Create a Payment Reconciliation Process in minutes, with Kognitos
From one chat prompt to a working payment reconciliation: Concierge generates the English process, you click Apply, and Kognitos runs it end-to-end.
What's in this video
A 6-minute demonstration of going from a blank Kognitos playground to a running payment reconciliation process using nothing but a chat. The chat is purpose-built to generate Kognitos processes — not a general assistant, but a process designer for the platform.
The prompt
The user types a multi-requirement brief: extract customer number, payment number, date, and total amount from a remittance with several payments in a table (with headers); extract the whole table excluding the last row; convert it to an Excel file; email it as an attachment with the payment number in the subject line.
What Kognitos does
- Clarifying questions: the chat first asks whether the remittance is structured or scanned, what extraction format applies (PDF here), the recipient email, whether the user knows the column names, and whether the subject line should include the payment value as a placeholder or literal.
- English process generation: the chat returns a step-by-step English process explaining its plan and the exact Kognitos statements that implement it.
- Apply and run: the user clicks Apply on each step bottom-up, then Accept, then runs the playground.
- Textract + GPT extraction: Kognitos uses a purpose-built engine that combines AWS Textract with GPT for high-accuracy extraction without training data.
- Email out: the run completes, the Excel file lands in the user's inbox, and the subject line contains the payment number as requested.
Why this is different from "chat with your data"
The chat doesn't answer questions about the document — it generates an auditable Kognitos process the user can run, modify, and ship. The runtime is deterministic and produces an English log of every step, so the result is reproducible across thousands of remittances, not a one-off chat answer.