Botminds adds “prompt automation” to accelerate document understanding

The artificial intelligence of large language models and ChatGPT triggers panic in some and a tendency to stare rabbit-like into the oncoming headlights of “the next big thing” for many others. But HFS Hot Vendor Botminds has wasted no time finding a real-world business use for this kind of generative AI, adding prompt automation to speed up intelligent document processing and making it possible for users to have a natural language conversation with their data.

“Prompts” have become hot with the rise of OpenAI’s ChatGPT and the likes of Google’s Bard. Prompts are the questions the user asks of the AI. Phrasing these questions smartly is becoming an art form of its own, with full-time roles such as “prompt engineer” already advertised on LinkedIn.

Business users can use natural language to create reusable data extraction models

Now Botminds, named an HFS Hot Vendor in January 2023, has added prompt automation to the AI platform at the heart of its intelligent document processing (IDP) and other automation capabilities.

The user simply types their query about an uploaded document or documents (e.g., I’m an HR manager, and I need to extract the name, email address, and telephone number from each candidate’s CV), tests their query, and when the query proves effective, saves it as a new model. By stitching together a series of prompted “models” in extraction pipelines, a business user can create bespoke and reusable data extraction models that would otherwise demand the creation of templates and the annotation of multiple terms to train the AI. The prompt model’s greatest advantage is its speed–it can accomplish these takes in a few minutes versus the few hours previously required to create templates and annotate content.

Chatminds allows users to have a conversation with their data

Botminds has also launched Chatminds, allowing users to upload unstructured data such as PDFs, emails, word documents, images, and audio files and ask questions across all the data.

With “chat” and natural language, users can interact with the information stored in uploaded company documents. You could, for example, type a request such as “Find the stage and start date of our Lithium Project” and get an answer such as “The Lithium project started in 2022 and is in the construction stage.” But you need not end there. You could continue conversationally interrogating the data with “What is the latest installed capacity of the project?” and receive the relevant answer.

Each response comes with a link to the source of the answer within the database, making the chat a new way of searching company data and building on Botminds’ established capabilities in searching across documents. Read our Hot Vendor report for more on the search-led approach that enabled Botminds’ switch to chat.

The Bottom Line: It’s easier said than done: A flexible, modular architecture is required to respond to the rate of change AI drives.

We are getting closer to being able to say, “Computer, do x, y, or z for me” in a very Star Trek way. But there are limitations; for example, Botminds admits the prompt model is not best suited to the most complex documents. But we applaud Botminds’ speed at bringing these generative AI applications to market, where customers are already enjoying the benefits.

IDP vendors are falling over themselves to enable generative AI capabilities in their platforms. The fact that Botminds has been the first to brief us on getting this to market is attributable to its modular architecture’s flexibility and its willingness to respond. For Botminds, search is dead—long live chat!