Responsible AI
PIE is built to say “I'm not sure yet,” not guess.
PIE answers from your own data, not the open web
When you ask PIE a question, it answers only from the receipts, notes, and corrections it has actually read from you. It does not draw on general internet knowledge to answer a question about your own spending, and it never invents a receipt or a number that isn't really there.
Every Insight explains itself
PIE never states something it learned without saying why: what it noticed, how confident it is, and what evidence backs it up. When the evidence is too thin to say anything useful yet, PIE says that plainly instead of manufacturing a confident-sounding answer.
Confidence is never faked
PIE checks its own AI-generated answers against what it actually has evidence for before showing them to you - it does not simply trust an AI response at face value. If PIE can't confirm something with real evidence, it will tell you it doesn't know rather than answer anyway.
You correct PIE, and it remembers
When you tell PIE it got something wrong, that correction is recorded and used the next time something similar comes up - PIE does not ask you to correct the same thing twice without explaining why it's asking again.