In 1987, the Nobel laureate economist Robert Solow famously observed, “You can see the computer age everywhere but in the productivity statistics.” Nearly four decades later, AI has produced a version of its own. Call it the abundance paradox.
At the level of the individual worker, the evidence of AI-driven abundance is easy to see. Researchers found that AI typically raises productivity on specific tasks 15% to 30% in real-world work settings, and 20% to 60% in controlled studies. Meanwhile, 8 in 10 professionals say AI has made them more productive at work, according to McKinsey’s 2026 State of AI survey of more than 1,700 respondents in 97 countries.
At the level of the company, however, the picture looks very different. Only 37% of McKinsey’s respondents can link AI to any impact on their organization’s earnings before interest and taxes, or EBIT—essentially unchanged from a year earlier.
Relatedly, the National Bureau of Economic Research reports that although 69% of firms were using AI, 89% of executives reported no impact on productivity over the previous three years. And PwC’s 2026 Global CEO Survey of more than 4,400 chief executives in 95 countries found that 56% had seen AI deliver neither higher revenue nor lower costs in the past year.
The gains individuals get from AI are not adding up to gains for the companies they work for. Closing this gap is now one of the most important tasks facing business leaders. Here’s how to do it.
From tools to systems
Companies expect abundance to arrive with the tools. But while tools can indeed raise individual productivity, collective, company-level abundance comes not just from the tools themselves but also from the system a company builds around the technology.
History offers a clear precedent. In a classic paper written in response to Solow’s paradox, economic historian Paul David argued that electricity had gone through the same thing. When older factories first electrified, owners typically bolted electric motors onto the existing system of shafts and belts, often keeping the old steam engines in place. The layout of the factory stayed the same, and the effect on measured manufacturing productivity growth remained small.
The breakthrough came when manufacturers redesigned the factory itself, giving each machine its own motor. That freed them to build lighter, single-story plants laid out around the flow of work. By David’s estimate, the spread of factory electric motors was associated with roughly half of the sharp acceleration in manufacturing productivity growth in the 1920s.
Most companies adopting AI today are still at the bolt-on stage. To truly benefit from the possibilities that AI unlocks, companies need systems. An abundance system has three parts.
1. Focus on what AI makes possible
Most companies aim AI at what it can save. In Deloitte’s 2026 State of AI in the Enterprise survey of more than 3,200 leaders in 24 countries, 66% reported efficiency and productivity gains, but only 20% said AI was helping them improve products and services or innovate.
Those savings are worth having. But savings have a ceiling: You can only cut what you already spend. The bigger prize is the work that was never viable in the first place.
I recently experienced this with my own team. Over a six-month period, we built a suite of apps that would have cost roughly five times as much and would have taken around two years longer without AI—economics that would have made the product nonviable from day one.
Now, this doesn’t mean that it was free to build those apps; they still cost a lot in terms of both time and money. But the point is that without AI we wouldn’t have been able to pursue such an ambitious project.
That is the key to unlocking abundance—moving from the scarcity mindset of savings to the abundance mindset of possibility.
The Abundance Principle: Don’t just ask how AI can cut costs. Also ask what it makes possible.
2. Build reusable components
Good software engineering means never writing the same code twice. Efficient approaches are inevitably modular—build a component once and reuse it across every product that needs it. New products can then be assembled from parts that already exist. I was an early adopter of this principle, building an early software company around it long before it became standard practice.
AI pulls most organizations in the opposite direction. Because generating something new is now so cheap, people start from scratch every time: a new draft or a new deck, used once and thrown away; a newly coded solution to the same problem someone else was tackling in the office next door last week. That produces enormous volume, but nothing compounds.
Unilever offers a glimpse of the alternative. The company has created digital twins of its products: AI-driven 3D replicas that hold every variant, label, packaging format, and language version in a single file. Each twin is built once, then used to generate imagery for every sales and marketing channel, serving as a reusable asset that multiplies value. That’s what an abundance system looks like in practice.
The Abundance Principle: Think modular—build assets that can be easily repurposed for different products and for different purposes.
3. Orchestrate resources for the AI workflow
When production multiplies, the pressure shifts downstream, to reviewing, testing, approving, and fitting the pieces together. Organizations that don’t redesign for this shift end up with more work piling up unfinished.
Software development shows this clearly. After GitHub Copilot’s launch, a study of open-source projects found that output rose, but so did rework, and the burden fell on the most experienced developers: They reviewed 6.5% more code while their own coding output fell 19%.
Google’s DORA research program has found the same pattern at scale, with time saved writing code often spent again checking it.
These findings do not suggest a problem with AI. In fact, AI was doing its job. The problem was that the system around it couldn’t absorb the extra output. The fix is to build systems that ensure extra oversight capacity is available to deal with the increased flow of work driven by AI tools.
The Abundance Principle: When output multiplies at one layer, the capacity challenge moves to the new bottleneck.
Four things to do now
Here are four things you can do immediately to start reaping the returns of AI-powered abundance.
- Revisit earlier ideas. List the projects you killed over the past three years because they cost too much or would take too long. Re-scope them with AI to see if the calculations have changed.
- Start a component library. Choose one function and inventory what’s worth reusing: frameworks, research, approved language, templates, code. Make it easy to find. Then set a simple rule: no one builds something new without checking the library first.
- Find the queue. Trace one AI-accelerated workflow from start to finish and look for where finished work sits waiting—for review, approval, or testing. Then move people there.
- Change the scoreboard. Alongside cost savings, track at least one metric for what’s newly possible, such as new products launched or projects attempted.
The system is the advantage
It took roughly four decades for the potential unlocked by electricity to show up in the productivity statistics, because that’s how long it took factories to be rebuilt around it. No one yet knows quite what the economic model of an age of abundance will look like, or who will capture the value when anyone can produce almost anything. But the growth in capacity is already real, and for companies to succeed, they need to design systems that can turn that extra capacity into real growth.
