Training the AI Assistant on Your Data

Getting started with Vendorful is easy! All you need to do is upload some documents so that the AI Assistant can learn about your business and its offerings. This should be a quick and easy process. In fact, strongly recommend that you don't overdo it. :-) Adding 100 documents, including content from years past, will not improve the output. We typically see the best results with up to 20 pieces of content for a single product. Roughly half of those should be questionnaires.

Currently, we split documents into two types:

• Non-spreadsheet documents like PDF, MS Word files, text files, etc.
• Spreadsheets, specifically xlsx and csv files
• URLs

We'll cover each of these below.

Uploading Non-Spreadsheet Documents

Most companies have lots of written content about their offerings. These can include one-pagers, white papers, case studies, and other marketing collateral. For companies where security issues are a concern, there are typically other documents like disaster recovery and business continuity plans, IT security policies, and the like.

So what do you upload? Anything that you think will help support an answer to a question that you might get in a questionnaire from a buyer.

There are several buttons that will prompt you to upload content to train the AI. However, no matter what is on the main screen, you can always get to your Content Library by clicking the link on the sidebar:

Once you are on your content library page, it's easy to upload your documents, there will be a button that you can click to access the uploader:

Clicking on "Import Content" will take you to a page where you'll see two options wrapped in rectangles: Import Spreadsheet and Import Text. Click "Import Text as seen below."

Now all you need to do is click "browse your files" or drag and drop the file you want right into the gray uploader box and then click the Upload button. Please note that the current beta implementation supports one file at a time. We will consider adding support for uploading multiple files simultaneously in the future. Once you've uploaded the file, Vendorful will process it. After the file has been processed, you will be taken to a page where you see something like this:

You don't need to concern yourself with the information on this screen as it simplify verifies that the data has been added to the Content Library. Later, if you want to revisit this page, you can simply click on the name of any uploaded content in your Content Library and you will be taken right to it.

Pro tip: To help your AI Assistant generate high-quality answers, make sure that you are using the right documents. The AI model trusts that users will provide authoritative training data. If you add pure "marketing fluff" to the content library, you run the risk of having that appear in your AI-generated answers. Incorporating strong documents like official cases studies, white papers, IT security documents and the like can significantly improve both the AI Assistant's success rate in drafting answers as well as he quality of the answers.

Uploading Spreadsheets

Great reference data for questionnaires like RFPs and Security Questionnaires are — surprise, surprise! — completed RFPs and Security Questionnaires.

To upload a spreadsheet, we're going to revisit the Import Content page referenced in the previous section. This time, however, we're going to make sure that "Import Spreadsheet" is selected.

As before you can click "browse your files" or drag and drop the spreadsheet file you want right into the gray uploader box and then click the Upload button. Once you've uploaded the file, you will be taken to an annotation page where you can easily mark questions and answers. This is a critical step as it provides much-needed context to the AI Assistant.

Let's begin by looking at the top part of the page:

At the top of the page, you will see instructions regarding how to label questions and answers and what to do when you are done. We'll see how this works in greater detail in a moment. Now locate the red arrow, which is not in the product UI, but is here simply to help direct your attention. The arrow is pointing to the sheets that make up the spreadsheet. Many spreadsheets will include multiple sheets. When you click to select a specific sheet, the content in the on-screen spreadsheet will update to reflect the data in that sheet.

Now to the annotation:

Labeling Questions

On the right side of the screen, you will see a widget that reads, "Selection Mode." Here, you can indicate whether you are going to label questions or answers or clearly something that has been labeled. We strongly recommend labeling the questions first. To do that, make sure that "QUESTIONS" is selected in the "Selection Mode" box and begin clicking in the cells that have questions. You will want to avoid section and subsection headers as well as other explanatory text. Remember, the focus is on the questions.

Another way to approach the labeling task is to bulk highlight. This can be done in two ways:

• Hold down the mouse button and drag the mouse across all the text that you labeled as a question
• Select the letter on top the column, which will cause everything in the column to be labeled a question

As cells are labeled, they will change color to light blue. Bear in mind Vendorful preserves the formatting of the spreadsheet that was uploaded, so you might see other things that are colored blue. In the example below, the spreadsheet has a darker blue hue for cells with section headers. The lighter blue cells are labeled as questions.

To unlabel cells that were incorrectly marked as questions, you can click "CLEAR" in the "Selection Mode" box and then click the cells in question.

Labeling Answers

The process of labeling answers is almost identical to labeling questions. Just as with question labels, you can click one cell at a time, drag the mouse across a number of cells, or click the letter at the column header. When an answer label is applied, the cell will turn a light green:

Some questionnaires will have multiple cells for the answer. The most common implementation involves a cell with something like a Yes/No answer and an adjacent cell where the respondent can add commentary. When you see a setup like this, you'll want to mark all relevant cells that container answer information as answers. This will ensure that all of the key data is indexed by Vendorful.

There are situations where you might not do this, however. For example, if one of the cells asks for something like a numerical rating of a capability (let's say 1 to 5), then you'll want to disregard it. The reason is that it's so specific and the AI won't really be able to apply a numerical rating from one questionnaire to another one.

Matching Answers to Questions

When questions and answers are in the same row are labeled, Vendorful assumes that they are connected to one another. If an answer is matched to a question, both it and the corresponding question take on a slightly darker hue.

Completing the Process

After you have completed the labeling process, you can click the "Add Content" button. Any unlabeled content will be ignored. All labeled cells will be added to your Content Library. After this process is complete, you will have an opportunity to remove specific question-answer pairs. This can be a good idea when dealing with a few situations:

• An RFP where you provided an answer that is atypical and might have been specific to the RFP in question.
• An answer that is no longer valid.
• An answer that will likely age quickly like "How many employees do you have?" for a fast-growing company.

Adding URLs

For many companies, a website is the content destination where the largest investment is made. So, in addition to allowing users to upload content, Vendorful also allows for the downloading of content from a website. Bear in mind that this isn't a scraper, but rather a way to get content from specific pages that can be helpful. Why would we take this approach? Imagine your blog content were incorporated into your responses. Most organizations use their blog for thought leadership, to opine on subjects related to their market, to develop how-tos, etc. This means that much of the content on a website will not be very effective in terms of providing subject matter expertise. However, by incorporating product pages, solution pages, FAQs, and the like into Vendorful, you can extend the value of your website content.

To pull website content into your Content Library, simply click the "Content Library" link in the left sidebar and then click the "Import Content" button on the page that appears. Now all you have to do is select the "Import URL" option.

The last part is easy; enter the URL and click the "Import URL" button. Vendorful will break the page into chunks automatically. You can add additional web pages the same way.

What's Next

You are ready to rock and roll! Maybe. :-) One thing we recommend you to is consider adding Labels to your content items. Labels allow you to specify which content is used to create answers. If you have multiple products, different positioning by market vertical/geography/etc., or want to otherwise create categories to classify your content, Labels can be incredibly powerful.

At this point, the AI Assistant is ready to roll up its sleeves and start answer your questionnaires and ad hoc questions. Moving forward, you can always revisit your content library to add or remove documents as well as individual question-answer pairs. This way, you can ensure that you are using the most up-to-date, relevant content.

The last thing you need to do is figure out what you're going to do with all that newly-discovered free time....