Most businesses rent SaaS products that solve part of the problem.
The tool is useful enough to justify the subscription, but not specific enough to fully match how the business actually works. So the team fills in the gaps with side spreadsheets, duplicate data entry, manual approvals, exported reports, Slack threads, and tribal knowledge. That hidden work has always been expensive. What is changing now is the economics of doing something about it.
Harvard Business Review recently put it this way: "Generative AI is dissolving the economic logic that made standardized enterprise software the only practical choice for most companies."1
That does not mean SaaS is going away, and it does not mean every company should start rebuilding payroll, accounting, email, CRM, payments, or ticketing systems. There are many places where renting software still makes sense, especially when the workflow is standard, the vendor is carrying compliance burden, or the business benefits from broadly accepted best practices. But not every workflow is standard.
Where the gap actually shows up
Take a proposal workflow. A company may rent a quoting or proposal tool, but the real work still happens around it. Sales pulls language from old proposals. Finance checks margin. Operations confirms capacity. Technical people answer feasibility questions. Leadership approves exceptions. Someone eventually reconciles what was promised against what can actually be delivered. The SaaS product handles the document, but the business is still carrying the workflow.
A reporting workflow has the same problem. The dashboard exists, but the team still exports data, cleans it up, combines it with spreadsheets, adds context from Slack or email, and turns it into something leadership can actually use. The rented tool provides visibility into part of the system, but people are still doing the connective work manually.
Customer support has the same shape. The ticketing system stores the request, but the real value depends on pulling customer history, order context, prior escalations, product details, internal policy, and judgment about what should happen next. If that context lives across several systems, the team still spends time assembling the picture before they can do the work.
What AI changes
AI-enabled engineering makes it faster and cheaper to design, prototype, test, and maintain narrow systems around specific workflows. AI can also be built directly into those workflows: summarizing context, flagging missing inputs, drafting first passes, comparing exceptions against known rules, preparing approval briefs, or helping route work to the right person.
The opportunity is not to replace every SaaS product. It is to ask where the business is paying for a partial fit and then paying its own people to close the gap every day. AI does not make ownership free — custom systems still need design, maintenance, security, governance, and human judgment. But it does make ownership more reachable for workflows where the gap is expensive enough to matter.
That is the build-vs-rent question worth revisiting. Not because SaaS is bad, but because the work around the tool may now be more solvable than it used to be.