Andrew Siemer, Founder & CEO of Inventive Group, firefighter, veteran. We are a software product team that gets stuff done, brilliantly.
The loudest version of the buy-versus-build conversation right now is also the least useful one: AI made custom software cheap, so SaaS is in trouble, and every company should start building its own tools.
I don’t think that’s right. AI did change the economics of software, but not evenly. It did not suddenly make every category of SaaS irrational, and it did not erase the advantages of vendor scale, compliance programs or integrated platforms.
What it changed most was the cost of building the layer around a business: the dashboards, internal tools, workflow interfaces and decision-support systems teams live in every day. The core system underneath those workflows is still often worth buying. The workflow wrapped around it increasingly is not.
The old answer was usually ‘buy.’
For a long time, the buy-versus-build decision was mostly settled by cost. Even when off-the-shelf software fit badly, most companies still bought it because the alternative was too expensive to justify. That logic made sense. SaaS vendors spread development cost across many customers, and buyers accepted awkward workflows as the price of avoiding a custom build.
What AI changed was not the desirability of fit. Companies have always wanted software that matched how they actually work. What changed was the price of getting closer to that fit. From what I see, the cost of building useful software at the edges of the business has dropped enough that old assumptions deserve a second look.
The edge moved faster than the core.
There is a meaningful difference between replacing a core system of record and replacing the rough workflow layer sitting on top of it. Systems of record still carry real weight. They hold the historical data, sit inside dense integration webs, support audit requirements and often serve as the operational source of truth for multiple teams. Replacing them is still expensive, risky and, in many cases, unnecessary.
The friction I see is in the layers teams touch all day: the extra screens, awkward handoffs, narrow tasks buried inside broad software and workflows that almost fit but never quite do. That is where business-specific nuance matters most, and where vendor generalization often fits least.
AI changed the cost of building that layer much faster than it changed the cost of replacing the core beneath it.
That distinction is already showing up in the market. In April 2026, The Wall Street Journal reported that large companies were not ripping out platforms like Salesforce, SAP and Workday. They were negotiating harder, customizing more and building smaller AI-driven tools around those systems instead. That is much closer to what I expect than a full-scale replacement cycle.
Breadth used to be a moat. Now it can be a tax.
For years, buyers tolerated mismatch. If a team used a narrow slice of a large product, but that slice solved a necessary problem, the waste was annoying but acceptable. Building a custom alternative would have cost too much.
Now the question is harder to avoid. If your team only needs a narrow slice of capability, and the rest of the product mostly introduces complexity, you are no longer just buying software. You are subsidizing breadth you do not need in order to access the part that matters.
That does not mean every narrow use case should become an internal build. It does mean feature breadth is becoming harder to defend as an automatic advantage.
What should still be bought?
If the platform is your source of truth, carries real regulatory burden or sits at the center of a dense and valuable integration ecosystem, it probably still belongs in the buy column.
Even recent research pushing back on the broad “SaaSocalypse” thesis lands in roughly the same place. A 2026 paper on agentic AI and enterprise software economics argues that the collapse story is overstated, while also finding that building is most compelling for commodity utilities and differentiating custom applications, and much less compelling for regulated and mission-critical systems.
That tracks with what I see. If the software is the ledger, the governed system or the operational backbone, you probably still want to rent it.
What should now be questioned?
The more interesting category is the software layer your team fights with every day: workflow overlays, internal dashboards, task-specific interfaces, narrow decision-support flows and process tools that are highly local to how your business actually runs. These are often the places where a company has adapted its behavior to fit the software, instead of the software fitting the business.
If the real frustration is not in the data model underneath but in the experience wrapped around it, owning that layer may now create more leverage than continuing to subscribe to a generic version of it. In the right cases, teams feel that benefit quickly in speed, usability and process fit.
This is why I think leaders should stop treating buy-versus-build as a philosophical preference and start treating it as a layer question.
Here’s a better test for the next renewal.
The next time a software renewal comes up, the useful question is not whether your company should become anti-SaaS. It is whether you are renting the right layer.
Is this platform your system of record, or is it mostly a workflow wrapper around work you already understand well? Is the pain sitting in the core, or in the interface around the core? Would owning this layer let you operate in a way that generic software never quite will?
I think the companies that benefit most from AI will not be the ones that try to rebuild everything. They will be the ones that finally separate what must be rented from what should be owned.
AI did not eliminate the need to buy software. It made the old boundary lazy. Keep the systems that earn their centrality. Rebuild the friction your team lives inside. That is where the economics moved.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?