Support
Contact Us
If the questionnaire is a Google Sheet, you should be able to click in the cells and start answering questions. We automatically infer the structure of the document (e.g. the questions you should be answering) from the content.
If the questionnaire is an Excel file, the easiest way is to open it in Google Sheets and then click File -> Save as Google Sheets. You can re-export as an Excel file later. WARNING: If you open Excel file in Google sheets and it looks like an imported Google Sheet, it still might not be imported. If you see a green ".XLSM" or ".XLS" badge next to the title, you still need to import as a Google Sheet as described previously.
If the questionnaire is a Google Doc, Word Doc PDF, or web portal, highlight the text of the question you want answered, and we will give you answers in the extension that you can 1-click copy. Unfortunately there is no way to answer multiple questions at once just yet.
We are not yet able to support questionnaires that you cannot open in a browser.
We natively support Google Sheets. We support all formats and layouts: multiple tabs, dropdown boxes, merged cells, etc.
For Excel Files, please first convert them to Google Sheet by uploading it to Google Drive and clicking "File -> Save as Google Sheets" (do not skip this step!).
Office365 support is coming.
For everything else, you can highlight the text of the question you want answered. We will answer the question in the extension, with a 1-click copy so you can easily fill out the questionnaire.
We can index the following file types: Google Drive files (Docs, Sheets, Slides, etc.), .doc/docx, .pdf, .xls/.xlsx, .ppt/.pptx, .txt. We can index previous questionnaires that are formatted as either a sheet (.xls/.xlsx/Google Sheet) or doc (.doc/docx/Google Doc). We automatically parse out questions and answers in those documents. We can index publicly available websites and allow you to crawl a tree of webpages to easily and quickly pull in all relevant documentation. This works particularly well for API documentation.
We cannot index the following file types: Any image (.png, .jpeg, .svg, etc.) audio (.mp3, .mpeg, .wav, etc.) or video (.mpeg4, .mov, .avi, etc.) format. Zip files (.zip, .gzip, .tar, etc.). Any application-specific editable filetype (Adobe Illustrator .ai files, Photoshop .ps files, Sketch .sketch files, etc.).
We minimize hallucinations by only answering questions we are confident are correct. In the future, we will offer visual indicators in the questionnaire and the extension itself on confidence level to flag answers that might need additional manual review.
Shoot over an email to contact@quilt.app and we will set up a custom demo for you and your team and get everyone onboarded!
We use a proprietary ML pipeline to index the documents you share with us according to their context. We answer questions straight from the indexed documents based on the text that we determine is the most relevant answer for the question. You can click through each question to see what source documents the answer is coming from.
We can only give answers from the documents you share with us. We do this to minimize hallucinations and false positives. Think of us like a brand new SE that is an expert on your documents only. In order for us to answer more questions, be sure to share more documents that could provide the right answer to those questions. Documents that tend to work well are typically information-dense: previously completed questionnaires, security certifications, and developer documentation.
If you have additional questions, please email contact@quilt.app for a detailed setup and walkthrough.
We have a proprietary way to determine the best answer from the source documents you share with us. You can click through to see where the answer is coming from in the "Answer History" section in the extension, as well as other answers we considered (coming soon).
We periodically re-index your source documents so that answer stay fresh and up to date. The frequency can be anywhere from daily to weekly depending on the source and the number of documents you have indexed. All of this happens automatically so you don't have to worry about it.
You can see previously answered questions in your "Answer History" in the extension. We will soon launch a way to interact with old questions and answers as well as indexed questions and answers in the web application.
For RFPs and security questionnaires, providing correct and accurate answers is only half the equation. We also have to provide those answers in the right context: be selective about what to say and what not to say, make sure the tone and phrasing matches the context, etc. We are constantly tuning our models with this in mind and have a lot of proprietary strategies around inferring the right context, but we are not perfect! We will continue to get better over time.
The quality of answers depends on the quality and completeness of the source documents you share with us and the kinds of questions that are being asked. In the best case - for instance, product questions for a product with very robust product or developer documentation - we can get to near 100% completion. However, there may be some questions we just are unable to answer due to not having enough context on the question itself or not having the right data sources shared with us.
We aim to get as close to 100% completion as possible. To this aim, we understand that in addition to getting answers factually correct, context and phrasing is also important. We are constantly tuning our models so they do the best job of inferring the correct context and providing correct answers phrased in the right ways.
We are building a way to manage questions and answers in the web application and give feedback to the model itself.
