What are Exceptions?
Learn what exceptions are in Kognitos and why they keep your automations accurate.
Overview
In Kognitos, exceptions occur when an automation cannot proceed without additional context or clarification. Instead of making an assumption that could lead to errors, the system raises a question (also called an exception) and routes it to a human for review.
This human-in-the-loop design ensures that business processes remain accurate, reliable, and trustworthy. Exceptions are not system failures — they are intentional checkpoints that keep automations accurate and reliable.
Why Do Exceptions Occur?
Exceptions can arise in Kognitos for a variety of reasons. The most common include:
Ambiguity
Multiple possible interpretations are possible, and the system cannot determine which is the correct one without user input.
Example: An email approval workflow asks, “Approved?” but doesn’t specify if it refers to cost, vendor, or timeline, so an exception is raised for review.
Missing Information
A required data input is not available, preventing the automation from completing the step.
Example: While processing an employee expense report, the system detects that a hotel receipt is missing the check-in date, so it raises an exception for confirmation.
Document Variability
The information exists in the document, but appears in an unexpected format or structure. This makes it difficult to extract with certainty.
Example: An invoice shows the total amount as “$5,000 (including VAT)” in a note at the bottom instead of in the standard “Total” field, so the system raises an exception for review.
Business Rules
A step requires human decision-making, often due to risk, compliance, or policy considerations.
Example: While reviewing a contract, the automation extracts a clause about early termination fees. Because company policy requires legal review of any termination penalties above $10,000, the system raises an exception so a human can confirm how to proceed.
How to Resolve Exceptions
Exceptions in Kognitos are resolved by providing guidance — giving the system the input it needs to understand how to proceed. This can include clarifying ambiguous values, pointing to the correct information in a document, or confirming a decision.
The Guidance Center makes this process efficient by allowing you to review open exceptions across automations in one place and provide the necessary clarification or correction. Once guidance is provided, the automation resumes from where it left off.
Guidance Examples:
Identifying the correct field in a scanned invoice that represents the “Total Amount Due.”
Selecting the correct shipping address from a document that lists multiple locations.
Confirming the contract start date when it appears in narrative text instead of a standard date field (e.g., “services commence on June 1st”).
Verifying the correct vendor name when variations or abbreviations appear across different documents (e.g., “Northstar Technologies” vs. “Northstar Tech LLC”).
Summary
Exceptions in Kognitos are not errors; they are intentional checkpoints that safeguard accuracy. By providing guidance in the Guidance Center, you ensure that automations continue smoothly, learn from your input, and become more resilient over time.
Last updated
Was this helpful?