Small and medium businesses struggle with fragmented data – Software 3.0 can change that

As organizations grow, their IT infrastructure naturally grows with them. At the beginning, a simple invoicing system, a shared drive with PDF files, and a few Excel sheets may be enough. But over time, as operations become more complex, the company accumulates more tools: sensors, ERP, CRM, databases, production monitoring software, and so on.

At the same time, the logistics of running the organization become harder to control. When a company has five people, everyone more or less talks to everyone else. But a 40-person organization is something completely different. Suddenly, accounting may have little visibility into what production is doing.

Management struggles to maintain a clear flow of information, because the company’s operations are now spread across five departments, each using a different application. To understand what is actually happening, managers either have to log into each application themselves, ask employees for updates and wait for answers, or rely on people’s self-reporting – even though those reports may be incomplete or intentionally misleading.

Over time, different parts of the company start operating with different versions of reality. The data that accounting sees may no longer match the signed contracts or HR records. Mistakes accumulate quietly in the background. Policies become harder to verify. Employees and contractors may begin to exploit the lack of oversight: invoicing at an outdated rate or failing to follow internal company policies.

As long as business is strong and the company is growing, these issues can remain hidden. But when a downturn hits and margins shrink, the need for oversight becomes urgent. Suddenly, someone has to verify that contract rates match invoice amounts and that the numbers across systems are consistent.

The first instinct is to have someone check everything manually across all systems. In some cases, especially when there are only a few contractors or transactions, this can work. But it is boring, slow, repetitive work that consumes valuable employee time. It also gives management yet another process to supervise. And as the number of contractors, contracts, and transactions grows, manual checking becomes less and less feasible.

Naturally, organizations want IT to take care of this. They want management to have a clear view of the company, track margins accurately, and eliminate repetitive manual work through AI and automation. This is exactly what we are doing in the City of Bydgoszcz. There, integrating data from different city departments was largely manual and required two people working for half a year. Unsurprisingly, Bydgoszcz was eager to automate it.

What these organizations really needed was strong data infrastructure.

The SMB data infrastructure gap

Unfortunately, SMBs are famously difficult to build software for. There are several reasons for this.

First, SMBs are highly cost-sensitive. They have less capital available than large corporations, and wasted spending has a much more direct impact on decision-makers. As a result, they are cautious about investing in software unless they know it will solve their problem.

This would be manageable if SMBs did not also require a high degree of customization. Some might call it enterprise-level customization; in many cases, it goes even beyond that. Large enterprises, because of their size, often converge around similar business processes and organizational structures. 

SMBs, by contrast, vary much more. They have more flexible processes, different operating philosophies, and often work in niche markets or combine several business models at once – for example, an e-commerce company that also manufactures some of the products it sells. For SMBs, this business logic is not a side detail, but their competitive advantage.

Unfortunately, customization is a very expensive part of making software. If an existing solution cannot simply be reused, someone has to write and maintain custom code. Data integration becomes more complex as the number of data sources grows, ETL pipelines need to be adapted as the business changes – these costs can grow very quickly.

This leaves SMBs with two common options, and both are flawed.

The first option is to work with IT consultants who can adapt or customize software to their needs. But SMBs are often disappointed with this route. Consultants rarely fully understand the company’s business logic, the work can create significant recurring costs, and the developers assigned to the project are not always the same senior specialists involved in the initial sales conversations. We have spoken with a mid-sized company where a basic two-system integration cost around €12,000 and still failed to address the actual problem. Worse still, fixing it meant hiring even more consultants through the same enterprise-style channels, which pushed the company to look for an in-house alternative.

The second option is to build the solution internally. More and more SMBs now have small IT teams that maintain their internal systems. These teams understand the business logic well, and asking them to build something may seem cheaper at first. But they are not deeply experienced in data infrastructure. As a result, they tend to build systems that are fragile, difficult to scale, and hard to maintain. And because they already have many other responsibilities, they fail to keep up with the ongoing work required. 

In the end, the SMB finds itself back at square one.

This unique combination of high customization needs and limited budgets means that SMBs remain deeply underserved when it comes to data infrastructure.

Why is this changing

The way software is built is changing quickly. Increasingly, the actual creation of code is being offloaded to AI agents – a shift often described as Software 3.0.

AI agents can dramatically reduce the time it takes to create software, especially application-level code. LLMs can already generate impressive prototypes and workflow applications with relatively little effort. But they still struggle when the underlying problem involves many fragmented data sources, complex business logic, or information that does not fit neatly into their context windows. They also hallucinate, which means their code still needs to be checked manually. They cannot reliably maintain complex projects on their own.

This creates both a major opportunity and a new risk for SMBs. Many SMBs already have small internal IT teams that understand the company’s operations and business needs. With AI agents, these teams can now generate far more custom application code than before, much faster and at much lower cost.

They can also “vibe-code” parts of their data infrastructure: generate Python scripts, connect a few systems, and build application logic around them. In the short term, this can work. But over time, it can also create a new kind of technical debt. The infrastructure becomes fragile, difficult to maintain, hard to scale, and expensive to extend as the amount of agent-generated code grows. Agentic AI uses tokens, and tokens cost money. More importantly, without the right foundation, the system becomes harder to understand, verify, and control.

This opens a window of opportunity for data infrastructure designed specifically for the Software 3.0 reality: a world where people prompt agents, and agents build around reliable foundations. There is already growing interest in tools designed for agents to work with directly, shifting more of the software engineering burden away from humans.

The opportunity is to build infrastructure that does not just store and move data, but actively enforces quality: integration standards, maintainability, scalability, provenance, and reliability. In other words, infrastructure that helps agents build the right way from the start.

For SMBs, this is the difference between AI-generated software that works for a week and infrastructure that can support the business for years. Instead of accumulating brittle scripts and one-off automations, internal teams can use agents to build on a foundation designed to keep their systems reliable, scalable, and under control.

That is the real promise of Software 3.0 for SMBs: not just faster code, but durable digital infrastructure for historically underserved companies.