Research you can check, source by source
Research asks the AI vendor’s own web search for professional facts only. The confidence of each finding is decided by the server from the sources actually returned — never by the model — and nothing is applied to a contact until you accept it.
Who uses it
- Managing director
- Partner
- Family-office executive
- Business-development director
What it does
- Professional research within your monthly research credits, each finding with its sources and their retrieval date.
- Confidence classes decided by the server: verified (the organisation’s own or a regulator’s domain), corroborated (two or more independent domains), unconfirmed, inferred, outdated (older than 18 months) or disputed.
- Monitoring of chosen contacts, with research-driven alerts (Intelligence).
- The Executive Concierge, which prepares briefings and drafts from your records (Intelligence).
- Every AI step through the governed AI engine, with its kill switches, budgets and ledger; without an AI provider, rules-based versions labelled “From your records”.
How it works
- Ask for research on a contact or organisation.
- Read the findings with their sources and confidence.
- Accept what you agree with; dismiss the rest.
The controls that protect it
- Sensitive categories are filtered out by the prompt and by a server-side filter; LinkedIn is not used.
- A sealed contact is never researched, monitored or put in a prompt.
- A URL the model named that the search did not return is marked unconfirmed.
What it does not do (yet)
- Research needs a provider that can search the web through Achekam (Anthropic or Google Gemini).
- Monitoring, alerts and the Concierge are in the Intelligence tier and above.
