Leveraging Ask HelixGPT for efficient incident resolution
Scenario 1
Scenario 2
How Ask HelixGPT works
You can use Ask HelixGPT to deliver intelligent, context-aware responses through enterprise knowledge and conversational understanding. Over time, its response generation capabilities have evolved to better handle a wider range of queries, including those with limited context or vague phrasing.
Agentic workflow
Use Ask HelixGPT with agent assistance to improve flexibility, handle ambiguous or weak prompts, and handle multi-step actions.
The following diagram depicts how Ask HelixGPT processes queries.
Retrieval Augmented Generation(RAG) based workflow
To get the most out of Ask HelixGPT through a RAG approach, use prompts that align with its capabilities. BMC HelixGPT uses published knowledge articles as reference data for generating summarized information. Therefore, when you use the Ask HelixGPT tab to find information, a published knowledge article relevant to the incident must already exist in the knowledge repository, and you must be entitled to view that knowledge article. If you are not entitled to view the knowledge article, BMC HelixGPT does not generate any response. Response generation is influenced by several factors, such as the accuracy of the knowledge articles, the model, and the prompt configuration. Therefore, the generated responses might vary.
The following diagram depicts how Ask HelixGPT processes incident resolution and conversational queries to generate responses.
Understanding this workflow will help you draft effective prompts.
Examples of effective prompts
These prompts are likely to return accurate and helpful responses.
Prompt example | How it works |
---|---|
My VPN connection is not working. | The question is answered only if a knowledge article about the incident is available. |
Summarize activity logs. | A summarized response is generated only if the incident has activity notes entered by the user. |
I need to reset my password. | Direct problem statement. The question is answered only if a knowledge article about the incident is available. |
Examples of weak prompts
These prompts may fail or return irrelevant results.
Prompt example | Why it fails | Recommended prompt |
---|---|---|
Can you find a knowledge article for this incident? | The incident context (summary) is not included in the question. | “Can you find a knowledge article for an incident where VPN connection is not working?” |
Tell me the current date and time. | The question is not related to the current incident in context. | |
What is the status of this ticket? | The question lacks an actionable problem description. |
Workflow for configuring and using Ask HelixGPT for BMC Helix ITSM
The following table describes the workflow for configuring and using Ask HelixGPT for BMC Helix ITSM:
Task | Component | Role | Action | Reference |
---|---|---|---|---|
1 | Not applicable | Technical contact (customer) | Contact your BMC account manager to get the Ask HelixGPT capability for BMC Helix ITSM. | Not applicable |
2 | BMC Helix Innovation Suite | Administrator | Configure Ask HelixGPT for BMC Helix ITSM. | Configuring Ask HelixGPT for BMC Helix ITSM |
3 | BMC Helix ITSM | Service desk agent | Use Ask HelixGPT to get information about possible resolutions for the incidents. | Resolving tickets with the help of Ask HelixGPT |