Current interaction
What a person has supplied in this conversation or request.
TECHNICAL PAGES
Knowledge & Context
The quality of an agent's work depends heavily on what it knows about the situation — and on its ability to recognise what it does not know.
Case context
The relevant client, job, workflow or prior-event facts that frame the work.
Organisational knowledge
Approved procedures, policies, service information and reference material.
Authoritative evidence
Reliable source material the agent can retrieve when the task requires it.
Grounding prevents confident guesswork.
Grounding means working from relevant, approved information rather than filling gaps with plausible-sounding assumptions. In professional accounting environments, an agent should be able to distinguish supplied facts, retrieved evidence, assumptions and missing information.
A good design makes uncertainty visible. Where evidence is insufficient, the appropriate output may be a question, a gap list or an escalation — not an invented answer.
PRACTICAL LENS
What this means for your firm
The usefulness and reliability of internal agents increasingly depends on the quality, availability and ownership of firm knowledge. Adoption can expose where procedures are unclear, reference material is outdated or client context is fragmented; those are operating issues to address, not problems for AI to hide.
PRACTICAL LENS
What this means for services to your clients
Client-facing AI needs relevant client-specific context and clear boundaries between authoritative information, assumptions and unknowns. A service should make it clear what evidence supports an output and when the firm or client needs to provide more information.