I saw a headline recently claiming AI is basically a waste of money, and that much of the current boom is companies building solutions for problems that do not exist. There is something in it. AI tools frequently create distraction loops, where the time spent making the tool work exceeds the time it saves.
You may have invested time, money or both into AI in the last twelve months. There is a reasonable chance you have questioned whether you are getting out what you have put in. The output is faster. It is not always better.
That disconnect is the defining challenge of how small businesses adopt AI right now. McKinsey's State of AI shows the great majority of businesses using AI somewhere in their operations, with a much smaller proportion reporting meaningful value capture. The Stanford AI Index confirms the same pattern at small business level. Use has increased. Value has not increased at the same rate.
The problem sits in the layer underneath AI.
It is not the tool
For over two years the public conversation about AI in business has been almost entirely about tools. Which tools. Which prompts. How to automate workflows. Which jobs are at risk. Tool conversations are useful for a window, because they make the abstract feel concrete. But they have a ceiling, and we have hit it.
When the printing press arrived in the 1400s it was revolutionary, welcomed by scholars and a growing middle class, and it caused panic among elite scribes. For a brief period the conversation was about which press to buy. Then it stopped being about the machine and became: what am I going to print, who am I trying to reach, what do I have worth printing. The tool became infrastructure. The questions became about capability.
That is the moment we are in with AI. The tools have commoditised. The differentiators sit underneath them, in the human, leadership, business and operational layers that determine what the tool gets used for.
AI is an opportunity rather than a fix. It does not repair what is not working. Often it amplifies it.
The multiplier
Capability × AI = Output. A multiplication rather than an addition. Ten plus zero is ten. Ten times zero is zero.
If your underlying capability is strong, AI is genuinely transformative. It compounds you. If your underlying capability is weak, AI gives you something else entirely: fast, mediocre output at scale.
If you have been wondering why AI is not delivering ROI in your business, this multiplication is usually the answer.
This is not a new pattern. Erik Brynjolfsson's work at Stanford, going back almost a decade, shows AI is most accurately understood as augmentation, a multiplier of existing human and organisational capability rather than a replacement for it. The companies that capture AI's value capture it because they have the capability infrastructure underneath. Those that do not, do not.
Because there are people with tools who want you to buy their tools, the popular conversation has been a selling-tools conversation. That frame requires you to believe the tool is enough. For some businesses it is, because they have already done the real work. For others it leads to tools and processes that distract from the work that would actually help.
The four capability layers
From the research I have led at 25eight, there are four capability layers AI amplifies, plus two enabling conditions: capacity and community. Three of the four are barely discussed in public. The fourth, the tools layer, is where almost all the attention goes. The disproportion is the diagnosis.
1. Human capability. The psychological foundation you bring to the work. Mindset, self-efficacy, creativity, resilience, and the originality that lets you see what others do not. AI is neither a confidence builder nor a creativity substitute. It amplifies whatever clarity, or lack of it, you bring.
2. Leadership capability. How you direct, decide and delegate, including to the AI agents you now run. An owner gets value from a tool personally, rolls it out, the team uses it badly, and the owner concludes AI does not work. The actual problem is upstream. The AI is exposing it.
3. Business capability. Strategy, vision, customer understanding, brand and the problem you solve, alongside operational integrity, systems, data and financial management. AI loves clean inputs. Faithful execution against unclear strategy and partial data produces faithful chaos at scale.
4. Digital and AI capability. Prompting, workflow design, model selection, knowing when to leave AI out of a task, recognising when it has produced something wrong. It matters, and it sits on top of the other three. It cannot compensate for weakness underneath.
A quick test for the business layer. If I walked in tomorrow and asked your team what the strategy is for the next twelve months and how it shapes the work they are doing this week, would the answers be coherent and consistent? What if I asked you? In many growing businesses the honest answer is that a lot still lives in the owner's head.
When AI is not worth it yet
So, is AI worth it for a small business? Sometimes the honest answer is: for yours, right now, no.
The real cost of AI for a small business is rarely the subscription. It is the owner's attention. Every hour spent trialling tools is an hour that did not go to customers, staff or cash flow, and for a business whose underlying system is struggling, that trade is a bad one.
Signs it is worth waiting:
- You cannot say clearly what you would point AI at. A tool without a target becomes a hobby.
- Your strategy, pricing or processes live mostly in your head. AI cannot multiply what is not written down.
- The team is already stretched to breaking. A new tool adds load before it removes any.
- The numbers you would feed it are unreliable. Automating on top of bad data scales the badness.
None of these mean never. They mean the sequence matters. Fix the layer underneath first, and the same tools that were a distraction become a multiplier. That is also the cheapest path to AI ROI a small business has, because capability work costs effort rather than subscriptions, and it pays off whether or not you ever adopt another tool.
If digital capability alone were the bottleneck, the extraordinary volume of free, high-quality AI training available right now would already have closed the gap. It has not, which tells you where the gap actually is.
What AI cannot create
There is one category of capability AI cannot create. It can amplify it and support it. It cannot manufacture it. That category is embedded human capability. Capability that lives in the body, shows up reliably under pressure, and survives the actual conditions of running a business.
Reading about leadership does not make you a leader. Knowing about strategy does not make you strategic. Chris Argyris made this point precisely in Teaching Smart People How to Learn: high achievers are particularly vulnerable to the gap between knowing and doing, because intellectual mastery often substitutes for behavioural mastery in their self-image. Pfeffer and Sutton extended the case at organisational level. Brinkerhoff's work on training transfer found that in unsupported environments only around thirty per cent of formal learning translates to sustained behaviour change.
These capabilities have to be built. In you, in your team, in your business. Through structured experience, real stakes and applied repetition. They cannot be downloaded, prompted or outsourced to a model.
That shifts the playing field. Smaller businesses now have the same knowledge, the same tools and the same insights as larger enterprises. What used to be enterprise advantage has been democratised. The competitive advantage moves to the layer that has not been: human capability, built deliberately, applied to your conditions.
A question worth sitting with
If capability multiplied by AI equals output, and AI is now roughly the same for everyone, then your output is determined almost entirely by your capability.
So: where is your capability strongest, where is it weakest, and what are you doing about it?
Sources
- McKinsey, State of AI
- Stanford Institute for Human-Centered AI, AI Index
- Brynjolfsson, E. Work on AI as augmentation, Stanford
- Bharadwaj, A. et al. Digital business strategy, MIS Quarterly, 2013
- Westerman, G., Bonnet, D. and McAfee, A. Leading Digital, 2014
- Argyris, C. Teaching Smart People How to Learn, Harvard Business Review, 1991
- Pfeffer, J. and Sutton, R. The Knowing-Doing Gap
- Brinkerhoff, R. Research on training transfer
You don't need every tool, only the useful ones
One practical email a week on making confident calls about AI, without losing what already works.