Why Problem-First Beats Technology-First

The technology industry has a persistent habit of starting with the technology and working backwards to the problem. A new capability emerges — machine learning, blockchain, large language models — and vendors race to apply it to every conceivable use case. Some of those applications are genuinely useful. Many are solutions looking for problems.

The problem-first approach inverts this. Instead of asking "what can we do with this technology?", you ask "what problem does this person have, and what is the best way to solve it?" Sometimes the best solution involves cutting-edge technology. Sometimes it involves a well-designed spreadsheet. The right answer depends on the problem, not on the technology du jour.

Consider bookkeeping for sole traders. The problem is clear: people need to record their income and expenses accurately, categorise them correctly, and submit tax returns on time. The best solution for this problem involves a combination of bank data integration (established technology), automated categorisation (AI/ML), deterministic tax calculation (explicit arithmetic), and HMRC API integration (standard web services). None of these technologies are individually exciting. Together, applied to a clearly defined problem, they produce a product that genuinely helps people.

A technology-first approach to the same problem might start with "let us build an AI that does your bookkeeping for you." The demo would be impressive — talk to the AI, describe your expenses, and it categorises everything automatically. But the reality would be a system that gets categorisation wrong 15% of the time, cannot guarantee correct tax calculations, and requires constant human verification to ensure accuracy. The technology is impressive. The product is unreliable.

The problem-first discipline is harder than it looks. It requires genuine understanding of the user's situation — not just what they say they want, but what they actually need. Users often describe their problems in terms of existing solutions: "I need better spreadsheet templates" when what they actually need is "I need to stop using spreadsheets for bookkeeping." Problem-first thinking looks past the stated requirement to the underlying need.

It also requires the discipline to choose boring technology when boring technology is the right answer. Using a simple conditional statement to determine a tax rate is less impressive than using a machine learning model, but it is guaranteed to be correct. Problem-first thinking chooses correctness over impressiveness every time. This is not anti-technology — it is pro-outcome.

The products that endure are the ones that solve real problems effectively, regardless of the technology they use. Nobody remembers which technology powers their favourite tool — they remember whether it works. The bookkeeping software that correctly calculates your tax and submits it on time is valuable whether it uses AI, a database, or a team of trained pigeons. The technology is the means. The solved problem is the value.

At neart.ai, every product starts with the problem. Who has this problem? How do they currently solve it? What is inadequate about their current solution? What would a genuinely good solution look like? Only after answering these questions do we consider which technologies to apply. This discipline means that our products are sometimes less technically exciting than competitors' demos — but they work reliably in the real world, which is all that matters.

The AI era makes problem-first thinking more important, not less. The temptation to apply AI to everything is strong. The results of undiscriminating AI application are products that are impressive in demos and disappointing in daily use. The antidote is relentless focus on the problem, with technology as the servant of the solution rather than the master of the product roadmap.