Creating Copilots for Analyzing Documents Using Suggested Questions and User Inputs

Credal's AI Copilots are tailor-made workflow assistants that are designed to handle a wide range of tasks like customer support and contract review. We're excited to announce two features that make setting up and using your Copilots even easier: User Inputs and Suggested Questions. For a quick refresher on Copilots, check out our Quick Start Guide.

Example Copilot Use Case: Financial Investment Analysis

This Copilot how-to will be useful for any workflow that requires doing an analysis repeatedly, each time with a new document. For this example, imagine that you're an associate at a private equity firm, tasked with reviewing three Confidential Information Memorandums (CIMs) and writing investment committee memos. 

The problem

Typically, this involves asking a set of repetitive questions and a significant amount of Control-F’ing, copy/pasting, and formatting per report, amounting to significant manual work. A Credal Copilot can be created to analyze the CIMs and have a draft of the deal memos in minutes. You can then spend your brainpower reviewing the more analytical and critical parts of the memo.

The Credal Copilot solution

Let’s try building a Copilot to solve this problem. We should be able to create an assistant in a few minutes, and we can then test and tune it from there to get the exact outputs that we want.

a) Name and description. 

Let’s start by configuring the Copilot and assigning a name and description that would clarify the Copilot’s purpose. 

b) AI model. 

We chose the default GPT-4o as our LLM model and given our use case, decided to set a “precise” creativity level to narrow the Copilot’s response to rely more on the context and data provided, rather than on unique or creative brainstorming.

c) Prompt. 

We added background for the Copilot to generate deal memos based on specific instructions, including workflow topic, intender users, expected input, and style guidelines. The prompt can be incredibly complex or specific, and detail what sections of a document to search, how to format numbers and citations, and how to perform certain calculations. Since this Copilot is a demo, we added context that the CIMs provided would be example documents with limited financial dates, so the Copilot would need to add 5 years to display “actual financials” up to 2024 (date of publishing). We even specifically instructed the Copilot not to mention this in the output for our test case! 

We then added Suggested Questions based on common prompts that would be helpful to the deal memo creation. 

To keep things simple (for now), we skipped the Model Q&A Pairs, which would provide the Copilot context on what a good example looks like. 

d) Data.

We then provided guidelines around the data that will be the source of truth for the Copilot by adding a CIM as a required User Input to prompt the user to link this data source. We specified the CIM as a Pinned data source by selecting “Include full contents from documents” for the Copilot to read the entirety of the document when generating a response since the full report would be needed for generating a high quality answer. 

The user would then be able to input a CIM in any document type, such as a PDF or Google Doc.

e) Deploy.

We’re now ready to deploy our Copilot! We published the Copilot to the Credal Web UI for users to easily find and select this workflow assistant in the Copilot dropdown. 

Let’s check out a snippet of our output using the sample document:

Making This Solution Possible With Suggested Questions and User Inputs

We were able to solve a pain point for a common use case with two new features: Suggested Questions and User Inputs. 

Suggested Questions allow you to save prompts that would be frequently used in a given Copilot, eliminating the need to repeatedly paste the same prompts. For instance, in a Copilot built to synthesize research reports, you can save the following prompts: Summarize this paper, Extract the results, and List citations. 

User Inputs recommend which document should be added to the Copilot. Whether it’s a research paper, an incident response Slack channel, or any other document, you can define what your Copilot needs to search to address the prompt. User Inputs can be either specified as Pinned or Searchable Data depending on the Copilot’s needs. 

The saved Suggested Questions enable the user to auto-populate the prompt in one click, with User Inputs displayed below the prompt. 

Start creating your custom Credal Copilot

This example use case highlights the versatility and power of Credal’s AI Copilots. With User Inputs and Suggested Questions, setting up and utilizing your Copilots has never been easier. 

Ready to enhance your productivity? Set up your Credal Copilot today and experience the difference these new features can make. Whether you're dealing with financial documents, incident reports, or any other complex data, Credal's AI Copilots are here to help.

Questions or feedback? Need ideas on suggested financial use cases? Reach out to sharon@credal.ai. We’d love to hear from you. 

If you’re not using Credal yet, drop us a line at sales@credal.ai to get started!

Building blocks towards secure AI apps

Credal gives you everything you need to supercharge your business using generative AI, securely.

Ready to dive in?

Get Started