Yoav Kutner is the Co-Founder of OroCommerce, a B2B-focused commerce platform for manufacturers, wholesalers and distributors.
When building gets easier, deciding what to build becomes more important.
I’ve watched how software development has changed for a long time. When I started in technology in the 1990s, a company’s entire IT department could be one person. That person built the software, maintained it and carried most of the knowledge about how it worked.
We may be heading toward a version of that again. AI makes it possible for one person or a small team to create software that once would have required a much larger development team. The barrier to building something new is getting lower, and the number of things we can build is going up.
That makes decisions about what we build and how we build it more important, not less. Here are a few reflections from what I’ve seen so far.
Don’t Make The Business Chase AI
Every new AI capability creates another decision for companies. One model gets better at coding, another at reasoning, another at multimodal tasks. If your business sells industrial equipment, you shouldn’t have to become an AI expert just to keep your technology current.
Your team has a business to run. You only have so much time to spend evaluating new models and deciding what needs to change every time something new comes along.
The technology underneath the business needs to keep pace. Core systems can remain stable while you introduce new AI capabilities where they make sense. The business gets the benefit of better technology without rethinking its operations every few months.
AI will keep changing. The foundation underneath the business has to be built for that reality.
Use AI For What It Does Best
The question, then, is where AI can do the job well enough to trust it.
AI is incredibly good at producing things. Code, text, images, summaries, first drafts. When the task is well defined and the output can be checked, there is a lot less reason to spend human time doing the first pass yourself.
The harder problem is dealing with rules that interact. AI might write a pricing rule in minutes. Ask it to account for customer-specific contracts, volume discounts, product exclusions, regional rules and exceptions that override each other, and the task changes. In research on interlocking rules, even leading models scored only around 50% on the more complex cases.
So, yes, experiment with AI. Just keep in mind that getting something to work is different from figuring out how it should work as part of a larger system.
Know What Is Worth Rebuilding
That raises a bigger question: If AI can help us build individual pieces of software, how much of an enterprise system should we build ourselves?
For some companies, building more of their own software will make sense. If you have the resources and want to find out how far an AI-built enterprise system can go, that is a valid journey. But I don’t think the lower cost of writing software alone is a good reason to start over.
For years, the value of enterprise software was partly in the code. But the code is no longer the scarce part. When you generate a new system, you don’t get everything that happened before it: the years of use, the problems people found, the mistakes that were fixed and the decisions that were tested in the real world.
I’ve seen enough enterprise software to know how much of that experience is invisible until you need it. An AI model can help you build a system, but it can’t give you the experience of learning what breaks, what scales and what customers actually need.
And that’s the risk of building everything yourself. If you build 100% of the system for your own needs, you also get 100% of the mistakes that someone else may have already made and solved.
As building gets easier, the things that are harder to generate become more valuable. In enterprise technology, that means architecture, deep industry knowledge and the ability to customize the software around the business.
Those are the things you can’t shortcut, and they’re easy to underestimate when building new software becomes cheap.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?