Using Generative AI for Decision Intelligence With Pyramid Analytics
About a decade ago, the arrival of what the industry called “self-service business intelligence” (SSBI) was heralded as a game-changer for BI software. SSBI helped make data-driven decision making more accessible by providing regular business users with the ability to explore interactive dashboards themselves, lowering dependencies on the assistance of technical teams. In retrospect, it seems that SSBI failed to deliver on these promises.
The challenges around SSBI are numerous, with the most glaring one being that, in order to access the insights it delivers, users still needed help from data scientists and admins to set up data exploration environments. Free reign to find answers to open-ended questions was still elusive.
Also Read: More than 500 AI Models Run Optimized on Intel Core Ultra Processors
Fortunately, with the arrival of generative AI and chatbots powered by large language models, businesses now have a way to empower decision makers with a new generation of conversational BI tools. As “GenBI,” as it’s been called, matures, decision intelligence has become far more accessible, but it’s also becoming more reliable, ensuring users can trust the insights they deliver.
In a recent interview with TechRound, Pyramid Analytics Co-founder and Chief Technology Officer Avi Perez spoke about the way his company has implemented AI capabilities and the challenges his team has faced in delivering maximum decision intelligence value with LLM integrations.
Insights Grounded in Truth
While LLMs are great at processing data according to instructions, they do tend to be “lousy at mathematics, and by extension, lousy at analysis,” Perez said.
On the other hand, what LLMs are especially good at is understanding what users are trying to ask. Armed with that understanding, they can go about formulating the most pertinent questions to ask BI tools on behalf of users, so that the software can serve up the insights they’re looking for.
Perez likens this to the process of baking a cake. While generative AI is highly skilled at creating recipes based on what the user has in their fridge, it can’t actually bake the cake itself. If you want to automate it, the actual task of baking it must be handed off to a robotic chef, so to speak. In that case, all the LLM does is provide the instructions to the robot, which goes ahead and follows them to the letter.
Pyramid Analytics does the same sort of thing with its GenBI tools. According to Perez, what happens under the hood is that the user asks a question, and then Pyramid goes and finds the relevant data to respond to it, handing a description of this data to the underlying LLM, which understands both the question and how to find the answer.
“It comes back with a recipe for the kind of question that I’m asking and therefore how to find the answer,” Perez explained. “Pyramid takes the recipe and goes and bakes the cake and effectively runs the query for the user.”
Also Read: Understanding Artificial Intelligence’s Role in Accelerating Africa’s Healthcare Momentum
Perez said this approach ensures the insights generated by its BI can be relied upon. But the software also goes further, formulating those insights and visualizing them for users, before responding.
Another benefit of this approach is security. According to Perez, Pyramid’s solution is “running in the sandbox of the customer on their private data.” That means the LLM never actually gets access to the customer’s underlying data – only a description of the data – meaning there is no opportunity for any sensitive information to be leaked.
LLMs for Every Occasion
Another key advantage of GenBI is its flexibility in terms of using different LLMs for different analysis types, verticals and business areas. While other AI integrations into BI tools involve black box LLMs managed by the software providers themselves, Pyramid has taken a more open approach, allowing users to toggle between models as they like.
Perez said it’s clear that we’re heading towards a world where companies will have access to hundreds of different LLMs that have all been fine-tuned for different applications.
“So maybe there’s an LLM that is tuned for accounting, another one is tuned for marketing, maybe an LLM is tuned on healthcare, maybe an LLM is tuned on insurance data,” he said. “Therefore, it makes a lot of sense that you’re able to switch out the LLM based on the data that you’re looking at.”
Perez said these models tend to be much smaller, making them easier to train and cheaper to run. And because they are trained on a very specific kind of content, they can also deliver responses much more quickly than broad, generic LLMs. As such, he believes the BI ecosystem will see a natural progression towards these smaller, faster and less expensive models that are more narrowly focused.
BI’s Eureka Moment
Eventually, Perez sees the BI landscape evolving into a situation where decision intelligence capabilities are no longer relegated to BI-specific apps and are rather integrated directly into all other types of software, using embedded analytics.
As Perez notes, context switching is a productivity killer, and decision makers are often reluctant to switch over from the standard applications they work in, such as their customer relationship management or accounting platform, to a separate tool so they can perform some analysis on the data they’re working with.
While many people have become used to these clunky workflows, that doesn’t make them like doing it, Perez stated. “If you can take the high-end analytical components and drop them right into the app itself, no one has to go anywhere,” he said. “As you’re looking at your CRM data, you can do the analysis right there on the spot and the two are highly synchronized and highly copacetic.”
This kind of embedded analytics is not new, with a number of classic BI tools having already integrated with other popular platforms in the past. But the concept has not really taken off yet, Perez said, because traditional BI itself is not very accessible.
For embedded analytics to become more widespread, it’s necessary to remove as much friction as possible from the analysis process itself, and that’s where GenBI is going to have a big impact, Perez added.
“The big trick is to do it very effectively, very simply. It needs to talk to the database directly, so there’s no copying and pasting. Connecting this all up to the rest of the story, it’d be fantastic if you can bring the generative BI or generative AI functionality into that embedded experience directly within the application seamlessly,” he said.
“Ultimately, the idea is to get to that eureka point, which is to tell somebody not just what did happen, not why it happened, but what could happen – and what they should do to make it a better outcome, more profitability, lower the cost, better quality, whatever the reason is,” he said.
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]
Comments are closed.