Domino topples AI model-to-mission mountain

Any models, any agents

The company says its approach works with models and agents, whether hosted in Domino or sourced externally.

“AI is moving into the most consequential work an enterprise does: discovering drugs, pricing risk, underwriting loans, and protecting national security. Coding assistants now let teams build the applications that put AI in expert hands. But to build a mission-critical tool, not merely a toy, those applications must do more than impress in a demo. They have to be governed, auditable, secured, scalable, and integrate with an enterprise’s complex technology ecosystem,” said the company, in a press statement.

That gap, between something that works in a demo and something that runs the business, is where most AI projects stall.

Domino CEO and co-founder Nick Elprin says that Domino is “already the proven platform” for building, operating, and governing AI models and agents in regulated environments.

“Applications define the next era of delivering AI transformation in the enterprise, but as coding assistants make it easier to build these new tools, organisations must find new ways to unlock innovation without making a mess,” said Elprin. “Domino is the best platform for enterprises to build, deliver, and govern the coming wave of AI applications.”

This release extends that foundation to the applications built on top, with the following new capabilities:

• Domino App Hub – A place to develop and deliver AI applications with new capabilities that include rapid previews to accelerate development; version control and staged deployment; approval gating to govern application deployment and review; and Domino Knowledge Manager, a customizable taxonomical organisation system that makes it easy for business stakeholders to find relevant apps at enterprise scale.

NOTE: Domino’s approval-gating capability governs application deployment using automated quality gates, ensuring models meet strict validation, performance, and compliance benchmarks before receiving mandatory stakeholder sign-off for production.

• Integrated Coding Assistants. GitHub Copilot, Claude Code and OpenAI Codex run inside Domino as first-class tools. Because they operate natively on the platform, software engineers can use them to develop, deploy and govern AI work. A built-in library of skills equips a user’s preferred assistant to perform actions that drive the data science lifecycle using Domino’s powerful platform services.

• High Performance Computing (HPC) workload support via Slurm integration. Applications and other workloads in Domino can now use Slurm, ensuring integration with a critical technology interface common in financial services and life sciences organisations.

NOTE: Domino governs application deployment via automated quality gates, ensuring models meet strict validation, performance, and compliance benchmarks before receiving mandatory stakeholder sign-off for production.

• Platform Extensions. A new framework that allows Domino customers and partners to embed their own tools and workflows directly into the Domino interface, tailoring the platform to an organisation’s own workflows and requirements. Partners are already building on it, including Appsilon, whose Axon.R extension validates R packages for life sciences.

When applications are built this way, the impact shows up in the work itself: better drugs in life sciences, sharper risk management in financial services, greater safety in the public sector. This is evidence that speed and governance together turn AI from a research project into the way the business runs.

Domino deep dive

The Computer Weekly Developer Network spoke to Matt Bonyak, principal product manager at Domino Data Lab for more.

CWDN: How does Domino’s approval gating prevent developers from treating critical compliance and AI model validation like an optional side quest?

Bonyak: Domino builds governance into the platform as a structural requirement.

In most development environments, compliance review and model validation are steps that happen alongside the production pipeline and are easily deferred under deadline pressure – they’re essentially honour-bound. Domino’s approval-gating changes that dynamic by embedding governance directly into the app publication workflow. Developers can build freely with Domino’s robust APIs, infrastructure hosting and identity management, but if approval gating is enabled, they cannot push a new app version to production without completing the governed publication process and obtaining required approvals. Domino enforces this as a hard gate built directly into the platform.

Reproducibility ability

What makes this meaningful for regulated enterprises is what sits beneath the gate: reproducibility ensured within each governed version, full audit trails across preview, versioning, and publication, and the ability to run older versions on demand for historical validation. Rather than reviewing a finished artefact in isolation, compliance teams are reviewing a reproducible, traceable state of the application built up across the entire development lifecycle within Domino’s system of record. That is the evidence standard regulated enterprises require at audit time.

CWDN: With every enterprise vendor claiming full-stack governance, how does Domino bridge the gap between demo and mission-critical deployment?

Bonyak: Many platforms can demonstrate governed AI in a controlled environment. The question enterprise buyers in regulated industries are actually asking is: what happens when a model is wrong, needs to be rolled back, or has to survive a regulatory audit two years from now?

Domino’s answer starts with something most competitors cannot replicate: it is already the system of record for how enterprise data science work gets done. Domino embeds governance in the environments where models are built, the workflows where apps are published and the infrastructure where everything runs. Audit trails, model lineage, access controls, and approval workflows are part of the daily developer experience, rather than a separate compliance module supplemented after the fact.

The specific capabilities matter here. Coding assistants operate inside governed environments, so developer velocity does not come at the cost of ungoverned shadow work. App versions in Domino are tied to reproducibility, so teams can start an older version on demand for validation or rollback, whereas present-day practices often rely simply on a documented reference point. And for enterprises running Slurm workloads on HPC infrastructure in air-gapped environments, Domino integrates that existing compute into the same governed platform rather than requiring a migration to a new stack.

This extends to infrastructure more broadly. Regulated enterprises rarely operate in a single cloud. Data residency requirements, regional restrictions, and existing on-premises investments mean that AI infrastructure is almost always distributed. Domino runs data science and machine learning workloads across any compute environment, whether cloud, on-premises, or air-gapped, from a single interface. Teams build and deploy once, without needing to re-architect for each environment or move data across boundaries to accommodate it. Once deployed, Domino manages the infrastructure and scales to support tens of thousands of users across the enterprise.

Every major competitor is optimised for one part of the problem: development, hosting, or regulatory credibility. Domino brings build, scale, and govern together in a single platform, which is what mission-critical deployments in a regulated industry actually require.

The last domino?

Will Domino Data Lab break open the path to real-world AI deployments in ways we had never previously imagined? In places, perhaps, but surely not everywhere. The model-to-mission mountain and monolith will still be there for many organisations and every tech vendor in this space is claiming to offer a “unified approach with full-stack governance” these days, so take those claims with a pinch of salt, or at least with enough scepticism to give yourself some balance. All that said, Domino approval gating capabilities are neatly packaged and (arguably) one of the real antidotes to the challenges ahead – and, crucially, embedding Copilot, Codex and Claude inside one governed environment may at least stop developers from treating compliance like an optional side quest.

Domino’s new capabilities are currently in private preview and will be generally available by Q3 2026.