Introducing Metal’s Autonomous Assistant

If you've read our recent post on AI’s broader applications for funds, you might remember our mention of Autonomous Behavior – the ability for Large Language Models (LLMs) to sequentially reason in order to achieve a goal.

This incredibly powerful capability opens up many new possibilities for investors, and when properly applied to a fund’s workflow, can be nothing short of transformative.

Today we’re thrilled to announce Metal’s first release in support of this functionality!

How It Works

Metal's datasources are highly capable of transforming both qualitative and quantitative financial data, making it fully compatible with LLMs. The Assistant takes this a step further, supporting a customizable and in-depth analysis of an entire datasource.

Simply select your data, input your request, and the Assistant will begin.

report prompt

The Assistant will list its thought process as it goes.

assistant thoughts

Once finished, your report is ready to review! Below is an example slide produced by this run.

assistant output

No Context Window Limitations

The time it takes to complete the task may vary depending on the amount of data being analyzed, but you are not limited by the LLMs context window. In other words, you can ask the Assistant to analyze thousands of documents with confidence.

For example, imagine you just got access to a data room filled with call transcripts, pitch decks, financial metrics, and more. You can use the Assistant to quickly extract and summarize key points, producing an output in either document or slide deck format.

The Assistant is incredibly flexible, you just need to give it instructions. For example, you can use the Assistant to:

• Understand key takeaways and sentiments from many expert call transcripts
• Accelerate company research by asking the Assistant to incorporate web results
• Perform a financial analyses on a portfolio company’s recent and historical reports

What’s Next?

More power, and more control. While this initial release is production ready, we will be adding additional customizations options – both in terms of what information the Assistant generates and the structure of its output, ensuring it conforms to your team’s formatting preferences.

If you would like to try this at your fund, please get in touch.