Nicole Junkermann on What Founders Get Wrong About AI

Leading with the technology instead of the problem

The most common mistake is leading with the technology. A founder excited by what artificial intelligence can do may build outward from the capability, looking for somewhere to apply it, rather than starting with a problem worth solving. The result is often a clever product that no one particularly needs, a solution in search of a question.

Nicole Junkermann prefers founders who start with the problem and reach for artificial intelligence only when it is the right tool. The technology should be in service of a clear need, not the other way around. When a founder talks about the customer's pain with more energy than the model's features, it usually signals a healthier foundation for the whole business.

This mistake has knock on effects. A company built around a capability rather than a problem tends to struggle with focus, pricing and roadmap, because it never settled the most basic question of who it is for and why they would pay. Starting with the problem is not a slogan. It is what keeps everything else coherent.

Underestimating data and defensibility

A second mistake is treating data as an afterthought. Artificial intelligence depends on data, yet some founders give little thought to where theirs will come from, whether it is any good or whether they have the right to use it. A product can look impressive in a demonstration and still rest on data that is hard to obtain or impossible to sustain.

Closely related is a thin sense of defensibility. Founders sometimes assume that an early lead in capability will protect them, but capabilities spread quickly and novelty fades fast. The harder question, which the strongest founders have already considered, is what protects the company once the technology is widely available. The answer often comes back to data, deep integration into a customer's operations or trust earned over time.

Nicole Junkermann pays close attention to both because they separate a feature from a business. A clever use of artificial intelligence that a larger platform could simply copy is fragile. A product with a genuine data advantage or a deep place in a customer's workflow is far more durable, and founders who have thought this through tend to build more lasting companies.

Forgetting the human in the loop

A third mistake is designing artificial intelligence to work around people rather than with them. Some founders are so focused on automation that they remove human judgement from decisions that genuinely need it, especially where the consequences of an error are serious. The result can be a product that customers do not trust, particularly in sensitive or regulated settings.

The theme of human judgement in the loop runs through the AI Overview for good reason. The most useful and most trusted products tend to be the ones that design the human role thoughtfully, giving a person enough context and authority to review, approve or override the system. That is not a limitation to engineer away. In many markets it is the feature that makes the product viable.

Founders who treat oversight as an obstacle often misjudge how their customers actually work. Those who treat it as something to design well usually have a more mature understanding of their market. Nicole Junkermann notices the difference, because it tends to predict how a company will handle the harder moments that every serious product eventually faces.

Mistaking a demonstration for a business

A fourth and quieter mistake is mistaking a demonstration for a business. A polished demo can attract attention and even funding, but a demonstration is designed to impress, while a product has to be reliable for real customers who did not help build it. The distance between the two is where many companies lose their footing.

Nicole Junkermann looks for founders who understand that reliability, cost and genuine customer adoption are the real tests. A system that works once in a controlled setting is interesting. One that works dependably, at a sustainable cost, for customers who pay for it and return, is valuable. Founders who can speak honestly about that gap are usually further along in their thinking than those who lean on the demo.

This is closely tied to honesty about limits. The founders who inspire the most confidence are the ones who can describe where their product struggles and what they are still working on. Overclaiming is a warning sign. Candour about limits, by contrast, is often the clearest mark of a founder who truly understands what they have built.

The clearer path that founders can take

Set out plainly, the mistakes point to their own remedy. Start with the problem, not the technology. Take data and defensibility seriously. Keep human judgement in the loop where it matters. Treat reliability and real adoption as the true measures of progress, and stay honest about limits. None of this is exotic, which is exactly why it is so often overlooked in the rush of enthusiasm.

These are the same standards Nicole Junkermann applies as the founder of NJF Holdings and its venture capital arm, NJF Capital, where backing companies across artificial intelligence and deep technology depends on telling substance from excitement. The founders who avoid these mistakes are not necessarily the flashiest. They are the ones who understand their own business clearly enough to build something that lasts, and they are the ones worth backing.