delivery gives you a basis to change how you charge. You can price for speed, reliability or guaranteed turnaround rather than time spent. That shifts the conversation away from cost and towards outcome. Done properly, it protects margin, lifts perceived value and creates headroom to grow without adding cost at the same rate.

The fourth step is to simplify processes before layering in more automation. Clear workflows allow the technology to compound rather than stall.

The broader lesson is that AI does not change outcomes on its own. It changes what is possible. The economy only moves when businesses convert that possibility into different decisions about work, pricing and capacity.

The economists are effectively saying that most firms will not do that fast enough to shift the macro numbers. That is the opportunity. If you move ahead of that curve, the gains are real and immediate. If you follow it, the gains stay hidden inside your own operation.