AI, Productivity and the Illusion of Efficiency

We are in the middle of one of the most aggressively hyped technology cycles in decades. Every major firm is “doing AI”. Boards are asking for AI strategies. Entire functions are being restructured around automation and augmentation.

So one would expect to see this showing up clearly in productivity data - but it is not.

Across the UK, US, and most advanced economies, productivity growth remains weak. Output per hour has barely shifted relative to pre-pandemic trends. Outside a handful of frontier firms and sectors, there is no broad-based acceleration. According to the UK's ONS, output per hour grew just 0.6% in the past year — barely above the pre-pandemic trend. The US has seen similar patterns outside a few tech-heavy sectors.

AI is delivering efficiency. It is not yet delivering growth.

Economists define productivity as output per unit of input (often labour). But efficiency is not the same thing. You can make a process leaner, cheaper or faster without increasing the total value of what’s produced.

AI, in its current stage, is often delivering private efficiency (cost savings for individual firms) rather than social productivity (aggregate growth). For instance:

• A legal firm uses AI to automate document review and reduces its wage bill, but the saved costs don’t necessarily become new output elsewhere.

• A call centre automating customer support might increase efficiency, but if displaced workers aren’t reemployed productively, aggregate output across the economy can even fall.

At the aggregate level, that distinction becomes critical. If labour is displaced and not reabsorbed into higher-value activity, total output does not rise in line with efficiency gains.

Throughout history, productivity gains have only translated into growth when the saved resources are reinvested in new activities, innovation or human capital. For instance steam power did not just make existing processes cheaper. It enabled entirely new forms of production. Computing did not just reduce admin costs. It created new industries.

The same logic applies here: AI creates capacity but growth depends on what you do with it. At a firm level, that means making deliberate choices about reinvestment:

• Do you use savings to reduce headcount or to build new capabilities?

• Do you compress costs or expand output?

• Do you automate to maintain margins or to increase market share?

Without that second step the impact of AI remains bounded. Businesses capture efficiency gains but do not shift their growth trajectory.

Most organisations are approaching it as an efficiency programme because that is the easiest way to justify investment. The returns are immediate and measurable and the narrative is familiar.

But this framing quietly constrains the upside. You get a more efficient cost base but not a fundamentally different growth trajectory.

The harder question is what to do with the capacity that efficiency creates. That requires a shift from thinking about optimisation to thinking about allocation: The firm has more effective capacity than before and the question is where it should be deployed.

The firms that benefit most from AI will be the ones that treat efficiency gains as an input into expansion. That might mean using freed capacity to accelerate product development, enter adjacent markets, deepen customer relationships, or increase the speed and scale of experimentation. It might mean redesigning roles so that time saved on lower-value tasks is systematically redirected towards higher-value work.

If your firm is adopting AI, the strategic question is not simply which tools to buy or which tasks to automate. It is how to turn efficiency gains into measurable commercial value. In practice, that means treating AI as a business redesign and capital allocation problem, not just a technology rollout.

This is where RG Economics comes in and can help in a few specific ways:

1. Run an AI value diagnostic: Most firms have a fragmented view of AI impact. Some teams report time savings, others report cost reductions, but very few have a clear, business-wide picture of where value is genuinely being created. This involves mapping AI use cases across functions, separating real efficiency gains from superficial ones, and identifying which activities are actually shifting margins, throughput, or revenue potential.

2. Quantify freed capacity and where it should go: Cost savings on their own are not strategically useful. The more important question is what capacity has been created and where it should go. This means translating efficiency gains into hours, roles, and throughput, and then assessing where reallocating that capacity would generate the highest return, whether through faster delivery, deeper customer coverage, or new product and service lines.

3. Model the commercial trade-offs between cost-cutting and reinvestment: This is fundamentally an economic allocation problem. Firms can extract savings immediately or reinvest them to drive future growth. Building simple but robust scenarios that compare these paths over a two to three year horizon makes the trade-offs explicit, and allows leadership teams to see the impact on revenue, margins and market position before committing.

4. Redesign workflows and roles so the gains are actually captured: As we've discussed, efficiency gains do not automatically translate into higher output. In most organisations, freed time is quickly absorbed unless workflows, responsibilities, and expectations are deliberately reset. This step focuses on restructuring how work is done so that AI-driven efficiency is converted into higher-value activity rather than dissipating across the business.

The underlying point is straightforward. AI-driven efficiency does not automatically become growth. It only does so when someone works through where value is being created, how resources should be reallocated and which operating changes are needed to make the gains stick. That is not a generic transformation exercise, it is a commercial and economic one. And that is precisely the kind of work RG Economics is designed to do.