What OutSystems actually shipped at ONE 2026
At the OutSystems ONE Conference in Amsterdam in early June 2026, OutSystems unveiled the Agentic Systems Platform, positioned as an evolution of the low-code platform into a full multi-agent enterprise runtime. The platform is built on top of the OutSystems Enterprise Context Graph — the company's metadata, identity, and lineage layer for the data and process objects that already flow through the low-code platform — and ships with three explicit core domains:
- Agentic Systems Engineering — the developer-facing surface for building, testing, and shipping agents, with Claude Code, Codex, and Kiro integrated as first-class agentic coding tools. Developers can work in the IDE they prefer (Claude Code, Codex), publish into the OutSystems runtime, and inherit the platform's existing governance and observability without re-implementing it.
- Agentic Enterprise Orchestration — the runtime-orchestration plane that coordinates multi-agent workflows, manages handoffs between agents, enforces governance and access policy across the agent fleet, and surfaces the audit trail to the existing OutSystems observability layer.
- Agentic Industry Solutions — pre-built agentic solutions for regulated-vertical use cases (financial services, public sector, healthcare, manufacturing), positioned as the higher-leverage starting point for customers in those verticals.
The operational details worth flagging:
- Open and BYO model: the platform is positioned as model-agnostic, with the customer choosing the underlying LLM provider per agent class and the platform handling the routing, observability, and governance plumbing underneath.
- Early access in Q2 2026, with the agentic coding, publishing, and platform-extensibility surfaces going live at the ONE conference itself.
- The Enterprise Context Graph as the governance plane: every agent, every tool call, every data object the agent touches, and every identity that authorizes a call is represented inside the Context Graph, which means the audit trail and the access-control surface are unified with the existing low-code platform's compliance posture rather than being a parallel system to maintain.
- Integration with existing OutSystems lifecycle management: agents are deployed, versioned, and governed through the same release-management pipeline that the customer already uses for low-code applications, rather than through a separate AI-specific operational surface.
The positioning is unambiguous. OutSystems is not shipping a chat interface. OutSystems is shipping a managed multi-agent runtime with low-code's existing governance, identity, and audit primitives wrapped around it, deliberately targeting the regulated-enterprise buyer who has historically chosen low-code precisely because of the governance posture.
Why low-code-as-agent-runtime is structurally significant
The headline temptation is to read this as another enterprise software vendor shipped agent tooling. Every major enterprise platform vendor has shipped some variant of that announcement in 2026 — Salesforce Agentforce, SAP Joule Studio 2.0, ServiceNow's agentic workflow, Microsoft Copilot Studio, Workday's agentic platform. The pattern is unmistakable: the vendors who own the enterprise system-of-record install base are racing to make agents a standard component of the platform layer rather than something the customer integrates from outside.
Low-code is structurally different from the system-of-record vendors in two ways that matter for how an agentic platform gets adopted.
The low-code install base is the install base that has historically priced governance highest. The reason a regulated enterprise buys OutSystems instead of building applications from scratch is governance — the platform's built-in identity, access control, audit, lifecycle management, and compliance reporting are the value proposition. That same buyer profile is the buyer profile that has been most resistant to the download a tool and start building agents narrative that has driven the AI specialist firm's market share in the last two years. The buyer that wouldn't let the AI specialist firm wire LangGraph directly into the production SAP instance will let OutSystems wire it in, because the governance posture is already there.
The low-code platform owns the metadata layer that the agentic platform needs. An agent that can credibly drive a multi-step enterprise workflow needs access to the data and process objects the workflow operates on, and to the identity and authorization surface that controls who can do what. In a vanilla integration, the AI specialist firm builds that metadata layer as a custom integration — usually expensive, usually brittle, usually a multi-quarter engagement. In the OutSystems case, the metadata layer is the Enterprise Context Graph, which already exists, which already covers the platform's existing applications, and which gets extended to cover agentic workloads as a platform feature rather than as a custom build. The integration cost that has been the dominant line item in regulated-enterprise agentic AI projects gets compressed materially.
The compression is meaningful. The work that an AI specialist firm would have priced at six to nine months of engineering for a regulated-enterprise deployment — building the metadata graph, wiring the identity surface, integrating the audit pipeline, standing up the multi-agent orchestration plane — gets shipped as platform plumbing. The remaining work — the eval discipline, the rubric authoring, the senior-review queue calibration, the workflow-specific agent design — does not disappear, but it is a different fraction of the total project, and the procurement-conversation shape moves accordingly.
What changes about the make-vs-configure conversation
Four shifts that follow when low-code becomes a credible agentic platform for regulated enterprises.
The make-vs-configure boundary moves. Work that was previously make — build the integration with the system of record, wire the audit trail, stand up the identity-aware tool-call surface, instrument the observability layer — becomes configure inside the low-code platform. Work that was previously configure — write the simple form-driven application, expose a CRUD interface on top of a database — becomes the agent does it directly. The boundary between we need an AI specialist firm to build this and our existing OutSystems team can configure this shifts upward. The CIO who walks into the next budget cycle without recognizing the shift will keep paying specialist-firm rates for work that is now in scope for the platform team.
The partner ecosystem economics rebalance. The OutSystems partner network is large, deep, and concentrated in regulated verticals. With the Agentic Systems Platform shipped, that network becomes a serious distribution channel for enterprise agentic AI work. The AI specialist firms that have built their books of business on we are the team that can wire LangGraph into your SAP instance find that the wiring is now platform-native. The specialists that retain pricing power are the ones whose value proposition is the eval discipline, the rubric authoring, the workflow design, and the senior-review queue calibration that the platform vendor cannot supply — not the ones whose value proposition is the integration plumbing.
The procurement conversation gets a credible incumbent option. A regulated-enterprise buyer evaluating agentic AI capability for the first time used to have a binary procurement choice: engage an AI specialist firm to build custom or wait for the system-of-record vendor to ship something. With OutSystems shipping, the buyer has a credible third option: deploy on the low-code platform we already operate, configured with the agentic capability the platform now ships, governed by the Enterprise Context Graph we already trust. That is a meaningfully different procurement shape, and it is the procurement shape that the CFO will find easiest to defend at the audit committee.
The lifecycle discipline gets a built-in answer. Agentic AI in production has, until this quarter, mostly been treated as a research-style operational artifact — deployed by a small team, owned by a small team, maintained by a small team, with the audit trail and the lifecycle management improvised. The Agentic Systems Platform brings the agent inside the existing OutSystems lifecycle — versioning, release management, environment promotion, rollback, audit — and that is the kind of operational discipline that the enterprise platform team has been quietly insisting on for two years without getting a vendor answer. The teams that adopt the discipline early get a meaningfully more reliable operational posture; the teams that defer it will keep operating agents the way they operated their first machine-learning model in 2019, with the same predictable consequences.
What this does not change
Three honest caveats.
It does not eliminate the eval discipline at the workload-specific boundary. A platform that ships agentic primitives is not a platform that ships agents that work on your workload. The eval discipline that grades the agent honestly on your business data, the rubrics that govern the senior-review queue, the gold sets that calibrate the multi-judge agreement — those still have to be built, and the platform vendor cannot build them. The teams that deploy the platform without the eval discipline get the failure mode that has defined the first wave of enterprise agentic AI: a capability that looks impressive in a demo, that decays in three weeks, and that no one trusts in six.
It does not collapse the model-vendor portability question. OutSystems ships BYO model, which is the right architectural choice. The customer still has to decide which models to deploy, how to route between them, how to grade them against the workload, and how to handle the inevitable relative-capability shifts as the underlying model landscape moves. The portability story has to be designed in, even when the platform is BYO-model in principle, because the integration glue between the platform's orchestration plane and the underlying model endpoints is where the soft coupling will accumulate if no one is watching.
It does not eliminate the senior-judgment work at the rubric boundary. Multi-agent orchestration is a hard distributed-systems problem, but the agent makes a confidently wrong decision and the system has no way to know problem is a senior-judgment problem, not a distributed-systems problem. The senior-review queue, the rubric authoring, the calibration discipline, the multi-judge agreement protocol — all the human-in-the-loop primitives that decide whether an agent's decision is trustworthy at production scale — are not in the platform. The customer that signs the procurement contract expecting the platform to supply senior judgment will pay for the missing piece on the back end, after the first production failure makes the cost visible.
Where Sonnet Code fits
A low-code platform shipping an enterprise-grade agentic runtime with native governance and audit primitives is the easy half of the regulated-enterprise procurement story. The hard half is the engineering and human-judgment discipline that turns the platform is configured into the agents are doing what the business needs, evaluated honestly, observed continuously, and governed by a senior-review queue calibrated for the failure modes that actually occur. AI development at Sonnet Code is the engineering half: extending the OutSystems Agentic Systems Platform's BYO-model surface with a routing layer that treats Claude Opus, MAI-Thinking-1, Gemini 3 Pro, and the open-weights frontier cohort as peer endpoints with workload-specific selection; wiring the Enterprise Context Graph audit trail into the customer's existing compliance observability surface; designing the multi-agent orchestration patterns that work inside the platform's governance posture rather than around it; and building the cost-per-successful-task attribution per agent class and per workflow so the platform's promise can be defended with numbers. AI training is the human-judgment half: the senior engineers, the domain experts, the bilingual reviewers who design the gold sets for the agentic workflows, calibrate the senior-review queue against the failure modes a regulated-enterprise workload actually produces, author the rubrics that the eval harness runs against, and serve as the senior-judge pool whose calibrated decisions make the difference between a platform that works and a platform that demos well.
The boundary between custom AI engineering and standard configuration work just moved inside the low-code box for the largest install base in the regulated-enterprise segment. The teams that recognize the shift early and rebuild the eval, governance, and senior-judgment disciplines on top of the new platform will compound a meaningful operational advantage through 2027. The teams that keep treating agentic AI as a custom-build specialist project will keep paying specialist-build rates for work that is now in scope for the platform team — and will keep losing the procurement conversation to the buyer down the road who figured it out a quarter earlier.
