The Infrastructure Trap: Why Your AI Investment Strategy Might Already Be Obsolete
Most AI spending is infrastructure spending dressed up as strategy. The businesses doing the buying feel progressive. The businesses doing the selling are the ones getting rich.
That gap matters more than any capability the tools actually deliver. Before you sign another SaaS contract or brief your team on a new AI workflow, you need to know which side of that transaction your business actually sits on. Because the research is starting to separate the categories clearly, and two of them end in the same place: you spent the money and someone else kept the margin.
The Three Categories That Actually Exist
Forget the broad claim that AI helps everyone. It does not. McKinsey's 2024 State of AI report puts measurable profit impact in roughly 5% of companies globally. That is not a technology failure. That is a structural one. The gains concentrate in businesses that meet a specific set of conditions, and most businesses do not.
The three categories are producers, distributors, and consumers. The order matters because it determines where value accumulates.
Producers build the models, own the training data, or sell access to the underlying infrastructure. Distributors wrap that infrastructure in a workflow and sell it to a vertical. Consumers buy the finished tool and use it to do existing work slightly faster. Most Irish SMEs are consumers. A growing number of mid-sized firms think they are distributors but are actually consumers with a prettier dashboard.
Irish companies are already losing ground in this race because the consumer position feels like participation when it is actually dependency.
The Four-Question Audit
Run this before the next budget conversation. The order matters because each question builds on the previous one.
Step 1: Where does your data live, and who owns it?
If your business generates data that nobody else can replicate, you are closer to the producer end of the spectrum. A logistics company with ten years of last-mile delivery patterns for Irish rural routes owns something a model cannot easily substitute. A recruitment agency using a generic AI screening tool owns nothing except the subscription fee.
Step 2: Does AI reduce your cost base or reduce your price?
This is the binary that most strategy conversations skip. Cost reduction is a producer move. It protects margin. Price reduction is a consumer move made under competitive pressure. It hands margin to the buyer. If your instinct is to use AI to undercut a competitor on price, you are not building an advantage. You are starting a race to the floor that a better-funded competitor will win faster.
Step 3: Are you training anything, or just prompting?
Prompting a generic model is using a tool. Training a model on your own data, your own customer interactions, your own product knowledge, is building an asset. Tools depreciate. Assets compound. A Dublin-based professional services firm that fine-tunes a model on ten years of its own client work has something defensible. A firm using the same off-the-shelf chatbot as its three nearest competitors has a shared feature.
Step 4: What would happen to your competitive position if the AI vendor raised prices by 40%?
If the answer is "we would pass it on" or "we would absorb it", you are a distributor with pricing power or a consumer with a cost problem. If the answer is "we would rebuild on a different model", you have genuine architectural flexibility. Most businesses have never asked this question. The vendors have.
Where the Trap Actually Springs
The infrastructure trap is not buying bad tools. It is buying good tools and mistaking the purchase for a strategy.
The €720 million figure attributed to failed or stalled AI projects in Irish business is not a story about technology that did not work. It is a story about investment that went into the consumer position and produced consumer returns, which is to say, marginal productivity gains that did not move the revenue line.
Picture it this way. A mid-sized Irish accountancy firm spends €80,000 implementing an AI document processing system. It saves four hours a week across the team. That is roughly €12,000 in recovered time annually if you value it generously. The vendor who sold the system is scaling that same product across 400 firms globally and compounding the return. The accountancy firm bought efficiency. The vendor built a business.
That is not a criticism of the accountancy firm. It is a description of the consumer position. The question is whether they knew that going in, and whether they made the decision with both eyes open.
What the Distributor Position Actually Requires
The middle category, distributor, is where most of the realistic opportunity sits for Irish businesses that are not building foundational models. But it requires three things that most AI strategy conversations ignore.
First, a defined vertical. Horizontal AI tools are the producer's territory. A Cork-based construction firm that builds an AI estimating tool trained on Irish planning codes, local material costs, and fifteen years of its own project data is doing something genuinely hard to copy. A generic estimating tool reskinned is not.
Second, proprietary feedback loops. Every customer interaction has to make the model smarter in a way that benefits your business, not the underlying vendor. This requires thinking about data ownership at contract stage, not after deployment.
Third, a pricing model that reflects scarcity. If your AI-powered service is priced identically to your previous service, you have not changed your market position. You have changed your cost structure while leaving the value capture on the table.
The Turn
None of this argues against buying AI tools. It argues against buying them without knowing which category you are operating in. The consumer position is not fatal. It is just honest. Some businesses should be consumers, run the tools well, and compete on execution. The mistake is funding the consumer position while telling yourself a producer story.
The audit is four questions. It takes an afternoon. It is worth more than most strategy days.
Know which category you are in, and build deliberately from there rather than spending your way toward a position you have not actually earned.
