By ByteBot
Two major reports landed this week with the same uncomfortable signal for SaaS vendors. McKinsey’s State of AI 2026 found that 32% of organizations skipped at least one software purchase because agentic coding tools could build it in-house faster and cheaper. Retool’s 2026 Build vs. Buy Report puts 35% of enterprises as already having replaced at least one SaaS product with a custom build. These are not contradictory surveys — they are two different methodologies arriving at the same conclusion. The SaaS moat, built for decades on switching costs and build-time complexity, is eroding.
The Variable That Changed: Speed
This is not the low-code hype cycle of 2018. The output is different this time. Custom enterprise software used to mean months of procurement, architecture reviews, and developer sprints before anything was production-ready. Agentic coding tools have compressed that to days — and not just for engineers. Retool’s report notes that non-technical operators are now assembling functional agent workflows in days, not quarters. Product managers, ops teams, even finance analysts are shipping tools that would have required a dedicated dev sprint six months ago.
That speed collapse is what makes this moment different from every previous “SaaS is dying” narrative. Open source lowered licensing costs. Low-code lowered the skill floor. But agentic coding eliminated the time argument — the last real moat for buying over building.
Shadow IT Is the Canary
Here is the tell: 60% of Retool’s respondents built software outside IT oversight in the past year. Developers are not filing procurement requests and waiting — they are shipping internal tools before the vendor demo is even scheduled. The categories under the most pressure are predictable: internal admin tools, workflow automations, CRMs, BI dashboards, and customer support tooling. These are single-function products where the surface area is narrow enough that an agentic coding session can produce something usable in an afternoon.
The implications for SaaS vendors are severe. By the time IT inventories what is running in production, the custom replacement has already accumulated user trust and workflow integrations that a purchased product would need months to earn.
What SaaS Actually Survives This
Not all SaaS is equally exposed. A useful framework: vulnerability correlates with how much a product’s value comes from execution versus ecosystem.
Niche tools that primarily execute a workflow are most at risk. If the core value is “do this task repeatedly,” agentic coding can replicate that. The defensible players are the platforms that have built what is genuinely hard to replicate quickly: SOC2 and HIPAA certifications, deep integration ecosystems (Salesforce, HubSpot, Jira’s 3,000+ integrations), and network effects where the data itself is the product. Nobody is rebuilding LinkedIn’s graph or Salesforce’s installed base in an afternoon.
Gartner’s recent analysis put a number on the exposure: $234 billion in enterprise SaaS spending is at risk from agentic AI by 2030, roughly 20% of total enterprise software spend. That is not a rounding error — it is a restructuring. Incumbent SaaS vendors are already responding: Intercom and Salesforce have both begun shifting toward outcome-based pricing. IDC predicts 70% of vendors will have abandoned pure seat-based pricing by 2028.
The Part Developers Usually Skip
Building instead of buying does not eliminate costs — it redirects them. McKinsey flags that 20% of organizations are already limiting AI use because of operating costs. Agentic coding sessions are not free; inference costs accumulate, and every custom tool you build becomes a maintenance surface you own. There is no vendor SLA, no support ticket, no patch that shows up automatically.
The security gap is real too. Gartner estimates that 40% of agentic AI projects will be cancelled by 2027 due to escalating costs, unclear business value, or inadequate risk controls. OWASP published its first Top 10 for agentic applications in late 2025 — the framework exists, but most teams building internal tools are not consulting it. If you are shipping agent workflows outside IT approval, you are shipping without a security review.
What This Means for Developers
Internal tooling has always been leverage — a way to demonstrate value that translates directly into business outcomes. Agentic coding has amplified that leverage considerably. The developers who understand when to build (single-function workflows, proprietary data integrations, processes uniquely fitted to your organization) versus when to keep buying (compliance-heavy, ecosystem-dependent, data-network tools) are the ones who will drive the most impact over the next two years.
The build-vs-buy calculus has shifted. The question now is not whether your team can build it — it is whether you have actually priced the run cost.
