Which Software Projects AI Makes Cheaper to Build

Enterprises are starting to act on cheaper software development. In a 2026 McKinsey survey, 32% of respondents said their organizations had decided against buying at least one software product or feature because they could build it internally with agentic coding tools. The savings are uneven, though. AI mainly reduces the cost of writing code, which is one stage of building software, and the stages around it now carry a larger share of the effort.

Code Is Getting Cheaper to Write

Coding agents can draft, test, and revise code from a stated requirement, cutting the engineering hours between a business request and a first working version. That lowers the entry price of a build project: the amount a team must spend before it knows whether the software works.

Use of these tools is now routine in professional development. AI-assisted coding has become a daily practice for a large share of developers, and AI-generated code makes up a growing portion of what teams commit to production.

Some teams have gone further and replaced purchased tools with custom builds. Most of the early evidence comes from teams already set up to build, so it shows where the change starts. How far it spreads beyond those teams is still unclear.

Verification Now Carries More of the Cost

Cheaper code generation does not reduce every hour in a build project at the same rate. Developers report limited confidence that AI-generated code is functionally correct, and many find that reviewing it takes more effort than reviewing code written by colleagues. That review work has to happen before the software can be trusted in a business process.

The effect on project economics follows from that. Writing hours fall, but review, testing, and requirement definition take up a larger share of what remains. A build estimate that cuts coding time without adding verification time understates the real cost.

Which SAP Projects Benefit Most

The largest savings go to projects where requirements are clear, outputs are easy to test, and interfaces are documented. In SAP landscapes, that typically means integrations over released APIs, side-by-side extensions, and internal workflow tools on SAP BTP. Projects that touch complex business logic or poorly documented custom code keep more of their verification burden.

SAP is adding coding-agent support to the tools where these projects are built. SAP’s September release of the Cloud Application Programming Model added a plugin for building AI agents directly from CAP services and exposing them through the Agent-to-Agent protocol. SAP developer sessions have covered using coding agents to build web apps with SAP’s design theme. SAP positions Joule Studio as its environment for building enterprise agents, applications, and agentic workflows. On the legacy side, SAP’s Q2 2026 release highlights list its S/4HANA custom code migration agent as generally available.

These changes apply to producing software. Maintaining, securing, and governing it once it runs is a separate cost.

What This Means for SAPinsiders

• Revisit projects shelved on development cost. Integrations and internal tools rejected in earlier budget cycles may now clear the business case. Reassessing that backlog can surface value that was previously out of reach.
• Shift build estimates toward review and testing. Coding hours may shrink while verification work grows. Estimates that reflect this give business cases a more accurate cost baseline.
• Prioritize projects with clear, testable requirements. Integrations over released APIs and well-defined workflow tools keep verification effort low. Starting there captures the largest share of the savings.