Blog

Build or Buy? AI Has Changed the Question

Spend any time on LinkedIn at the moment and you will run into the argument: why pay for software when AI can build exactly what you need in an afternoon? A few years ago that would have been a silly question. Today it is a fair one and the answer matters a great deal to any firm that holds client data.

Why we always bought

For most of the life of business software, the decision made itself. When a firm needed a CRM, a document system or a workflow tool, it went out and bought one. Building your own meant hiring developers which was out of reach for nearly every practice.

What you see when you buy is the feature list but most of what you actually pay for sits underneath it. The provider keeps the product running, ships new features, patches vulnerabilities, stores your data and maintains the integrations with the other tools you use. Making sure your login page is not the weak point is somebody else's full-time job.

What changed

AI started out as a personal productivity aid. It drafted an email, tidied a file note or reviewed a letter before it went out. Useful, but contained to one person and one task.

It has moved well beyond that. Agentic workflows can carry a process across several steps without a person driving each one. AI coding tools can turn a plain English description into working software. Someone with no development background can describe the fee calculator they have always wanted and have a working version before lunch.

That is what sits behind the build argument. For the first time, a firm can realistically replace a piece of software it used to buy.

What do you take on when you build your own software?

A tool you build yourself is often excellent on day one. It matches your process exactly, with none of the fields you never use and none of the workarounds.

Day one is the easy part. Your requirements change. The systems your tool connects to release updates and change how they talk to each other. A build that fit perfectly in March is out of date by September and keeping it current is now your job. Nobody is shipping fixes for it overnight.

Security is the bigger issue. If you ship an internal tool you do not fully understand, there are questions you may not be able to answer:

  • Where the data lives: Which servers hold your client records and which country are they in?
  • How it travels: Is the data encrypted in transit and when it is stored?
  • Who can get in: How are users authenticated and how are their passwords kept?
  • Where it is shared: Which outside services does the tool pass data to along the way?

A software provider answers all of these for you. When you build, they are yours to answer. For a firm holding tax file numbers, bank details and health information, that is a lot to take on for the sake of a nicer screen.

When does it make sense to build your own tools?

None of this means you should never build. It means being choosy about what.

The safest candidates are tools that run locally on your own machine: a fee calculator, a reporting dashboard or a model you use to visualise the business performance. Many of these are what developers call client-side tools, where the code runs in your browser or on your computer with no server behind it. That alone is not enough, because code in your browser can still send data elsewhere if it is written to. Provided the tool keeps everything on your machine and does not store or process data anywhere external, a number of the security questions above fall away.

Should you build or buy? Buy the platform then build on top

The option we think is better again sits between the two. Buy a system then build on top of it.

Many modern products like monday.com, HubSpot and Pipedrive give you the baseline: the data model, the hosting, the user logins, the security and the maintenance. They carry the risk, you then build the custom pieces your firm needs inside the platform, using its AI tools or your own.

The detail that matters is where the custom app runs. Shipped inside your own tenant, it works off the platform's data model and sits behind the platform's security. A custom app built inside monday.com that uses only client-side code and makes no calls to outside services never processes data externally. It acts as the logged-in user, so it can see and do only what that person is already allowed to.

You get the screen, the calculation or the workflow the off-the-shelf product was never going to give you. The vendor still does the heavy lifting on security and upkeep.

Where this is heading

We expect this middle approach to grow quickly over the next few years. The build versus buy debate assumes you have to pick a side. Most firms will not need to. They will buy the foundation then build the parts that make it theirs.

Before you build anything, ask two things: where will the data live and who fixes the tool when it breaks? If the answers are “I'm not sure” and “me”, stop there. You are about to take on the job a software provider used to do for you.

Let’s talk

Ready to automate your business?

Discover how tailored automation can free up time, reduce errors and give your team capacity to focus on clients. Start with a no-pressure conversation about your business and goals.