The wave of personalized software has arrived. Users can increasingly build and use the small applications they need without waiting for a company to turn them into a product and sell them. At the same time, generative AI has lowered the threshold for producing a viable product at an early stage without significant capital. Two old asymmetries are weakening: entrepreneurs’ dependence on capital and users’ dependence on application developers. In the fierce competition created by these mutually reinforcing trends, what should software companies do?

The situation in stock markets is striking. Even strong financial results are failing to support share prices. Microsoft met or exceeded earnings expectations, yet its stock fell by roughly 11 percent, wiping out around $400 billion in market value. Salesforce announced growth, but its shares declined by 5.6 percent. ServiceNow beat expectations, yet its stock fell by approximately 11 percent; SAP met expectations, but its stock dropped by around 16 percent. Despite their acquisitions and AI features, companies once expected to be the biggest beneficiaries of AI are increasingly valued not on today’s earnings, but on what their business models might look like five years from now. AI is putting pressure on per-user licensing: as agents take over tasks previously performed by multiple people, adding more user seats ceases to be a dependable engine of growth.

None of this means software is finished. But horizontal, general-purpose software can no longer expect to grow simply by expanding its feature list. Software development is not just writing code. Understanding needs, designing, testing, deploying, operating, securing, and maintaining systems remain essential. These are precisely the stages where vibe coding often runs into difficulties. Large language models can lose track of context and fabricate solutions. AI workflows also create new attack surfaces through risks such as prompt injection.

For small and midsize software firms, the way forward is to move competition away from code and toward three forms of value: domain expertise, trust, and embeddedness in the customer’s workflow. Rather than selling generic products with ever larger sets of features, firms need to specialize deeply in particular subsectors. They must understand the exceptions, regulations, data definitions, and process realities of those subsectors and embed AI within that understanding. They also need to find ways to move pricing away from per-user licenses and toward value based on outcomes or transactions. The question is not “How many users do we have?” but “How much of the customer’s costs do we reduce, or what additional value do we create?”

The wave of personalized software may be an opportunity to build an ecosystem rather than merely another competitor. If users are going to build their own solutions, could your product provide the secure framework around them? Can you supply the integrations, permissions management, audit trails, version control, rollback mechanisms, and human approvals when needed?

While software firms use AI internally to gain efficiency and speed, what they should sell to customers is not speed alone, but reliability. Code has become cheaper; trust has become more valuable. Rather than competing with low-cost code, software firms should find ways to create and capture a new layer of value by assuming responsibility for risk, compliance, and sustainable operations on their customers’ behalf.