Overview of Voyant's Intelligent Fact Find. Voyant’s Intelligent Fact Find uses Retrieval Augmented Generation (RAG) focused on document question answering. Unlike traditional generative AI, it does not create new content.
Instead, it leverages the Vertex AI API and Optical Character Recognition (OCR) to extract and structure existing data from documents. No learning, training, or content generation occurs. The system answers a carefully engineered prompt, which poses a question to the document to retrieve specific data points. These responses are returned in Voyant's structured format, streamlining the adviser’s workflow.
Key Technical Details
1. Data Sources & Architecture
• Data Sources: Adviser-uploaded documents (PDFs, scanned images, .docx, .csv, etc.). Advisers are responsible for redacting any sensitive client data before upload.
• Architecture: OCR digitizes the text and creates a document-specific corpus. A prompt engineered to retrieve specific data points is sent to the Vertex AI API. The structured response is then displayed in AdviserGo. Advisers must verify the accuracy of the extracted data.
2. Algorithms
• Uses the Vertex AI API.
• Relies on precisely engineered prompts to extract specific data points.
• Voyant systems and data remain separate from the Vertex AI environment. No client data is included in the prompt.
3. Deployment Pattern
• Intelligent Fact Find is currently in internal beta and only available for client testing on a case by case basis. Not yet available in all regions.
Transcript of training
In today's training, I will be walking you through how to use the Voyant iFact Find import tool. I’m going to start by going to the plus button in the bottom right-hand corner and creating a client. I have Brian Miller, and I’ll enter his date of birth, the plan start date, and the state of residence. You’ll want to at least create the clients in Voyant and enter the people involved, so we’ll create Brian, his spouse, and then add their child. Once we’re done with that, we’ll import their information.
Next, we’ll go down to the plus button and select the Intelligent Fact Find option. Once you open this, you can choose a file to import. You’ll see several file options, including PDF, JPG, JPEG, PNG, CSV, and TXT. One thing to keep in mind is the file size—it will not accept files over 5 megabytes, so make sure to check that.
I’m going to use the US Fact Find that was specifically created for this tool, but you can use any file containing client data, and it should be accepted. I’ll go ahead and choose my file and import it. Depending on the size of the file, this could take a minute or two to process. Once it’s finished, a banner will pop up indicating that it has been processed, and you can hit “Complete.”
The system will then ask you to review all of the items it found in the document. You can choose to include or exclude items, and you can click on any item to make adjustments if needed. If you want to exclude something, just toggle it off so it won’t be brought into the plan. If everything looks correct, you can import the data into the plan.
If the tool finds an item that needs adjustment, it will display a red banner. For example, in this case, Nina’s Roth needs a value adjustment. It says, “Roth percentage must have a value.” We’ll go in and give her Roth contribution a value, change it to a percentage, and click Done to save. Now the plan has been brought over correctly.
You can click the arrow to review all of the data imported from the fact find. I hope this was helpful. If you have any questions or need assistance importing data, you can click the name of your client in the top right-hand corner, select Request Support, enter your question into the text box, and share client access.
*If you need assistance, you can click the name of your client in the top right corner of your plan and from the drop-down select "request support". Or email us at support@planwithvoyant.com.