First of all, with Augmend, you leverage the full multi-modal capabilities of LLMs. So handwritten text, languages and pictures are all easily processed. Secondly, the way Augmend is architected makes it possible to fine-tune its application to use cases in any functional or industry domain where sufficient public data is available.
Here are some example use cases where Augmend really shines
- The most obvious use case is
automating extraction of data from documents to Excel. If you are doing
this
manually today, you can let Augmend do it, and free up your time for value added tasks.
This
use case has several application scenarios:
- You do not have a database so you currently store data in files, and manually extract it into Excel
- You have a database but Excel is your staging solution, i.e., you first manually extract data in Excel and then it is uploaded into an enterprise system. For example you may be using Winshuttle or ProcessRunner to upload data from Excel into SAP
- You have a bunch of highly complex documents, such as letters, contracts, SOPs, Work orders, etc., and you want to structure data in them for quantitative insights. This is different than being able to simply chat with the documents. Refer to the articles on Augmend vs RAG and how stuffing many documents in your RAG pipeline is not useful in this scenario. Also refer to the CRL analysis article for an example
- You not only want to extract
data
but apply some kind of standardization or harmonization to the dataset so
that
you can easily do quantitative analysis
- Your data needs to be extracted from documents that have different languages. So your data may end up having an mix of words, e.g., fat, vet, fett, etc.
- Your data needs to be extracted from documents that come from multiple stakeholders that has variations, e.g., labels on consumer goods products or hand-written notes of operators
- You want to categorize your data as you extract it. For example as you extract data you want to attach a service or parts label to your data
- You want to create metadata from pictures, i.e., there is no text to extract but you want create standardized metadata on pictures, e.g., information on actions, emotions, colors, textures, entities out. Simple example could be description of a shirt, package, etc.
- You want to enrich metadata using external datasets. Here you can fully leverage the ability of Augmend to tap into the publicly-available WWW information
When Augmend might not be the best fit
Of course, no tool is perfect for every situation. Here is when using Augmend will not offer a complete solution or might not be the best fit:
- Augmend focuses on data extraction into structured format. So, it does not provide any analytics capabilities. This was conscious choice as typical analytics use cases need combining the structured data extracted by Augmend with other datasets. This is best done in Excel or other advanced tools such as PowerBI, datalakes, etc.
- Augmend is an overkill for high-volume extraction of super-standardized documents. If your document format hardly changes you are better of using time-tested OCR techniques
- Augmend for the foreseeable future will operate in the cloud and use third party LLM models from reputed vendors such as Google. There are guarantees that Google gives on not using your data for training models. But if you absolutely cannot live with this because your data is very sensitive, then Augmend at least right now is not the right solution for you. This is probably true only for exceptional situations – most enterprises make use of some kind of cloud already so your data is already in the cloud. For example, if you are using Office 365 your data already resides in the cloud.