Copilot for process discovery

Navigating across process mining visualizations to extract insights for improvement is often very time-consuming. To enable a faster and more convenient approach to process discovery, we have an an AI-powered Copilot.

Apromore Copilot is a chat interface that allows us to ask process discovery questions in Process Discoverer. The chat conversation is based on the process map displayed in Process Discoverer and can answer quick questions about the process performance. So, rather than navigating the barrage of information in Process Discoverer, we can simply ask Apromore.

To use our AI Copilot, open an event log in Process Discoverer and click Ask Apromore.

This opens the Apromore copilot chat window.

In the chat input field, we can ask process discovery questions about the event log we have opened. Let’s begin by asking what the most frequent activity in the process is. Type What is the most frequent activity in this process? Press Enter.

We see a progress bar indicating that Apromore Copilot is finding the answer to our question.

After a few seconds, Apromore Copilot returns the answer.

We can confirm this in the process map by going to the Total frequency overlay.

Note

By default, Copilot is disabled in client instances. To turn on Apromore Copilot in your tenancy, please contact your account executive or customer support. Once enabled, Copilot enables the user to submit up to 50 prompts per calendar.

Let’s say we want to dig deeper. We can ask why this activity is the most frequent.

The chatbot tells us it has the highest total frequency of 1108, which is correct.

Note

• AI models are non-deterministic. Hence, given the same prompt, the model may return a different response.

• Apromore Copilot understands the context based on the chat history in a session. Notice that we did not specify the activity in question in the second prompt, but Apromore Copilot correctly determined it from our earlier prompt. We can clear the chat history by closing and reopening Apromore Copilot or refreshing the Process discoverer page.

Let’s ask Apromore Copilot to determine the bottlenecks in the process.

The model correctly identifies “Approve Purchase Order” as a processing time bottleneck because it has the highest average processing time of 1.06 days.

It is important to note that when we apply a filter to the log, answers from Apromore Copilot will be based on the filtered process map. To demonstrate this, let us remove the most frequent activity, “Analyze Request for Quotation”.

Apromore Copilot detects that a filter has been applied and informs us that previous answers may no longer be valid.

Now, let us ask what the most frequent activity in the process is.

Apromore Copilot provided its answer based on the filtered log being displayed.