Previously published on Nuclia.com. Nuclia is now Progress Agentic RAG.
This week, as I was researching for my next blog, I took Nuclia for a spin to help me answer a burning question: "What are the real business benefits of GenAI?"
Like many people embarking on such a mini project like this, I took to Google to start looking for insights and stats from my usual favourite sources—Forbes, HBS, British Chamber of Commerce—and of course I started with keyword search.
Questions such as "Businesses using AI", "How many companies have embraced AI?", "What's the percentage of companies adopting AI?" and others yielded about 1,700,000,000 results, and—as Google proudly told me—all in 0.39 seconds.
Great! It's a hot topic and there's lots of content. But now I have to trawl through and read all the articles (well, at least maybe the top 10, as there's a deadline), watch the videos, decipher the statistical significance of the stats and somehow summarise it all into something useful.
So, in the interests of getting a balanced view, I uploaded 12 resources from credible and reliable sources to Nuclia. These included videos, reports, URLs with tables and charts and many, many stats. This took me all of about 25 seconds in actual upload time, and it probably took me longer to do the copy and pasting than for the platform to process the resources.
In addition to uploading the resources, I chose my data region (Europe), embeddings model (English, Nuclia model) and which language learning model (LLM) I wanted to use (Anthropic Claude 3 Opus, my personal bias as it's new and really, really good!)
Below is my resource list, clearly labeled. I'm also notified where there are embedded links or other reports.
Now onto searching. This is why generative AI (GenAI) is so magical. Search doesn't stop at results–we want answers. And, as a sales person, I want to know from those answers about the business benefit; no feature based arguments around here.
My first question was, "What is the benefit of GenAI to a business?"
That was a good start. I went on to chat with the AI to find the answer about specific use cases.
What I find particularly beneficial about generative answers is the use of verbs—quite literally, what action we can drive. I often talk to our clients about seeing Nuclia as the enabler of new things—not replacing something a human was doing previously but allowing you to do more with your digital assets while always keeping a human in the loop, or creating opportunities that weren't even on your radar or would have been too costly and taken years of human effort to achieve.
Next, I asked Nuclia a question that I knew the LLM didn't have enough input to answer, to prove out our "no hallucination" promise: "Which companies should I target likely to be most interested in GenAI?" Gold star for this answer.
Lastly, as I have to stop playing around with my favourite toy, I asked Nuclia "What's the best pizza topping?" If you managed to catch any of the reports about the training data used by Anthropic, this was the question the engineers used to see how well Claude searched, and it was the point at which the LLM responded by supposing it was indeed being tested. Metacognition had arrived. For my case, though, I've asked it to show that we don't leverage the LLMs for any content or training data, only to derive context from your own data. Another gold star Claude, you've done well.
If you'd like to see how Nuclia could work with your data and deliver the business benefits we've outlined here, check out our demos or get started with a free trial.