First of all, Happy New Year, everyone! I needed to get that out of the way. We hope everyone had a fantastic holiday season!
We’re excited for 2025 at BIS. I’ve been feeding y’all information about the upcoming version and the ground-breaking AI features that we’re using to get better results faster than ever.
But you might be asking yourself “Where does this new technology fit or why should I care?” Fair enough.
Let me try to shed some light on this whole AI-Vector-Indexed-Assistant thing.
We try to use Grooper internally as much as possible.
Our accounting, marketing, and production EOB processing services, as well as other ad hoc processes around the company, rely on Grooper to get things done faster and more accurately than people can.
But this AI technology has taken us to a new level. Here’s how…
A 300% Improvement On One Task
I created an AI Assistant with an early version of 2024.1 (read: dev version) to help my sales teams respond to Request for Proposals (RFPs) faster and more efficiently.
Why RFPs? Responding to RFPs is one of the more unpopular tasks a salesperson has to do. So, we don’t do very many of them unless we can influence the creation of the document.
There’s always a chance that a response can win business, but it’s a very low-probability task. As I said, we don’t do it very often.
However, our new RFP Helper AI Assistant knows everything about Grooper because Grooper’s new AI Assistants default to having all the Grooper wiki information accessible to them. I also added all the RFPs we’ve written over the years, as well as all our sales collateral, such as use cases and blog posts. It’s a lot of information.
Grooper takes those documents and creates vector indexes with OpenAI’s biggest and best embedding model. That means we have all that data categorized by meaning and context.
Today, our salespeople or administrative personnel can simply ask the AI Assistant the RFP questions. We then get back an answer that was created in reference to all that vector-indexed content. And since we have all that reference data for the LLM to use as example data, the responses are in line with our corporate talk tracks and messaging (no hallucinations).
We’re seeing a 300% reduction in the amount of time it takes us to create an RFP response! And this assistant only took as long as it took to OCR all the documents I fed into it.
The first version was live in less than 15 minutes.
A Better Way to Perform Data Discovery
However, the more significant Grooper revelation is the ability to search inside documents in Grooper instead of starting from scratch when building data models.
For example, we process a lot of Explanation of Benefits (EOB) documents daily. EOBs are a real pain to process because the details vary across every single medical provider.
But we now group similar EOBs together by their common features before we create a definition for them. Find the needles in the hundreds of thousands of pages we process each day. Then build data models that have the most immediate business impact.
In the past, we’d have you (the client), give us the documents that you wanted us to focus on. That ends up being a massive delay as you try to gather a good representation of your documents for us to start building definitions and models.
Get Immediate Value on Day One
But today, you get value from Grooper faster and easier than ever before. Just start gathering documents and get them into Grooper.
And don’t worry about defining things perfectly before you begin. That’s the older, slower way of doing things.
Now you can get value on day one by scanning or importing documents and starting to search through those documents and take action on them.
We’re in a new era for document processing.
But What About the Bots?
Calling Grooper’s new AI Assistant capability a “bot” drastically discounts what we’re doing.
That’s because the assistant is closer to an “Agent” than a bot. The assistant first looks at the prompt then determines what resources it needs to answer the question. This includes:
- Vector-indexed documents
- Databases
- External websites
- Other resources
- Asking the user for more info
Each assistant gets its own resources, so you can quickly build assistants with access to only the required resources.
With this method, you can have customer service assistants in the same system as an accounting assistant. Here are just a few use cases for an AI Assistant:
- Regulatory Compliance Assistance
- In industries with changing compliance regulations, these assistants can quickly find what’s changed and analyze existing data for risk.
- Financial Reporting and Analysis
- An assistant that can consolidate data from internal sources like ERPs and external market data to create reports, trend analysis, and predictive insights on demand.
- Legal and Contract Review
- An assistant to compare contracts, find questionable clauses, suggest improvements, and ensure compliance.
- Onboarding Help
- This friendly new-hire assistant has access to all the organization’s internal knowledge and policies. This assistant is always ready to answer questions – even questions that employees might not want to ask their company’s personnel.
- A Gateway to Knowledge
- This is a simple assistant with access to all your internal documents and data across all repositories. It can find new relationships that you didn’t even know existed.
I can’t stress enough how easy these assistants are to create. I created several this week, and the longest part of the process is OCR’ing documents that weren’t already text-searchable.
Give us a call today to see these Assistants in action!
