Intelligence work rarely fails for want of information. It fails when the thread between the question and the answer gets cut.

A requirement is raised in one system, collection is tasked in another, the material lands in a repository, the analysis happens in a document, the review happens in email, and the report goes out as an attachment. Each step is defensible on its own. Together they lose the connective tissue. Six months later the report still exists, and nobody can say which version of which source supported the key judgement, who signed it off, or what was actually authorised for release.

CORTEX exists to keep that thread intact. It manages the intelligence lifecycle end to end: requirement, collection, retained evidence, analysis, reporting, controlled release and feedback, in one governed workspace rather than six tools and a convention.

The question stays attached to the work

Direction records what needs to be known, why a decision depends on it, who owns it, when it is needed and where the gaps are. Collection then hangs off that: sources, source-use authority, plans, tasking, attempts and results all stay linked to the requirement that caused them. Collection is not treated as a separate activity that happens to produce material.

Material is preserved before it is processed, and the original is kept apart from anything extracted from it. Evidence references point at the specific passage or the specific cells an analyst used, so an assessment can be reopened against the version of the material that actually informed it rather than whatever the source looks like now.

Analysis is structured around hypotheses, assumptions, supporting and contradictory evidence, gaps, probability and confidence. The judgement is explicitly human. CORTEX is built to support an analyst, not to stand in for one.

Reprocessing must not rewrite history

Original material, extraction revisions and evidence references are all retained, so running a source through again does not quietly change the evidential basis of an assessment someone has already acted on. A reviewer can walk backwards from a judgement to the evidence, from the evidence to the retained original, and from there to the requirement that started it. Products are versioned and come out as real PDF, DOCX or structured JSON with that lineage intact.

A draft is not a release

Writing a report and authorising its release are separate acts. A release binds one exact revision to its handling restrictions and its intended audience, approval is independent of authorship, and changing either the content or the audience requires a new decision. Nor does preparing or downloading a package count as receipt: a named recipient acknowledges the exact package before any feedback is recorded against it.

That feedback goes back to Direction, where someone accountable decides whether the requirement is satisfied, reopened, retasked or turned into further work. It is a cycle rather than a publishing pipeline.

Deployment follows the environment, not the vendor

CORTEX runs on-premise, in private cloud, or in an approved public-cloud deployment. The core application, storage and processing can all sit locally. An offline policy disables documented external capabilities, and an air-gapped profile is available where physical isolation is required, subject to assurance of the intended environment. Connected deployments can enable approved feeds, mapping services and model endpoints under organisational policy. The intelligence workflow should not depend on a public cloud service being reachable.

AI assists; people decide

Model assistance is optional and sits on top of the product rather than underneath it. A local model can run on the CORTEX node, or an approved deployment can use a configured remote endpoint. It can produce an editable first draft from bounded, authorised data, with generation provenance kept for review. Interpretation and editing remain the analyst’s; review remains a human decision; release remains independently controlled. The core lifecycle works with no cloud AI at all.

Built with the doctrine in mind

The terminology and workflow follow Joint Doctrine Publication 2-00 on intelligence, counter-intelligence and security support to joint operations: requirement-led direction, accountable collection, all-source evidence, explicit uncertainty, review, production and dissemination. CORTEX does not replace tradecraft or organisational authority. It is the environment in which those responsibilities can be exercised, recorded and audited afterwards.

What you end up with

The requirement, the source authority, the original information, the retained evidence, the judgement, the report revision, the release decision, the acknowledgement and the feedback all remain part of one traceable lifecycle. It is not another document repository. It is built around the work that produces the documents.

CORTEX is available to explore at cortex.aic.io. For a demonstration or to discuss an evaluation, get in touch at commercial@aicuk.ltd.