o9 Solutions

The Enterprise Knowledge Graph connects internal and external data across supply chain, commercial and financial systems, named by category rather than by product, and the agents recommend actions and guide execution. No named integration catalog, custom tool support or authenticated action framework is documented; agent action in outside systems is not asserted, and execution is guided inside the platform.

Sourceo9 Solutions, o9solutions.com/digital-brain and /ai-innovationread 2026-09-08

Composite agents integrate a sequence of atomic agents, each retrieving, synthesizing and acting on specific inputs, to complete cross-functional planning exercises inside the Enterprise Knowledge Graph, shipped July 2024 and extended by the specialized planning and execution agents of June 2025; the AI Innovation page names advanced orchestration as a current architecture property. That is multi-step, multi-agent execution under o9's own composite layer. No workflow designer of the BPMN kind, and no named runtime beyond the composite layer, is documented.

Sourceo9solutions.com/news/o9-enhances-its-digital-brain-platform-with-generative-ai-powered-composite-agents-to-execute-complex-cross-functional-planningread 2026-09-08

The patented Enterprise Knowledge Graph is a persistent, maintained structure connecting internal and external data, business relationships and tribal knowledge into a system of record of digital knowledge that the agents reason over, described on the AI Innovation page as providing enterprise context with advanced retrieval, and on the Digital Brain page as a living digital twin.

Sourceo9 Solutions, o9solutions.com/ai-innovation, /digital-brain and AI agents release 2025-06-03read 2026-09-08

o9's AI Innovation page names a flexible-autonomy spectrum from human-in-the-loop to touchless execution, symbolic enterprise guardrails on the agents, and policy controls, and the agents page has them recommending actions and guiding execution rather than acting unilaterally. The human in the loop mode and the guardrails are stated, but no review and approval surface, approval gate or escalation mechanism is documented, and the touchless end of the spectrum is o9's stated direction.

Sourceo9 Solutions, o9solutions.com/ai-innovation and AI agents release 2025-06-03read 2026-09-08

ISO/IEC 27001, ISO/IEC 27017, SOC 1, SOC 2, TISAX and C5 appear on o9's own Security page under a Compliance Certifications and Frameworks heading without hedge, alongside GDPR, CCPA and NIST CSF 1.1 as frameworks, plus a BitSight rating and a vulnerability disclosure policy.

The same page names single sign-on, domain claiming, enterprise mobility management support and administrative access control as platform features, and the AI Innovation page names role-based access, auditability and policy controls for the agentic layer.

No auditor, report date or trust portal is published, the heading mixes certifications with regulations, and RBAC is a single phrase on the AI page rather than a documented permission model.

Sourceo9 Solutions, o9solutions.com/security and /ai-innovationread 2026-09-08

The agents answer o9's four questions, including what happened and why, through post-game analysis and causal root-cause diagnosis; the AI Innovation page names explainability by design and auditability among the governed-and-secure properties, and the Security page names system monitoring, logging and alerting at the infrastructure level. Decision explanation is surfaced and audit is asserted, but no agent run trace, decision log surface, retention period or reconstruction artifact is documented; explanation of a recommendation is not a record of what an agent did.

Sourceo9 Solutions, o9solutions.com/ai-innovation and /securityread 2026-09-08

The AI Innovation page calls the Enterprise Knowledge Graph the platform's memory, logic and enterprise context, but it is the knowledge graph, an application data structure over the customer's data, not an agent memory layer with a stated scope and lifetime. Composite agents continuously learn from feedback, but that learning is absorbed into the agents rather than kept as stored memory. No session, user, team or workflow-scoped agent state with a lifetime is documented.

Sourceo9 Solutions, o9solutions.com/ai-innovation and composite agents release 2024-07-01read 2026-09-08

No product hosting region, region list, customer environment option or selection surface is published on o9's legal pages or Security page. The January 2026 privacy policy is website-scoped, places the website's hosting in the Netherlands, names group offices in eleven countries and covers transfers with SCCs or binding corporate rules; the Security page lists ISO 27017 and C5 as cloud-control attestations, which concern cloud security rather than data residency. The platform is described throughout as cloud-based delivery with nothing for a buyer to choose.

Sourceo9 Solutions, o9solutions.com/privacy-policy section 4 and /securityread 2026-09-08

The June 2025 release ships specialized AI agents for management, planner and front-line roles across 30-plus industry verticals with industry-specific knowledge models, and the AI Innovation page describes them as domain-aware and purpose-built for planning, execution and learning. The agents come in named role families rather than a browsable catalog; no agent is named individually and no template library or marketplace is documented.

Sourceo9solutions.com/news/o9-is-raising-the-bar-for-enterprise-planning-and-executionread 2026-09-08

Described as an always-on digital representation of the enterprise, the platform continuously senses changes, simulates outcomes and recommends actions, and the agents continuously sense, analyze and recommend, with real-time risk and opportunity visibility as one of the four questions they serve. Work reaches the agents on continuous signal rather than on request. No schedule, webhook, inbound queue or channel surface is separately documented, and the sensing claim is made in platform prose rather than through a documented trigger mechanism.

Sourceo9 Solutions, o9solutions.com/digital-brain and /ai-innovationread 2026-09-08

o9's AI Innovation page discloses that its AI architecture supports multiple LLMs, and the 2024 composite-agent release describes atomic agents coupled with LLM systems. The multiple models are a capability on o9's side; no customer or admin model selection surface, named provider list or routing control is documented. The disclosure is one line in a list of architecture properties, with no provider named.

Sourceo9 Solutions, o9solutions.com/ai-innovation and composite agents release 2024-07-01read 2026-09-08

No API, SDK, MCP server or developer portal for calling o9 from outside is documented. Low code configuration is for building inside the platform, not for outside callers driving it. The o9 Academy is training rather than documentation, and the AI Innovation page's self-service configuration and extensibility line names no interface.

Sourceo9solutions.com/ai-innovationread 2026-09-08

No evaluation harness, scored test cases, release gate or controlled post-deployment experiment measuring the agents is documented. Post-game analysis measures the customer's plan against execution and diagnoses forecast inaccuracies, so what is tested is the plan, not the agent, and the self-learning models that improve every cycle have no documented evaluation loop. The productivity and forecast-accuracy figures are outcome marketing. The product optimizes the customer's operations rather than testing the agent.

Sourceo9 Solutions, o9solutions.com/ai-innovation and AI agents release 2025-06-03read 2026-09-08

The agents sense, analyze and recommend over the Enterprise Knowledge Graph and guide execution inside the Digital Brain; no browser, desktop or remote computer control is documented.

Sourceo9 Solutions, o9solutions.com platform pagesread 2026-09-08