Question 1

What kind of manual classification work can I automate by connecting Nyckel to my other business tools?

Accepted Answer

Nyckel automates the manual review and sorting of images and text that would otherwise require human judgment. When connected to other platforms via automation workflows, incoming images or text can be automatically classified, routed to appropriate teams, logged in databases, or trigger downstream actions without any manual intervention. Common examples include sorting incoming submissions by category, flagging content that meets specific criteria, or organizing files based on their content rather than metadata.

Question 2

How does data flow between Nyckel and the rest of my automation setup?

Accepted Answer

Nyckel receives images or text through API calls from automation platforms, processes them through your trained classifier, and returns structured predictions that can be mapped to fields in other tools. The classification results (category label, confidence score) are then available as outputs to route data downstream, trigger conditional logic, or update records in connected systems. Data flows in one direction for classification: your system sends content to Nyckel, receives the classification result, and uses that result to inform what happens next in your workflow.

Question 3

What are common problems when automating Nyckel classifications across multiple workflows?

Accepted Answer

Classifier accuracy depends heavily on the quality and diversity of training data, so models trained on limited examples may misclassify edge cases in production. Automations can fail if Nyckel receives corrupt image files, unsupported formats, or text that falls outside the scope of your trained categories. Additionally, if your business processes or categories change significantly over time, your existing classifier may drift and require retraining to maintain accuracy across updated workflows.

Question 4

How should I set up my Nyckel classifier so it works reliably when automated?

Accepted Answer

Train your classifier on representative samples of real data you will actually process, including variations and edge cases you expect to encounter. Define clear, non-overlapping categories that match your actual business workflow needs, and avoid creating categories that are too similar or subjective. Document your classification rules and review classifier predictions regularly in production to catch drift early, then retrain with new examples if accuracy drops, especially if the types of incoming data change.

Automate the manual work in Nyckel

Nyckel by the numbers

Top 5 Nyckel friction points when done manually

Manual image sorting before Nyckel

Nyckel predictions require constant verification

Retraining Nyckel after data changes

Exporting Nyckel results for downstream systems

Handling Nyckel edge cases and rejections

What you can automate in Nyckel

Classify leads from new HubSpot form submissions with Nyckel

Import samples to Nyckel from new Google Sheets rows

Send Slack channel messages for new Zendesk tickets classified by Nyckel

Send training images to Nyckel from new Google Sheets rows

Start training in Nyckel when new messages are posted to Slack channels

Update new MySQL database rows with Nyckel classifications

Tools commonly connected to Nyckel

Nyckel Automation questions

Ready to automate your work in Nyckel?