Conversational search is the first module of the Cleed.ai Co-Agent. It queries every one of your document sources at once and answers in plain language — citing its sources every time.
Procedures, notes, contracts and references sit scattered across SharePoint, a document management system, an intranet, shared drives and half a dozen business tools, none of which talk to each other. Answering a routine question means asking two colleagues and opening four folders. Worse, expertise leaves with the people who hold it: know-how that was never written down anywhere searchable simply disappears when someone changes role. And when teams turn to public AI tools to fill the gap, your confidential documents go with them.
An employee types a question the way they'd ask a colleague. The Co-Agent queries all connected sources simultaneously — SharePoint, document management systems, internal databases, institutional websites and business APIs — reads the relevant documents, and returns a contextualised answer with a direct link to the original file and its last-updated date. Existing access rights apply at query time, so a user only ever receives answers drawn from documents they are entitled to see. The knowledge base re-synchronises as soon as a source document changes, which means no stale answers circulating months after a procedure was revised. Every query is logged — question, sources consulted, extracts returned — and exportable by an administrator.
Conversational search was first deployed across local government, where the volume of procedural documentation is highest. The same problem exists wherever knowledge is written down and hard to retrieve: legal and procurement teams searching regulatory texts and precedent, corporate teams navigating internal policies and onboarding material, retail teams needing product knowledge and store procedures, education teams handling course material and administrative process.
Existing access rights apply at query time. A user only ever receives answers drawn from documents they are entitled to see — no shadow index, no accidental disclosure, and no need to rebuild your permission model before starting.
The knowledge base updates as soon as a source document changes. No manual re-indexing, and no stale answers still circulating months after a procedure was revised — the failure mode that quietly erodes trust in every internal search tool.
Question, sources consulted and extracts returned are logged for every request and exportable by an administrator — for audit, for quality review, and for identifying exactly where your documentation is missing.
Data hosted in France, with AES-256 encryption at rest and in transit, TLS 1.3, and regular third-party penetration audits. Each organisation gets a dedicated, isolated environment, and no personally identifiable data is passed to language models. Processing takes place exclusively within the EU, with automated deletion after three years, on infrastructure not subject to the US Cloud Act. For organisations with the highest requirements, SecNumCloud 3.2 hosting on S3ns (Thales and Google Cloud), certified by ANSSI, is available as an option, as are Microsoft Azure France and on-premise deployments.
See also: document generationSee also: writing assistance← Back to the AI Co-Agent platform