Formic AI unveils Boreal, an explainable model for firms

Canadian artificial intelligence start-up Formic AI has launched Boreal, an explainable language model that it says produces verifiable and audit-ready outputs for enterprise users.

The Toronto-based company is targeting organisations that have slowed or cancelled generative AI projects because of concerns over accuracy, regulatory exposure and dependence on foreign vendors.

Formic AI positions Boreal as a locally developed alternative for Canadian businesses that want stronger control over data, traceable decision-making and clearer evidence behind machine-generated answers.

Focus on verification

Boreal restructures unstructured text such as documents and records into a structured knowledge graph. It then uses that graph as the basis for answering queries.

The system checks each answer against the original source materials before it displays a response. It logs the reasoning steps in an audit trail so reviewers can trace how the result was produced.

Formic AI contrasts this with conventional large language models that rely mainly on probabilistic next-word prediction and that often operate as opaque systems.

"Trust comes from being able to check the answer and the path that produced it," said Daniel Escott, Chief Executive Officer at Formic AI. "Boreal gives teams a clear link back to source documents and a simple audit trail, so decisions are grounded in verifiable evidence rather than guesswork."

The company is aiming Boreal at high-stakes professional environments that face regulatory, legal or compliance obligations. These include sectors where traceability of information sources and clear accountability for automated outputs are becoming mandatory.

Hybrid technical approach

Boreal operates on a structured, queryable graph rather than only on free-form text prediction. Formic AI describes this as a shift from opaque inference towards evidence-based interpretation, with the knowledge graph acting as a backbone for reasoning.

The model uses deterministic checks that block unsupported statements before they reach end-users. It also records the route taken through the source corpus, which creates a traceable path for auditors or internal reviewers.

Formic AI has combined different AI techniques inside the product. The company separates analytical steps from natural language drafting functions.

"Boreal pairs a neuro-symbolic engine for precise, auditable analysis with a generative fluency engine for drafting and conversational tasks," said Varun Ranganathan, Chief Technology Officer at Formic AI. "That hybrid design keeps day-to-day work efficient while preserving the ability to show how every result was assembled."

The company says Boreal uses a precomputed graph-traversal method that reduces the computing required for retrieval and integration. It argues that this approach supports more economical deployment and aligns with efforts to cut the energy footprint associated with AI workloads.

Governance and control

Formic AI has framed Boreal as an AI system that can sit within strict governance and security boundaries. The model supports on-premises and air-gapped deployments for organisations that do not want data to leave their own infrastructure.

It can also act as a control layer around existing tools that rely on other language models. In that configuration it adds source grounding, applies checks that aim to prevent unsupported content, and provides an audit trail. This allows enterprises to retain current workflows while inserting an additional verification step.

The company is pitching Boreal to boards and technology leaders who face what it describes as a "Crisis of Trust" in AI. Many organisations remain cautious about using black-box systems that they cannot fully explain to regulators, customers or staff.

Industry analysts have warned that a large share of early "agentic AI" projects risk cancellation because of escalating costs, unclear business value or inadequate risk controls. Vendors that can demonstrate explainability and local accountability are attracting interest from risk-averse buyers.

Academic partnership

Formic AI has also disclosed an academic collaboration linked to the new model. The company has partnered with York University's Connected Minds programme on a research project centred on Boreal.

Through the programme's Prototyping Award, Formic AI has received funding and non-financial support such as commercialisation assistance. The firm says this relationship will support ongoing research into explainable AI and structured knowledge representations.

Formic AI presents itself as a Canadian AI platform that focuses on transparent citations and verifiable answers based on an organisation's own knowledge archives. It has identified regulated industries and legal teams as core customer segments.