# Dec 14th, 2024: \[EN\] Top 5 Prompts to Maximize Your Observability AI Assistant Experience

**URL:** <https://discuss.elastic.co/t/dec-14th-2024-en-top-5-prompts-to-maximize-your-observability-ai-assistant-experience/370215>\
**Category:** Advent Calendar\
**Created:** [December 14, 2024, 8:00am UTC](https://discuss.elastic.co/t/dec-14th-2024-en-top-5-prompts-to-maximize-your-observability-ai-assistant-experience/370215 "2024-12-14T08:00:06Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Zoia\_AUBRY](https://avatars.discourse-cdn.com/v4/letter/z/b2d939/32.png) [@Zoia\_AUBRY](https://discuss.elastic.co/u/Zoia_AUBRY)\
**Post date:** [December 14, 2024, 8:00am UTC](https://discuss.elastic.co/t/dec-14th-2024-en-top-5-prompts-to-maximize-your-observability-ai-assistant-experience/370215/1 "2024-12-14T08:00:06Z")

</div>

**Top 5 Prompts to Maximize Your Observability AI Assistant Experience**

Over the past three years as a Customer Engineer, I’ve met hundreds of customers, and one of the most frequently asked questions has been: "I _have my data in Elastic; how can I leverage it to gain the most advantage_?".

If this applies to you, then this blog and AI Assistant is for you.

Released in version 8.9, the Elastic Observability AI Assistant introduced a range of powerful new features. Today, I’ll share **my top 5 favorite prompts** to help you make the most of these enhancements.

In this example, I’m using [Azure Open AI](https://www.elastic.co/guide/en/security/current/assistant-connect-to-azure-openai.html) with GPT-4o model combining an ecommerce dataset with APM metrics to increase the business value.

**#5. Are there any errors in APM services, _show the source code_**

 ![](https://us1.discourse-cdn.com/elastic/original/3X/b/e/beeb63143ce3cef2422f36ec78553cbbb176d044.png)

Bonus question: How to resolve?  
_Tip : Connect AI assistant also to your Internal KBs to get more precise results_

**#4. Create _separate_ alerts for these errors if reoccur**

 ![](https://us1.discourse-cdn.com/elastic/original/3X/b/3/b3cf148bf3566c2d8282a390b1750cc8c1256014.jpeg)

At this time, the AI Assistant cannot create alerts for you… _ **yet** _

However, it can generate an ES|QL query that you can easily copy and paste into the body of Observability rules while creating alerts.

Now let’s move to our ecommerce data.

**#3 Based on my index\_ name can you help me create 10 dashboards that can accelerate my business**

 ![](https://us1.discourse-cdn.com/elastic/original/3X/c/f/cffc8146ded4349f000a35845150a36150a21247.png)

_Tip : Next, you can instruct the AI Assistant to create all of them for you and save them as visualizations within your dashboards._  
_You can even define the type of visualization and time range if needed._

**#2 Create ML job to detect any unexpected behavior in the orders in index\_name**

Bonus: The AI Assistant can **set up ML jobs for you**.

 ![](https://us1.discourse-cdn.com/elastic/original/3X/4/a/4a49f7134cbd70535e1508969d084a7aa6819de9.jpeg)

You can also define Bucket Span and Prune Window while creating the prompt.

**#1. Analyze my active apm-'\*' alerts and give me 20 more alerts for better monitoring of my applications**

 ![](https://us1.discourse-cdn.com/elastic/original/3X/f/2/f25e235929c21c3ff473834644ebdf36a222daf1.png)

There are countless options to consider, and with direct access to your datasets, an AI assistant can deliver highly accurate results tailored to your specific use case.

I’d also like to share a few tips for using the AI Assistant effectively:

1. **Be Specific :** Keep your questions clear and concise; avoid lengthy text.

2. **Know Your Data** : Before asking the AI for help, ensure you know your index or .ds-\* names and the available fields. To check these, you can open any index in Discover and expand a document to find the field names you need.

3. **Monitor your spends** : To avoid unexpected costs, [monitor your token usage,](https://www.elastic.co/guide/en/observability/current/obs-ai-assistant.html#obs-ai-token-limits) as retrieving responses can consume a significant number of tokens.