Faster doesn’t mean better
AI is fundamentally reshaping the way developers and engineers design and build products. Historically, software design and interfaces have focused on how a product reaches the customer, which to date has almost always been a human. Now the rules are being rewritten once again, with systems adapting to work for both people and autonomous agents.
For developers, there is little doubt that AI adoption is growing rapidly. In fact, 72% of developers now use them daily.
Coding tasks that once took weeks now take hours or even minutes. The technology has radically reduced the cost of both writing and shipping software, making its production quicker than ever before.
Yet, more software does not necessarily mean greater impact. Previously, the bottleneck was mechanical: writing code and implementing features simply took a lot of time. Now, the constraint has shifted. Teams may be building faster with AI , but they are not always building better.
Scaling wrong decisions, faster
AI doesn’t just accelerate software development; it accelerates decision-making. Engineering constraints once acted as a natural filter, as proposing a new feature or product involved significant engineering effort, time and skill, which forced teams to prioritise and invest carefully.
However, AI removes much of that friction, allowing organisations to test, iterate, and deploy new products at unprecedented speed. Whilst this can provide vast opportunities, it also means bad ideas are produced more quickly and poor user experiences reach customers faster. The gains from increasing the speed to market are overshadowed by the burdens of unnecessary features, overlapping products and increased technical debt.
For example, imagine an online retailer wanted to make customer journeys easier and more seamless, to improve sales conversion. Since AI makes development so fast, un-coordinated teams launch an AI product recommender, AI shopping assistant, AI search experience and an AI buying guide. Confusing, right? Instead of feeling guided throughout the purchasing process, customers become overwhelmed with options and unsure of which tools to use or how. This feeling means they are less likely to enjoy their experience, less inclined to investigate products, and more likely to be lost to a sea of competitors with more streamlined offerings. Not only that, but any changes made to product data must be integrated into all the different AI experiences, increasing the workload for teams and eating into the time savings that AI coding tools generated.
Historically, bad ideas were limited by the time and effort required to execute them, encouraging thoughtful product design. Today, bad ideas spin up quickly and persist before their flaws become clear, at which point it may be too late.
More products, same usage
Despite a surge in software creation driven by AI, user engagement has not kept pace. MIT research found that whilst AI coding tools dramatically increased coding activity (by up to 180%) and led to new apps being created, this failed to translate into higher overall app usage. Large task-level AI productivity gains translated only partially into shipped and used software. Evidently, faster software production does not automatically generate more value for the customer, but can instead result in wasted business investment and effort.
This disconnect highlights a fundamental imbalance, in that the supply of software is growing rapidly but user attention and need remains finite. Simply building and creating more is no longer enough for businesses.
Customer experience quality is the real metric
As AI reduces the cost and speed of building software, the ability to measure its impact and determine what to build or improve next becomes increasingly valuable. Organisations need to move beyond simply deploying features and instead focus on understanding customer behaviour in real time. Which changes improve retention, and which create friction? Which experiences make an impact on the customer, and drive measurable business gains?
AI should be used to understand, then improve the customer experience. Again, take the example of our online retailer. This time, instead of using AI to develop multiple conflicting tools, the business builds a carefully considered shopping agent to answer product queries, share recommendations and help customers navigate around the website. AI is used to speed up its development once decided, but the process of making the decision is clear and thoughtful.
The harder question is what happens next. Are customers really using the shopping assistant, or abandoning it after one interaction? Do the recommendations increase basket size, or distract customers from making a purchase? Have product recommendations truly improved conversion?
Without measuring customer behaviour in real time, the retailer is out of the loop. However, the right insights and analytics enable it to identify which experiences perform best, genuinely improving engagement, retention and revenue. Winning companies will be the ones that learn the fastest, and use learned insights to make better product decisions that can then execute quickly with help from AI.
From dashboards to continuous intelligence
As AI makes its mark on the software development process, the way teams consume and react to product insights will naturally evolve too. Rather than relying solely on dashboards and manual analysis, organisations will use AI to autonomously surface anomalies, identify behavioural trends and proactively recommend actions.
The future of product development will be rooted in enabling continuous insights to understand how customers respond to products – and using this to improve at the pace AI now allows. Businesses will increasingly take customer behaviour insights and use them to refine products and systems, removing those that interrupt the customer journey. Our online retailer can create a continuous feedback loop, fine-tuning how customers interact with its AI shopping assistant for maximum satisfaction.
More haste, less speed
Software teams looking to get ahead must treat the technology as more than just a tool to build faster and create more. Approaching AI with haste allows time for strategic value to direct you, instead of drowning in features nobody wants to use.
AI has solved the problem of building and refining software, but in doing so has exposed another question: How will my product meaningfully improve the user experience?
