Waste of time?

What's the ROI of AI?

I saw a headline this week that said that AI was basically a waste of money and that much of the current AI boom is a "fake it till you make it" scenario, with companies building solutions for problems that don't exist. And it’s true, AI tools frequently create "distraction loops" where time spent trying to make the tool work exceeds the time it saves.

What do you think?

You may have invested time, money, or both into AI in the last twelve months. Maybe a little, maybe quite a lot?  And there's a reasonable chance you've questioned whether you’re getting out what you have put in or whether it is a complete distraction. Unless you are an alumni of a recent 25eight program that is!   

You may have considered that the output is faster, but it is not always better.

I consider this disconnect to be the defining challenge of how small businesses adopt AI right now. McKinsey's 2024 State of AI shows that the great majority of businesses are using AI somewhere in their operations, but a much smaller proportion report meaningful value capture. The investment curve and the value curve are decoupling. Stanford's AI Index 2024 confirms the same pattern at the small business level: use has increased; value has not increased at the same rate.

And I think the problem sits in the layer underneath AI.

It’s not the tool

For the last two and a half years, the public discourse on AI in business has been almost entirely focused on tools. Which tools to use. Which prompts to use. How to automate workflows. Which jobs are at risk. Which competitors are racing ahead? Tools conversations are useful for a window because they orient us; they help make the abstract feel concrete. But they have a ceiling. And I think we have hit it.

When the printing press arrived in the 1400s, it was revolutionary. It was welcomed by scholars and a growing middle class, while equally causing panic among elite scribes and anxiety among religious authorities. It was an early form of "tech panic" with strong opinions on both sides.

Rather than go into the debates (which I actually do think are interestingly similar to the conversations happening today about laziness and loss of jobs), a lot of the conversation for a brief period was: which printing press should I buy? The Gutenberg, a steam-powered cylinder press, rotary or offset printer, perhaps?

Then the conversation stopped being about this ‘miracle of technology’ machine of ideas and it became: what am I going to print? Who am I trying to reach? What do I have to say that's worth printing? The tool became the infrastructure. The questions then became about capability. 

That's the moment we are currently in with AI. The tools have commoditised. The differentiators are no longer in the tool stack itself, they are sitting underneath it, in the human, leadership, business, and operational layers that determine what the tool gets used for.

I think we have reached a point where awareness of AI is reasonably high and a large proportion of businesses are using or intend to adopt AI within the next 12 months. We’ve gone from not using AI at all to deciding whether it’s Chat GPT, Claude or other AI chat bots or integrated workspace tools. It’s not about the tool now, it’s about how you use it, what you use it for and the outputs. 

And this is where it can get a bit tricky for business. Because AI is an opportunity and not a fix. It doesn’t fix what may not currently be working as well as it could. In fact, often it amplifies it. 

The framework

Here is a simple framework to consider:

Capability × AI = Output.

Capability multiplied by AI. Equals output. It is a multiplication, not an addition. In numbers: 10 + 0 is 10, but 10 × 0 is 0.

This is important to understand because, if your underlying capability is strong, AI is genuinely transformative. It compounds you. If your underlying capability is weak, AI gives you something different. It gives you fast, mediocre output at scale.

Weak capability multiplied by AI equals fast mediocrity at scale.

This is not a new pattern. There's research from Erik Brynjolfsson at Stanford, going back almost a decade, that shows AI is most accurately understood as augmentation, a multiplier of existing human and organisational capability, rather than a replacement or substitute. The companies that capture AI's value capture it because they have the capability infrastructure underneath. Those that don't, don't.

But 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, this is enough because they have done the real work in their business and just need a tool to leverage it. For others, and this isn't a failing, it leads to using tools and processes that aren't relevant, that distract from doing the work that would actually benefit them. 

The tool sells the promise of the outcome and you get something tangible. The capability layer isn't always visible, so for small businesses, often time and resource poor, the tool seems like the better option at the time.

But once you can see the multiplier model, you can't unsee it. The question hopefully shifts. It stops being "which AI tool should I buy next?" and it starts being "what am I actually multiplying?"

The four capability layers

From the research I've led at 25eight, I consider there to be four capability layers AI amplifies in business, plus two enabling conditions: capacity and community. Three of the four are barely being discussed in public. The fourth, the AI tools layer, is where almost all the public attention has been going. The disproportion is the diagnosis.

Layer one: Human Capability. This is the psychological foundation that you bring to your work. Your mindset. The belief that you can act under uncertainty. Your self-efficacy. Your creativity. Your resilience. The originality that lets you see something other people don't. Without these inner conditions, judgement gets reactive, decisions get fearful, and the team feels it.

This is also the layer that motivational AI content can't reach. AI is not a confidence builder. AI is not a creativity substitute. The AI is going to amplify whatever clarity, or lack of clarity, you bring to it. An unsure business owner who questions their worth will get a very different output from a confident business owner who won't settle for what they believe is the right standard.

Layer two: Leadership Capability. This is how you direct, decide, and delegate inside your business, including the people, the systems, and now the AI agents you are running. It includes your judgement, your decision-making capacity, your delegation discipline, and the self-awareness that lets you tell when your own bias is in the room.

The pattern I see most often: a business owner subscribes to an AI tool, gets value from it personally, tries to roll it out to the team, the team uses it badly, and the owner concludes the AI doesn't work. Almost always, the actual problem is upstream. The owner has not decided, clearly, what the AI is for. Has not articulated, clearly, how the team's work is meant to change with it. Has not designed, clearly, the operating rhythm that makes adoption inevitable rather than optional. That is not an AI problem. That is a leadership capability problem. The AI is exposing it.

There's foundational research from Andy Bharadwaj at MIS Quarterly going back to 2013, before the current AI wave even started, showing that the prerequisite for digital advantage isn't the technology itself. It is the leadership capability to align technology choices with strategic intent and to lead the operational change required to capture the value. Westerman, Bonnet, and McAfee at MIT made the same finding through different research a year later in Leading Digital. The capability layer underneath the technology determines whether the technology pays off.

Layer three: Business Capability. This is broader than a lot of people assume. It includes the strategy, the vision, the mission, the customer understanding, the brand, and the problem the business solves, alongside the operational integrity, the systems, the data, the workflows, and the financial management.

A quick test for this layer. If I walked into your business tomorrow and asked your team what the strategy is for the next twelve months and how it shapes the work they're doing this week, would the answers be coherent and consistent across the team? What if I asked you that question? In many growing businesses, the honest answer is "I don't know" or "there's a vague idea that's been communicated once" but a lot still lives in the founder's head or is assumed that everyone knows.

Strategy lives mostly in the founder's head, if there is a strategy at all. Data lives across three CRMs (or forms of) and two note-taking apps. The operating rhythm is intuitive rather than designed. All of that worked when the business was small enough to fit inside the founder's head.

The moment AI is brought in at scale, that infrastructure has to harden. AI loves clean inputs. Faithful execution against unclear strategy and partial data produces faithful chaos at scale. The AI does exactly what you asked it to do, applied to the data you gave it, and you get back the same business you had, just faster.

I feel this myself. My data cleaning habits have to be impeccable. Files saved in the right place, a clear operating system, roles and responsibilities, no room for ambiguity, or the success rate decreases.

Layer four: Digital and AI Capability. This is the layer the public conversation focuses on. It is the technology skill itself: prompting, workflow design, model selection, knowing when not to use AI, recognising when AI has produced something wrong. It matters. But it sits on top of the other three layers. It cannot compensate for weakness underneath.

If you find yourself thinking, "I just need the right AI tool, the right prompt, the right course," can I get you to pause and consider whether the capability gap might actually be elsewhere? The volume of free, high-quality AI training online right now is extraordinary. If digital capability alone were the bottleneck, the gap would already be closing.

What AI cannot create

There is one category of capability AI cannot create. It can amplify it. It can support it. And it cannot manufacture it.

That category is embedded human capability. Capability that lives in the body. Capability that shows up reliably under pressure. Capability that survives the actual conditions of running a business.

Reading about leadership doesn't make you a leader. Knowing about strategy doesn't make you strategic. Watching a course on resilience doesn't make you resilient. Chris Argyris made this point precisely in his 1991 Harvard Business Review essay, 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's The Knowing-Doing Gap extended the case at the organisational level: smart organisations know what to do and do not do it because the conditions for doing have not been built. Brinkerhoff's work on training transfer found that, in unsupported environments, only around 30% 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 can't be downloaded. They can't be prompted. They can't be outsourced to a model.

This is what I mean when I talk about the Capability Era. Information has become essentially free. AI has commoditised it. Anyone can know almost anything almost instantly. The competitive advantage of knowing things has collapsed to almost zero.

What is rising in value, in inverse proportion, is the embedded capability to act, to decide, to lead, to build, under real conditions, in your specific context.

I find this exciting because it shifts the playing field. Smaller businesses now have access to the same knowledge, the same tools, and the same insights as larger enterprises. What used to be enterprise advantage has been democratised. Much like the printing press in the 15th century empowered people to think freely, challenge old authorities, and print in everyday languages rather than only Latin, AI is democratising knowledge. The competitive advantage of AI moves to the layer that hasn't been democratised: human capability, built deliberately, applied to the specific conditions of your business.

A question worth considering

If ‘ Capability X AI = Output’, and AI is now roughly the same for everyone, then your output is going to be determined, almost entirely, by your capability.

Can I get you to consider: where is your capability strongest, where is it weakest, and what are you doing about it?

If you're not sure of the answer, and you want a way to get an honest representation of where your capability is sitting, we have a complementary diagnostic that may help. The 25eight Small Business Capability Gap Diagnostic takes about ten minutes, and gives you a snapshot of where your strongest leverage is sitting, where your capability gaps are, and your AI readiness across our four capability domains plus our two enabling conditions: capacity and community.

Make sure you get the most return on any investment you put into your business, time or money, by finding your focus through the diagnostic, then making the next decision from there. Link in the comments.

The conversation we should be having about AI in 2026 isn't about the tool, the prompt, or the agent. It's about the human capabilities behind them.

Map your capabilities for free at https://www.25eight.co/the-small-business-capability-gap