08 / Market Briefings · Field briefing

02 / Resolve the contradiction

Resolve the contradiction

Start with a decision and a scope. “What is happening in AI?” is too broad to verify. “Which document-answering approaches should a small professional team trial?” creates a tractable question about features, data boundaries, support, cost and evaluation requirements.

Create a claim log before drafting the narrative. Each material statement needs an original source, a date, a precise passage or table, and a note about what it actually supports. Record the publication date separately from the date you accessed it. An accessible page may describe an old release.

Distinguish independent observations, vendor claims and your own calculations. A vendor demonstration is evidence of what the vendor presented, not independent confirmation of production reliability. A funding announcement says little about whether a tool solves your task. A repeated press-release sentence is still one source, even when several outlets repeat it.

For an invented product comparison, two sources appear to disagree about a document-page limit. Inspect the model version, interface, units and date before averaging or choosing one number. Both may be correct for different configurations. If the discrepancy cannot be resolved, preserve it in the conclusion.

Ask the assistant to work with a supplied evidence set and make source gaps explicit. Where tools can browse, verify that cited pages were actually retrieved and that the relevant passages support the claims. Structured citation support can help preserve locations, but it cannot establish independence or resolve a misleading source by itself.

Use an evidence-to-conclusion chain: observation, interpretation, decision implication. If the observation is that a provider offers a new feature, the interpretation may be that an additional pilot option exists. The implication is to test it against your criteria, not declare the market transformed.

Market-size forecasts require especially clear definitions and assumptions. Revenue, spending, users and addressable organisations are different quantities. This introductory guide does not invent a market forecast or rank providers without comparative evidence.

04 / Go deeper

Useful words. Important details.

Primary source

Original material supporting a claim, interpreted within its scope.

Interpretation

The meaning you infer from evidence; it is not the observation itself.

Independent evidence

Evidence produced separately from the interested party making a claim.

Uncertainty

A recorded limit, unresolved conflict or assumption that affects a conclusion.

The technical notes qualify the explanation; they are not product-specific assurance or a certification.

05 / Evidence & scope

Inspect the source.
Keep the context.

Sources were accessed on 4 October 2026. These technical and security references support the mechanisms described; they do not validate our fictional examples or establish comparative product performance.

1. Anthropic · Citations

   Structured citation locations in a provider implementation. Citations do not establish source authority or factual correctness.

2. Microsoft · Using Generative AI Responsibly

   Responsible-use principles, representative testing and operational ownership. Technical guidance, not legal advice.

3. OWASP · Misinformation, 2026

   Unsupported outputs, stale context, omissions and claim-check-act patterns. A risk framework, not a research accuracy benchmark.