How UST Codon Brings Governed AI Development to SAP

SAP teams weighing AI for application and workflow development have to settle how a generated change reaches a live system safely. Each enterprise platform relies on its own tools and specialist skills, and UST says general-purpose AI assistants lack the platform depth and governance to work safely in a live system.

UST’s answer arrived in September, when it launched UST Codon, a workspace where developers describe an application or workflow in plain English and AI builds it on the platform they name, working from the customer’s own configuration.

SAP is one of five platforms supported, alongside ServiceNow, Salesforce, Oracle Fusion, and Workday, and UST is connecting additional platforms to Codon. No change reaches a live business system without expert review, and governance controls apply to every write.

Building on SAP in Plain Language

A developer states the requirement in plain English and names the target platform, which sends the request to SAP. Codon plans the work, selects the matching governed Model Context Protocol (MCP) server, and drafts the build against the customer’s configuration.

Codon can build configuration, business logic, integrations, extensions, and reporting across SAP. Configuration covers how the system is set up, business logic the rules behind processes, and extensions the custom additions to standard functions. Each platform has its own governed server and software development kit, with the same rules applied.

Codon generates the change directly on the platform through a governed, audited connection. Access defaults to reading, and actions are attributed to the developer where the platform supports it. Role-based routing sends each developer to the surface that fits the work, and developers can work without scripting or switching between platforms.

Human Validation Before Go-Live

A senior UST practitioner reviews and validates every output before anything is promoted. Codon drafts the work, and a person confirms it before it moves on.

That validation comes from UST’s enterprise-platform specialists, and the same practices deliver SAP and other platform work for large clients. The review happens before the work reaches the customer’s environment.

Approval hands the build back to the customer’s process. The validated build moves through the customer’s existing development path and governance path before it becomes a working change in the live system.

Governance and Audit Controls

A set of controls applies to every write on a live system. Codon classifies each tool call as a read or a write and Blocks unknown or ambiguous operations, a behavior called failing closed. Individual developers are recorded as the acting party where the platform supports it, and an approved-build phase and production controls govern deployment.

Destructive actions require explicit intent, and activity goes into a tamper-evident, hash-chained audit trail, which links each record to the one before it so an altered entry breaks the chain. The trail comes with a dedicated security-event stream.

Cost is previewed before an AI-driven build begins. The built-in estimator combines token rates and expert input, so customers see the full build cost upfront. Each customer runs in a single-tenant environment on AWS Fargate with default-deny outbound access, which blocks outgoing connections unless they are explicitly allowed.

UST reports that internal deployments sped platform development by about 50% while all requests were classified, attributed, and audited, and that applications that once took months now ship in as little as a week and at a significantly lower cost.

What This Means for SAPinsiders

Validation sets the pace of AI development. When a workspace drafts changes quickly, the review step decides how fast they reach production. Who staffs that review, and how it fits a customer’s own release process, becomes part of how development with AI is run.

AI governance moves to the moment of action. Classifying each call as a read or a write and recording who acted puts control at the point where a change is made. The audit trail then gives later reviewers a record they can verify.

Governance can be shared across platforms. Enterprises that run SAP next to other platforms need review and audit controls on each one. A common set of controls, applied through each platform’s own connection, gives teams one consistent answer to review and audit across the whole estate.