Skip to main content

​
Why confidence levels exist

AngelAI draws its answers from the AngelHub knowledge base — 33 documents covering every aspect of angel investment. Not every question has a direct answer in the knowledge base. Confidence levels tell you how well the knowledge base covers the question you asked, so you know how much weight to give the response.

​
The four confidence levels

​
High

The knowledge base directly addresses your question. The response is grounded in specific documents written for this purpose. What to do: Use the response as a substantive starting point. Check the source badges to see which documents were used. You can follow up with more specific questions to go deeper. Example questions that typically return high confidence:
  • “What is the standard vesting structure for founder shares?”
  • “How does weighted average anti-dilution work?”
  • “What are the EIS eligibility requirements for a SaaS company?”
  • “What stress tests should I apply to a financial model?”

​
Medium

The knowledge base partially addresses the question. AngelAI has drawn on relevant content but has also applied general reasoning to fill gaps. What to do: Use the response as a guide but recognise that parts of it may be less grounded than others. The source badges will show which parts are knowledge-base-backed. Consider following up with a more specific question that the knowledge base is more likely to cover directly. Example questions that typically return medium confidence:
  • Sector-specific questions in niche industries not fully covered in the sector reference
  • Questions that combine multiple topics (e.g. “How does EIS interact with an ASA conversion for a university spinout?”)
  • Questions about very recent legal or regulatory changes

​
Low

The knowledge base has limited coverage of this question. AngelAI is reasoning primarily from general knowledge rather than the specific documents. What to do: Treat the response as a direction to explore rather than a reliable answer. Verify the key points independently. If the question has significant financial or legal implications, speak to a qualified adviser. Example questions that typically return low confidence:
  • Questions outside the scope of UK angel investment (e.g. US-specific legal structures)
  • Highly specific technical questions about niche sectors
  • Questions about very recent events or regulatory changes

​
Not in knowledge base

The question is not covered by the knowledge base at all. AngelAI will tell you this explicitly rather than attempting an answer. What to do: This is the system being honest rather than guessing. Use it as a signal to seek information elsewhere — a solicitor, an accountant, or an industry specialist. If you see this frequently on a particular topic, it may indicate a gap in the AngelHub knowledge base. The platform logs these responses so the knowledge base can be improved over time.

​
The source badges

Every AngelAI response shows source badges — the specific knowledge base documents the response drew on. For example: vg1 — Valuation Methodology Guide
wp10 — Sector Reference
rf1 — Red Flags and Cognitive Bias Guide
lg1 — Legal Documents Guide
wntbb1 — What Needs to Be Believed Framework Click any badge to read the source document in full. This is the best way to go deeper on a topic AngelAI has touched on. Even a high-confidence response on a legal or financial question is not professional advice. AngelAI explains the framework — what the law says, how a structure typically works, what questions to ask a solicitor. It does not advise on your specific situation. For questions with material financial or legal implications, always verify with a qualified professional.