AI use cases in business processes
AI helpers (the AI assistant and AI bots) in Pyrus can automate routine processes, analyze data, and make decisions based on rules and templates to speed up the processing of inquiries, improve the quality of responses, and free up employees to focus on non-routine tasks.
ServiceDesk
Voice message processing
Problem: Support staff have to manually classify queries, create tasks, and evaluate conversations after every incoming call.
Solution: Using its integration with telephony systems, Pyrus automatically creates a task when a call comes in and, after the conversation ends, attaches an audio file of the call to the task. The AI bot transcribes the recording, classifies the query, fills in the task’s Request Type field, analyzes the conversation for compliance with internal service standards, and fills in the Compliance with Regulations field.
Learn more about the capabilities of telephony integration in the Extensions section
You are a tech support rep. Transcribe the text from the last audio recording attached to the task. If transcription fails, add the following comment: "Error: Unable to process audio file. Manual review required." Reject the task and end the instruction. If transcription is successful, fill in the Request Type field according to the customer’s issue. Evaluate the conversation for compliance with the customer interaction procedures https://pyrus.com/t#kb/article/analitik-AAAAAAAAAAA. Fill in the Compliance with Regulations field with the appropriate value. Approve the task.
Request Type catalog format
| Request type | Nature of the request |
|---|---|
| Question/consultation | Requests for information about the company’s products or services |
| Bug | Report of a technical failure that needs to be fixed |
| Improvement suggestion | Ideas for improving, optimizing, or modifying existing products, processes, or systems |
| Other | Inquiries that do not fall into the categories above |
Compliance with Regulations catalog format
| Score | Evaluation criteria |
|---|---|
| 1 | Rude communication, jargon, negative remarks; no solution to the customer’s problem was provided |
| 2 | Procedures violated, procedure steps skipped, incorrect answer; the problem was not solved |
| 3 | Technically correct answer, procedures formally followed, but not customer-oriented enough |
| 4 | Correct solution provided, but the conversation dragged on or the customer waited a long time for a response |
| 5 | All procedure steps completed. Professionalism, empathy, high-quality solution. The customer is satisfied |
Filtering spam and mass mailings
Problem: Customer support frequently receives mass mailings and spam, which takes employees time to identify and remove.
Solution: The AI bot analyzes incoming messages, identifying signs of spam (such as the words “free,” “earnings,” or “promotion,” or suspicious URLs) and mass mailings (similar messages sent from the same source). Tasks where the bot identifies such signs are marked by checking the Spam/Mailing box.
You are a tech support rep. Analyze all of the user’s messages in the task. If the messages are spam (for example, they contain suspicious links, offers to make money, or advertisements) or mass mailings (for example, identical or similar messages sent from the same source), fill in the Spam/Mailing field in the task. If the messages are not spam or mass mailings, leave the Spam/Mailing field blank. Approve the task.
Tip: The Spam/Mailing field is a Checkbox type field. You can configure simplified workflow in the form to process inquiries marked as spam.
Importing data from other tasks
Problem: Requests received at the shared Support email require employees to search for customer information (company name, city, and business area) manually.
Solution: The AI bot identifies the customer and retrieves additional information from another task, automating the entry of the required data.
Try to find information about the customer. Use the organization name and email in the current task as search criteria. If no matches are found, leave the comment: "Unable to identify customer." Reject the task. If a relevant task is found, combine the values of the Last Name and First Name fields from the found task and enter them in the current task’s Customer field. Copy the values of the City and Business Area fields from the found task to the current task. Approve the task.
Tip: We recommend that you check which forms and tasks the bot has access to. It can work only with objects to which it has been granted access.
Prioritizing and classifying inquiries
Problem: Incoming requests cover different topics and have different priorities, requiring operators to spend time classifying them and determining their urgency.
Solution: The AI bot analyzes the text of an incoming request, determines its topic, and assigns a priority according to established criteria, speeding up the routing of the task to the appropriate specialist.
You are a tech support rep. Compile a summary of the task. Fill in the Request Type and Priority fields based on the customer’s request. Approve the task.
Request Type catalog format
| Request type | Nature of the request |
|---|---|
| Question/consultation | Requests for information about the company’s products or services |
| Bug | Report of a technical failure that needs to be fixed |
| Improvement suggestion | Ideas for improving, optimizing, or modifying existing products, processes, or systems |
| Other | Inquiries that do not fall into the categories above |
Priority catalog format
| Priority | Nature of the request |
|---|---|
| P1 (Critical) | The system is down; the service cannot be provided |
| P2 (Urgent) | A serious malfunction affecting the user, but workarounds exist |
| P3 (Standard) | A typical question that must be resolved during business hours |
| P4 (Low) | A request that does not require immediate attention |
Classifying multilingual inquiries
Problem: Your support department operates in multiple languages. Inquiries need to be routed to specialists who speak the appropriate language.
Solution: The AI bot analyzes the text of an incoming request, determines the language and topic of the inquiry, and then enters the corresponding values in the form, ensuring that the task is routed to the appropriate specialist.
Tip: To record the language of the customer’s inquiry in the task, add a Choice type field to the form template.
You are a tech support rep. Determine the language of the customer’s inquiry (Russian, English, Chinese, or another language). Fill in the Inquiry Language field with the appropriate value. Fill in the Request Type field based on the customer’s request. Approve the task.
Request Type catalog format
| Request type | Nature of the request |
|---|---|
| Question/consultation | Requests for information about the company’s products or services |
| Bug | Report of a technical failure that needs to be fixed |
| Improvement suggestion | Ideas for improving, optimizing, or modifying existing products, processes, or systems |
| Other | Inquiries that do not fall into the categories above |
Answering frequently asked questions (FAQ)
Problem: Support staff spend a lot of time answering repetitive questions, which reduces their efficiency and distracts them from more complex tasks.
Solution: The AI bot analyzes the customer’s request, determines the topic of the inquiry, and searches for a relevant answer in the specified section of the corporate Knowledge Base (KB). If a suitable answer is found, the bot prepares it as a ready-to-use draft for the operator, speeding up the process of resolving routine requests.
You are a tech support rep. Analyze the text and attached files in the task. Fill in the Topic field with a brief description of the problem. Find the answer to the user’s question in the Knowledge Base, in the section https://pyrus.com/t#kb/topic/sometext-AAAAAAAAAAA. If you find relevant information, write the answer to the user’s question in a comment in the task. If no answer is found, leave the following comment: "No information related to the user’s request was found in the specified section of the Knowledge Base." Approve the task.
Tip: Recommendations for structuring the FAQ section in your Knowledge Base.
Create a separate section in the Knowledge Base for frequently asked questions. You can name this section something like Frequently Asked Questions or FAQ.
Format each individual question, typical scenario, or solution as a separate article.
Each article should begin with a clear, informative title that briefly describes the case or problem (for example, “How do I reset my password?”, “File upload errors,” or “Instructions for processing a return”).
Include only articles that correspond to the current functionality and processes in the FAQ section.
Evaluating support rep performance
Problem: Support management needs to objectively evaluate operators’ performance, identify systemic communication issues, and monitor compliance with corporate service standards when handling a high volume of inquiries.
Solution: When a task is closed, the AI bot automatically analyzes the correspondence between the operator and the customer, compares the operator’s actions with the procedures in the Knowledge Base, compiles a brief summary of the interaction, and assigns a score from 1 to 5 in the Score field. If the score is low (1–2), the bot initiates an additional review.
Tip: To assign a score to a Support employee, add a Number type field to the form template.
You are the head of Technical Support. Review the customer interaction procedures in the Knowledge Base, in the section https://pyrus.com/t#kb/topic/sometext-AAAAAAAAAAA. Put together a brief summary of the operator’s performance based on their compliance with the customer interaction procedures. Enter this summary in the Operator Performance field. Fill in the Score field with a value from 1 to 5, where 5 means full compliance with the job description and 1 means complete non-compliance. If the Score field is 1 or 2, add the user https://pyrus.com/t#pp12345 as an Approver at the Inquiry Evaluation step. Approve the task.
CRM
Selecting the appropriate sales funnel
Problem: When a new potential customer (a lead) comes in, the manager needs to determine which sales funnel to move the deal to. This requires knowledge of the specifics of each customer type, product, and market segment.
Solution: The AI bot analyzes information about the customer, their request, the potential product/service, and other factors to automatically determine the most appropriate sales funnel and fill in the key task fields.
Tip: To fill in the Customer Type and Service/Product fields, add two corresponding Choice type fields to the form template.
Analyze the text and attached files in the task.
Fill in the Topic field with a brief description
of the problem.
Try to find information about the customer.
Use the organization name and email
in the current task as search criteria.
Determine the customer type
(B2C, B2B, B2G, internal customer).
Fill in the Customer Type field
with the appropriate value.
Based on the customer’s request, determine
the service or product they need
(Consulting and Audit, Software Development,
Equipment and Hardware, Education and Training).
Fill in the Service/Product field
with the appropriate value.
If you need additional information
to determine the funnel, leave the following comment:
"To determine the funnel accurately, we need to clarify:
{a clarifying question, for example,
the customer’s type of business}".
Approve the task.
Tip: We recommend checking which forms and tasks the bot has access to. It can work only with objects to which it has been granted access.
Optimizing contract approval
Problem: Contract approval takes a long time because employees have to manually compare versions of documents and look for discrepancies with the data in CRM form fields.
Solution: The AI bot automatically compares the new version of a document with the previous version or a standard template, identifies and documents all differences, and checks whether the key task fields match. This minimizes manual data-entry errors and speeds up the deal cycle.
Review the last document attached to the task. If multiple versions of the document are attached, compare them with each other, identify all changes, and add a description of them in a comment. Compare the information in the document with the data in the following task fields: Customer Name (must match the customer name in the document); Deal Total (must match the amount in the document); Payment Terms (must correspond to the terms in the document). If any information in the document does not match one of the task fields, or if there is a significant difference between document versions, leave the following comment: "The review identified the following discrepancy/difference: [provide a detailed description of each discrepancy or change]." Reject the task. If all checks are completed successfully and no discrepancies are found, leave the following comment: "The information in the document(s) fully matches the task data and contains no incorrect changes." Approve the task.
Answering frequently asked questions (FAQ)
Problem: Sales managers spend a lot of time composing responses to simple questions, which reduces their efficiency and distracts them from their core tasks.
Solution: The AI bot analyzes the customer’s request, determines the Topic of the inquiry, and searches for a relevant answer in the specified section of the corporate Knowledge Base (KB). If a suitable answer is found, the bot prepares it as a personalized draft for the manager.
You are a sales manager. Analyze the text and attached files in the task. Fill in the Topic field with a brief description of the problem. Find the answer to the user’s question in the Knowledge Base, in the section https://pyrus.com/t#kb/topic/sometext-AAAAAAAA. If relevant information is found, write the answer to the user’s question in a comment in the task. If no answer is found, leave the following comment: "No information related to the user’s request was found in the specified section of the Knowledge Base." Approve the task.
Tip: Recommendations for structuring the FAQ section in the Knowledge Base:
Create a separate section in the Knowledge Base specifically for frequently asked questions. You can name the section something like Frequently Asked Questions or FAQ.
Format each individual question, typical scenario, or solution as a separate article.
Each article should begin with a clear, informative title that briefly describes the case or problem (for example, “What is the cost of the product or service?”, “How do I process a return?”, or “How can I check the status of my inquiry?”).
Include only articles that correspond to the current functionality and processes in the FAQ section.
Evaluating manager performance
Problem: The head of Sales needs to objectively evaluate managers’ performance, identify systemic communication issues, and monitor compliance with corporate standards for interacting with customers.
Solution: When a task is closed, the AI bot automatically analyzes the correspondence between the manager and the customer, compares the manager’s actions with the procedures in the Knowledge Base, prepares a brief summary of the interaction, and assigns a score from 1 to 5 in the Score field. If the score is low (1–2), the bot initiates an additional review.
Tip: To fill in the Score field, add a Number type field to the form template.
You are the head of Sales. Review the customer interaction procedures in the Knowledge Base, in the section https://pyrus.com/t#kb/topic/sometext-AAAAAAA. Put together a brief summary of the manager’s performance based on their compliance with the customer interaction procedures. Enter this summary in the Manager Performance field. Fill in the Score field with a value from 1 to 5, where 5 means full compliance with the job description and 1 means complete non-compliance. If the Score field is 1 or 2, add the user https://pyrus.com/t#pp12345 as an Approver at the Inquiry Evaluation step. Approve the task.
HRM
Analyzing candidate resumes
Problem: An HR specialist needs to quickly analyze a large number of resumes to select the most suitable candidates for an open position, which is a time-consuming and subjective process.
Solution: The AI bot automatically analyzes the resume, compares it with the job requirements, highlights the candidate’s strengths and weaknesses, and provides a structured assessment of how well the candidate matches the position.
Tip: To specify the job title, add a Choice type field to the form template. To assign a score indicating how well the candidate matches the position, add a Number type field to the form template.
You are an HR rep. Analyze the last candidate resume attached to the task. If the resume is missing, leave the following comment: "Error: Resume not found." Reject the task and end the instruction. Review the job requirements in the related task. Compare the information in the resume with the job requirements. Prepare a brief candidate summary that includes: the job title; the candidate’s key strengths in the context of this position (experience, skills, education); potential weaknesses or gaps relative to the requirements. Post the summary as a comment in the task. Fill in the Job Match Score field with a value from 1 to 10, where 10 means full compliance with the job requirements and 1 means complete non-compliance. Approve the task.
Comparing multiple candidates’ resumes
Problem: When there are several suitable candidates, an HR specialist needs to objectively compare them based on key criteria, which requires significant time and effort to avoid subjectivity.
Solution: The AI assistant compares candidates’ resumes based on specified criteria (related to the current job opening) and provides structured analysis with recommendations.
Review the job requirements in the current task. Review all related tasks containing candidate resumes. For each candidate: analyze the resume, identify their strengths, how well they meet the requirements, and their weaknesses and areas of non-compliance. Compare the candidates against each other based on the key job criteria. Compile a response containing: a brief comparison of each candidate’s advantages and disadvantages relative to the position; a recommendation for the most suitable candidate, with explanation of reasons.
Transcribing and summarizing interview results
Problem: Manually transcribing audio recordings of interviews and meetings to document agreements made requires significant time and is prone to errors.
Solution: The AI assistant automatically converts audio into text, extracts key agreements related to HR processes (such as hiring and onboarding terms), and prepares a structured summary.
Analyze all files and comments attached to the task. Identify and extract all key agreements reached with the candidate/employee. Describe all key agreements with the candidate.