AI-generated article. This article was researched and drafted using AI tools and published automatically, and its featured image was generated by AI. Facts are drawn from the sources cited in the text.
For years, the advice to a small business owner was simple. Do not build software. Buy it. Someone else has already solved the problem, they keep it running, and a monthly fee lets you skip the hard part. That logic held for a long time. It is starting to wobble, and the reason is the AI build vs buy question now landing on businesses that never thought of themselves as software companies.
A third of companies stopped buying and started building
McKinsey’s State of AI 2026 survey, published on August 25 and drawn from 1,719 respondents across 97 countries, found that 32 percent of organizations decided against buying at least one off-the-shelf software product because they could build it themselves with agentic coding tools. In the tech sector that figure climbs to 41 percent. Retool’s 2026 Build vs. Buy report, based on 817 builders, points the same direction. 35 percent of teams have already replaced at least one SaaS tool with something they built, and 78 percent expect to build more custom internal tools this year.
What changed is speed. AI-assisted development has pulled the cost of a small internal tool down from months of engineering work to, in some cases, a few days. For a business paying a full subscription just to use one or two features of a product, that shift is worth noticing.
The part the headlines skip: someone has to run it
Building is not the finish line. It is the start of ownership. When a business buys software, the vendor carries the maintenance, the security patches, the uptime. Build the same thing in-house and all of that moves onto your side of the table.
Those costs are real, and they tend to show up late. Analysts have started naming a new category, comprehension debt, which accumulates when people ship AI-generated code they do not fully understand. One widely cited review of AI-generated codebases found that a large majority contained at least one critical security vulnerability. Gartner has projected that 40 percent of agentic AI projects will be canceled by 2027, often because the running costs climb past whatever was saved at the start. The pattern many teams describe is quiet. Year one feels like a win. Year two the maintenance bill arrives.
What the AI build vs buy shift means if you are not a software company
Most small businesses are not about to replace their accounting software with a homemade version, and they should not. The build or buy decision is rarely all or nothing. The more useful version of the question is smaller. Is there one repetitive job, a report copied out by hand every week, a form that feeds a spreadsheet, that a simple custom tool could handle, while everything else stays bought?
There is a governance edge worth naming too. Retool’s report flags that much of this building is happening as shadow IT, outside any oversight. That matters more for a small team than a large one, because there is no separate department to catch a quietly built tool that ends up holding customer data.
A first step that is not building anything
A reasonable place to start is not code. It is a list. Write down the software the business pays for, mark which features actually get used, and notice where a whole subscription is covering one small job. That list often answers the build or buy question before anyone writes a line of code. For the jobs that do look buildable, keeping a person on review and writing down who owns the tool once it exists is what separates something useful from a liability. Projects that skip that step are the ones that tend to stall before they ever reach daily use.
The falling price of building is genuinely good news. It hands small businesses options they did not have two years ago. It also moves a responsibility onto them that used to sit with a vendor. So the question worth sitting with is not whether a business can build its own tools now. Increasingly, it can. The question is which ones are worth owning for the next five years, and which are better left to someone whose full-time job is keeping them running.
