AI in every workflow, on a ledger you can audit.
Pixie, the AI built into Pixcel, works inside daily servicing: it investigates loans, triages overnight job problems, and drafts complaint, hardship and collections replies. It suggests; your team decides.
It is safe to use because of what sits underneath. Every balance comes from a recorded event history, Pixie works with each user’s own permissions, and every answer is logged with the records it used.
Faster operations, with the guardrails a lender needs.
AI built into the work your team already does, not a chatbot on the side. And through the same API, you can build agents of your own.
Loan investigation
Ask why a loan is in arrears or how a balance was reached, and get an answer from the loan’s own records, with sources.
Overnight job triage
When a nightly servicing job hits a problem, Pixie explains what failed and why, so your team can fix it sooner.
Complaints and hardship
Pixie drafts a summary and a reply for your team to review. Regulatory timeframes are calculated by the platform, not by the AI.
Collections drafts
Reminders and messages drafted with the right figures and tone, ready for your team to review and send.
Private and in your region
Personal data is masked before anything reaches the AI, and AI runs in your data’s region. Pixcel refuses any model that would send data outside it.
Build your own agents
Agents you build on the Pixcel API go through the same validation, permissions and audit trail as your team. Every action is attributable.
Why auditable AI starts with an auditable ledger
Pixie drafts and recommends. A person on your team confirms any change.
Every balance comes from a recorded event history, so AI answers trace back to real events.
Pixie works with each user’s own permissions, and API actions, human or agent, are checked against the same roles.
Every question and answer is logged, with the records it used.
Common questions
Pixie is the AI built into Pixcel. It investigates loans, triages overnight job problems, and drafts complaint, hardship and collections replies, using your own servicing records and citing the records it used. Your team reviews and decides.
Personal data is masked before anything reaches the AI model, processing stays in your region, and every question and answer is logged with the records it cited. Pixie never changes a balance on its own: any change it suggests needs a person to confirm it.
Yes. Everything your team does in Pixcel is available through the API, with the same validation, permissions and audit trail. An agent you build acts like any other user: it can only do what its role allows, and every action it takes is recorded and attributable.
It depends on what sits underneath it. Pixcel records every balance change as a time-ordered event and every action against a person or system, so AI answers can be checked and AI actions can be traced. Regulatory dates are calculated by the platform, not the model, and a person confirms any change.
See Pixie on a real loan
We’ll show you Pixie investigating a loan and drafting a reply on the real product, and talk through what you could build on the API.