Why AI Spending Works

The First-order Effect of Cheaper, Faster AI

Last week, Mustafa Suleyman explained on the Moonshots podcast how all businesses can benefit from heavy AI spending. We are moving away from an environment defined by operating systems, browsers, apps and search engines. We are moving towards agents and companions.

Suleyman co-founded DeepMind and is now CEO of Microsoft AI. He knows what he is talking about. This shift is for real.

For the last forty years, productivity gains came from teaching humans how to use software. For those that didn’t code this meant menus, workflows, dashboards and shortcuts. Entire industries grew around training people to operate machines more efficiently.

Agentic AI flips that relationship because now the machine understands the human. You explain what you want and the system figures out how to do it.

The biggest criticism of today’s AI boom is cost. The capital expenditure on data centres, chips, power and talent is colossal. To many it looks excessive, even reckless.

Yet this spending represents the competition to provide the next generation of productivity tools. Not all the big tech companies can win. Every small business can however, because they will use the tools that turn out to be the best.

Whether or not we’re in an AI investment bubble, the spending lays the groundwork for something businesses have always wanted. Faster tools, cheaper execution and less friction between intent and outcome.

As models get faster and cheaper, and as agents become more capable, the marginal cost of cognitive labour trends downwards. This covers drafting, analysing, forecasting, scheduling, reconciling, researching and coordinating. All of it becomes less expensive and more available.

This is why Suleyman positions Microsoft as a “platform of platforms.” It does not need to own the best models when it already sits inside business workflows. Email, documents, spreadsheets, CRM and ERP all revolve around Microsoft 365. When agents live inside those tools, productivity gains will follow.

For management teams in smaller companies, the implication is blunt. You do not need to build models, or understand transformers, or follow every AI announcement. Your job is simpler and harder at the same time. Embed AI where work already happens.

You don’t need your own little investment bubble. Adopting as many tools as possible is uncoordinated and expensive. The winners will be the firms that redesign their processes around delegation to a careful selection of machines.

That means less time spent operating software and more time spent defining outcomes, checking quality and making decisions. Humans get to move closer to the action and will require a better understanding of how a company makes money. The resulting increase in agency will raise job satisfaction.

That is the first-order effect of cheaper, faster AI. The second-order effect is more interesting.

Why Small Businesses Are Structurally Well-Placed

Small businesses are often framed as victims of technological change. They have fewer resources, smaller budgets and less room for error. That framing misses what makes them powerful.

Small firms have three structural advantages. These are speed of decision-making, proximity to customers and flexibility in process. AI amplifies all three.

Large organisations struggle to deploy agentic systems because their processes are rigid and political. Small businesses can change how work gets done in weeks.

They also sit closer to real customer problems. That matters because AI is valuable when paired with context. A small firm knows its edge cases and what “good” looks like. That allows it to supervise AI output more effectively than a distant corporate function.

Flexibility is the final advantage. Small companies are not locked into legacy workflows. They can let agents draft proposals, prepare estimates, manage follow-ups, handle first-line support and reconcile accounts, without redesigning an entire organisation chart.

You can already see this playing out in industries where small businesses flourish today.

Professional services is the obvious one. Marketing agencies, consultants and design studios trade on expertise and responsiveness. AI agents now handle research, first drafts, reporting and administration. The human layer focuses on judgement, creativity and client trust. Margins improve without a commensurate increase in headcount.

Construction and trades are another. Small contractors already win on local knowledge and speed. AI-powered estimation, scheduling, compliance checks and procurement reduce overhead and error. A five-person firm can now operate with the back-office capability of a much larger competitor.

E-commerce and niche retail provide a third example. Product research, pricing, ad optimisation, customer service and inventory forecasting are increasingly automated. Small operators can test, adapt and personalise faster than large brands burdened by process and brand risk.

In each case, AI concentrates the human advantage. The competitive edge shifts from scale of labour to quality of orchestration.

What Has to Change

This opportunity does not come for free.

Managers require new skills and must move from doing and reviewing to specifying and supervising. This mirrors the shift from operating tools to designing systems. It draws on an age-old theme of what makes for the best managers. They do not need to be the smartest person in the room, but they should be the clearest.

As a result, the businesses that win will be those that reinvent with AI, rather than treating it as another software to learn.

Questions to Ask and Answer

1. Where in my business does a human have to operate software, rather than define outcomes?

2. If AI could handle half of our cognitive workload, what would my team’s time be best spent on instead?

3. What would stop a smaller, faster competitor from using AI to undercut or outpace us within 24 months?