AI Agents Move Deeper Into Enterprise Workflows

Coding Agents Compress Enterprise Launch Timelines

Stampli, a B2B payments platform, used OpenAI's Codex and ChatGPT Work to condense what would have been weeks of launch production into days, according to an OpenAI case study. The constraint was real: a fixed deadline with design resources already committed elsewhere. The result illustrates a pattern practitioners are increasingly reporting — AI coding agents acting as force multipliers when human capacity is the bottleneck, not the idea itself.

Separately, Replit announced Free Mode powered by GPT-5.6 Luna, enabling users to build working software without token-cost friction. While the Stampli case targets existing enterprise teams, Replit's move signals an intent to broaden the population of people who can ship software, potentially expanding the pool of agent-assisted builders significantly.

Data Privacy Policy Gets Concrete: OpenAI's Zero Data Retention

OpenAI has reaffirmed Zero Data Retention (ZDR) for eligible API customers, meaning inputs and outputs are not stored after a request completes. The announcement also previews Private Safety Processing, a mechanism designed to allow advanced safety evaluations without requiring access to customer data in identifiable form. This directly addresses a tension that has slowed enterprise adoption: organisations in regulated industries need safety guardrails but cannot expose sensitive data to third-party retention.

For AI product and engineering teams, the practical implication is that ZDR-eligible deployments can now be positioned more credibly in procurement conversations involving legal, compliance, and security stakeholders. The Private Safety Processing preview is worth monitoring as it matures, since it attempts to decouple safety infrastructure from data custody — a non-trivial architectural challenge.

Ecosystem Expansion: Partner Networks and Open-Source Harnesses

Anthropic's Claude Partner Network has added adesso SE, a European IT services and consulting firm, to its directory on partnerhub.claude.com. Partner network growth is a lagging but meaningful indicator of enterprise readiness: it signals that system integrators see sufficient customer demand to formalise a practice around a model provider.

On the open-source side, OneCLI (YC S26) launched on Hacker News as an OSS sandboxed agent harness built for teams. The project, hosted on GitHub, focuses on providing a controlled execution environment for agents — addressing the containment and reproducibility concerns that often block teams from moving agent prototypes into shared or production environments. Together, these two items suggest the agent middleware layer is filling in from both the enterprise-partner and developer-community directions simultaneously.

Key takeaways

• Stampli's use of Codex and ChatGPT Work demonstrates coding agents as a viable deadline-pressure tool for enterprise launch teams.
• Replit's GPT-5.6 Luna Free Mode extends agent-assisted software creation to a broader, non-enterprise audience.
• OpenAI's Zero Data Retention reaffirmation and Private Safety Processing preview offer a concrete path for regulated-industry API adoption.
• adesso SE joining Anthropic's Claude Partner Network reflects growing SI investment in Claude-based enterprise practices.
• OneCLI's open-source sandboxed agent harness addresses containment and reproducibility gaps that slow team-level agent deployment.

Sources

• How ChatGPT Work helps Stampli move ideas to market — OpenAI
• adesso SE | Claude Partner Network Directory - partnerhub.claude.com — Anthropic
• Offering Zero Data Retention for frontier models — OpenAI
• Launch HN: OneCLI (YC S26) – OSS sandboxed agent harness for teams — Hacker News
• Replit expands access to software creation with GPT-5.6 Luna — OpenAI