This might suggest that custom software is becoming less important. If technology is becoming more accessible and off-the-shelf software increasingly includes AI capabilities, why invest in software built specifically for your organisation?
For many processes, there is no need to. If a standard CRM, ERP or other SaaS solution does what an organisation needs, custom software is not automatically the better choice.
It becomes a different story when processes are unique, multiple systems need to work together, or technology is part of what differentiates an organisation.
And that is exactly where AI does not reduce the need for custom software. It may even increase it.
Business processes remain unique
Many organisations use the same types of systems: CRM, ERP, planning software, customer portals or a combination of SaaS solutions. Yet no two companies operate in exactly the same way.
The way an order is processed, a customer is onboarded, a product is configured or a service is delivered has often evolved over many years. These processes contain business knowledge, customer agreements, exceptions and sometimes an important part of what differentiates the organisation.
AI does not change that.
An AI model can help process information, recognise patterns, analyse documents or automate certain tasks. But it does not automatically determine how an organisation wants to operate.
Imagine using AI to automatically recognise, extract and classify incoming invoices. The technology to do this is becoming increasingly accessible. But then the company-specific part begins: which information needs to be verified, which exceptions require human approval, which financial system should receive the data, and which subsequent actions should be triggered automatically?
The AI capability may be generic. The process around it often is not.
AI makes technology more accessible. The real challenge is still making that technology work for the organisation.
AI does not solve integration problems
AI rarely operates in isolation. It becomes part of an existing IT landscape.
And that landscape rarely consists of a single system.
CRM, ERP, financial software, portals, databases, external APIs and legacy applications all need to exchange information. Data may be spread across different sources, definitions may differ, and not every system provides modern integration capabilities.
Adding AI on top of that environment does not automatically remove this complexity.
In fact, when AI becomes part of actual business processes, reliable integrations and good data become even more important.
Consider the same invoice processing example. AI can recognise the supplier, amounts and products in a document. But if that information then needs to be checked against an ERP system, linked to a purchase order and sent to financial software, an AI model alone is not enough.
The value emerges when everything works together.
That requires software and integrations that ensure information reaches the right place at the right time.
Software remains the foundation
AI is therefore primarily a new capability within software. It does not replace the software around it.
An AI model can analyse a document, make a prediction, classify information or generate a proposal. Something still needs to happen with the result.
Who is allowed to see it? When is human review required? Where is the result stored? Which system should perform the next action? What happens when information is missing or an integration is temporarily unavailable?
These are not AI questions. They are software questions.
Architecture, security, data, integrations and user experience therefore remain important, especially when AI is no longer used simply as an assistant but becomes part of operational and business-critical processes.
AI makes custom software faster to build
AI is, however, fundamentally changing one aspect of custom software: how we develop it.
Development teams can use AI to support writing and reviewing code, analysing existing software, creating tests and documentation, and exploring potential solutions.
Work that previously required a significant amount of manual development can therefore be completed faster.
But speed is not the most important consequence.
More interesting is the ability for teams to move from an idea to a working solution faster. An initial version can be tested with users earlier, assumptions can be validated sooner, and adjustments can be made earlier in the development process.
This shifts some of the focus from producing code to making the right decisions around that code.
What should we build? Which components can remain standard? Where does custom development actually add value? How do we connect existing systems? And where can AI genuinely improve a process?
Ultimately, those questions determine the quality of the solution.
Not everything needs to be custom
This does not mean that every organisation suddenly needs more custom software simply because AI is becoming more important.
For many processes, standard software is the best choice. There is little value in rebuilding functionality that an existing solution already handles well.
The decision becomes more interesting where standard software stops fitting the organisation.
That could be a specific business process, a combination of different systems, a large amount of manual work or an AI application that depends on an organisation's own data and processes.
Custom software does not have to mean building everything from scratch either. It can be a relatively small software layer that connects existing systems, standard software and AI.
That is increasingly where the interesting opportunities are.
From building software to building the right software
The development of AI does not make software development less important.
Some of the technical work is becoming faster and more accessible. At the same time, decisions around processes, data, integrations and architecture are becoming more important.
The question therefore shifts from simply how do we build this? to what should we actually build?
For organisations with largely standardised processes, the answer may still be an existing SaaS solution. For organisations with specific processes, complex integrations or ambitions to make AI part of their operations, custom software may play a bigger role.
Not because everything needs to become custom, but because technology only creates real value when it fits the way an organisation works.
As technology becomes increasingly accessible, the difference will ultimately not be access to AI.
It will be how effectively you apply technology within your own organisation.
Wondering where standard software stops and custom software starts adding value?
We are happy to explore your processes, integrations and the role AI could play in them.
