Buy vs Build Just Flipped

For as long as enterprise software has existed, the default answer to “how do we solve this?” has been “buy something.” Need a CRM? Salesforce. Project management? Jira. Delivery orchestration? Find a vendor. The reasoning was sound: building custom software was expensive, slow, and risky. Buying was faster, safer, and came with support.

That reasoning just broke.

The Numbers

Retool’s 2026 Build vs. Buy Report landed last week. The headline: 35% of enterprises have already replaced at least one SaaS tool with custom-built software. 78% plan to build more in 2026.

That’s not a trend piece. That’s a structural shift, already underway.

The cost of building custom software has collapsed. That’s a structural change, not a cyclical one.

— David Hsu, Retool CEO

The rest of the data is just as stark:

• 60% of respondents built software outside IT oversight in the past year
• 51% have shipped production tools currently in use by their teams
• Half of those report saving 6+ hours per week
• 78% expect to build more custom internal tools in 2026

The shadow IT stat is the most telling. Developers and teams aren’t waiting for procurement cycles. They’re building what they need and shipping it before anyone signs a purchase order.

Why “Buy” Was Never Really “Buy”

The enterprise software industry sold a convenient fiction: buy our product and your problem is solved. But anyone who’s lived through a Salesforce implementation, a Jira migration, or a ServiceNow rollout knows the truth.

“Buy” was never buy. It was:

• Rent at increasing annual rates
• Customize for months with consultants
• Integrate with your existing systems at enormous cost
• Accept the vendor’s workflow, even when it doesn’t match yours
• Lock in to a platform where your data lives on someone else’s terms

The total cost of ownership for enterprise SaaS has always been far higher than the sticker price. But the alternative: building custom: was even more expensive. So you paid Salesforce $150/seat/month, hired an implementation partner, and spent six months configuring something that was 70% of what you actually needed.

What Changed

The “build” side of the equation collapsed. The “buy” side didn’t.

SaaS pricing, integration complexity, and vendor lock-in stayed the same. But a two-person team with agentic coding tools can now scaffold a custom replacement in days. Not a polished product. But a working tool that fits your workflow exactly, integrates with your systems natively, and costs you nothing per seat.

The categories getting replaced first aren’t the big platforms. They’re the narrow workflow tools that were always 80% generic and 20% custom:

• Admin panels and dashboards: The stuff that used to require a dedicated internal tools team
• Approval flows and workflows: The automation layer you used to buy from ServiceNow
• CRMs: The customer data layer that Salesforce charges $150/seat for
• Project management: The boards and tickets you pay Monday, Linear, or Jira for
• Customer support tools: The ticketing systems with per-agent pricing

AI made the 80% free. It made the 20% fast. And suddenly, the value proposition of paying a SaaS vendor to give you a generic version of what you need looks a lot weaker.

The Seat Compression Problem

The market isn’t just pricing in build-vs-buy. It’s pricing in something more fundamental: seat compression.

The entire SaaS revenue model depends on humans sitting in chairs using software. Per-seat, per-month. But when AI agents do the work that humans used to do, the seats empty. If 10 agents can do the work of 100 people, you don’t need 100 Jira seats. You need 10. That’s a 90% revenue hit for the same work output.

Atlassian’s Q3 earnings showed enterprise seat count declining for the first time in the company’s history. The stock dropped 35% on the report. Not because revenue was bad: it grew 23%. Because the leading indicator turned negative. Fewer seats means fewer renewals means the growth story is over.

This is why my post on PM tools designed for humans hit a nerve. Standups, sprint planning, retros, Confluence: these are compensation mechanisms for human cognitive limits. Agents don’t have those limits. They don’t need the tools that manage them.

I’ve Been Living This

I can’t share specifics, but I’ve watched this play out in my own projects. In one case, we built a custom delivery orchestration platform to replace an incumbent SaaS vendor. In another, we built Flux: a Kanban board designed for AI agents that replaces the need for Monday, Linear, or GitHub project boards.

Both followed the same pattern: the SaaS tool was 70% of what we needed, the remaining 30% required constant workarounds, and the total cost of ownership (licensing + customization + integration + training) exceeded what it would take to build a custom solution that fit perfectly.

A year ago, “build” would have been a multi-month project requiring a dedicated team. Today, with agentic coding tools, the initial scaffold takes days. The integration and hardening still takes weeks. But the total investment is a fraction of the annual SaaS spend it replaces.

When “Buy” Still Wins

Intellectual honesty matters. There are categories where buying still beats building:

• Compliance-heavy systems: SOC 2, HIPAA, PCI-DSS compliance is expensive to build and maintain. If a vendor already has the certifications, that’s genuine value.
• Network-effect platforms: Slack, GitHub, and similar tools where the value comes from who else is on the platform, not just the features.
• Deep domain expertise: Accounting software (QuickBooks, Xero), payment processing (Stripe), and similar tools where decades of domain-specific edge cases are baked in.
• Security infrastructure: Auth, encryption, and identity management where getting it wrong has catastrophic consequences.

Nobody’s ripping out Salesforce wholesale. But the bar for purchased software just got permanently higher.

— David Hsu, Retool CEO

The shift isn’t “build everything.” It’s “stop defaulting to buy.” Every SaaS purchase should now face the question: “Could we build this faster and cheaper than the annual subscription cost?” For an increasing number of tools, the answer is yes.

What This Means

If you’re a decision-maker evaluating software:

• Challenge every renewal. Before auto-renewing that SaaS contract, ask what it would cost to build a custom replacement. The answer changed dramatically in the last 12 months.
• Start with the narrow tools. Don’t try to replace Salesforce. Start with the admin panel, the dashboard, the approval workflow. The wins compound.
• Account for the real TCO of “buy.” Add up licensing, customization, integration, training, and the ongoing cost of working around features that don’t fit. Compare that to the build cost, not just the sticker price.
• Watch the shadow IT signal. If your developers are building unauthorized tools to avoid the official SaaS stack, that’s not a governance problem. It’s a market signal. The tools aren’t serving them.

The buy-vs-build calculus flipped because the cost of building collapsed while the cost of buying stayed the same. Corporate procurement processes, designed for an era when building was hard, haven’t caught up. The teams that recognize this first will have better tools, lower costs, and faster iteration. The ones that don’t will keep paying enterprise rates for 70% solutions.