Thoughts from Boston Conference

I’ll reveal where I heard the term “intelligent business.” I was sitting in on a panel segment at the Imagination in Action event in April, held here at MIT. (Disclaimer: April’s IIA event is an annual conference that I help to facilitate.) Prakul Sharma of Deloitte was interviewing Cindi Howson of Thoughtspot, Manuela Veloso of Carnegie Mellon University, Josep Puig Ruiz, the director of the Nasdaq Data Science division, and Al Bugerai of Snowflake, about the push to build an intelligent business.

“I define (an intelligent business) based on the ease and speed at which any person in an organization can get to the insights, the data they need, whether it's to manage the business to make a decision,” Howson said. “In most organizations, there's a bottleneck. They have to play ‘whisper down the line’ and ask an expert, a data analyst, a data scientist.”

Bugerai talked about how things are handled at his company.

“At Snowflake, we say ‘you can't have an AI strategy without a data strategy,’” he said. “The AI is as good as the data that you have, and the grounds to the truth of the data, and then the semantics of that data where you store the business context.”

“The data for the decisions needs to be accessible, but needs also to be kind of decoded,” Veloso added, speaking to a common challenge. “So data, one way or another, ends up being not understood by everyone, and so there is this problem between the actual access to the data and to access to the person.”

Ruiz had more to say about what this looks like in practice.

“We see many companies that are data rich, but they're insight poor, and I think the key aspect is how do you move from reactive reporting, like using data to understand what happened in the past, how do you move to proactive execution? So, using your data to understand what's happening now, and to make decisions now.”

More on Data Management

The panel continued to go over some of the problems with data, and potential solutions, on the way to building an intelligent business.

“You can't have an AI strategy without a good data foundation,” Howson said. “Most organizations, their data is fragmented. And if we bring in unstructured data, that is even more of a mess. In fact, what a lot of people are calling hallucinations, it's not. It's actually bad data. So you can't go back and fix all of that. Instead, you do have to go use case by use case.”

“I think the customers know about their business workflows and business context very well,” Bugerai said, “and then, in a way, we think that if the business context layer lives at the semantic layer, the semantic layer dictates which parts of the data it will retrieve when you ask it a question.”

As Sharma asked whether this concept is “new,” Howson argued that the semantic layer was pioneered in the ‘90s.

“What is new is the AI context,” she said, referencing the difference between a memory state, and a continual learning state, “the ability to crowdsource it while you're learning.”

Market Equations

Taking the discussion to an economic lens, Ruiz noted that U.S. markets are the most liquid in the world.

“There’s a huge amount of data, so you need to be able to compute it, disseminate it and allow people to use it with very, very low latency,” he said.

Relatable AI Use

I thought that this set of comments from Howson was important in stressing the human side of AI. It’s something I’ve heard before, especially in light of instances like Tobi Lutke at Shopify literally demanding that employees use AI, something that Howson indirectly referenced in explaining that people need to be brought in, to really thrive with AI tools.

“Boards that are more AI literate will have a higher ROI from AI, because they will set appropriate expectations, and will not boil the ocean,” she said. “The ones that just say, ‘I want AI and I want it now,’ or let's take the Wall Street Journal headline, ‘Use AI or be fired,’ which puts fear in many workers' minds, they will undermine and sabotage efforts.”

Of logistical efforts, Howson suggested this approach:

“You have to take a high enough value use case, clean or ready enough data, and start with that, and then scale,” she said, recommending that companies do two more things: bring people along on the journey, and aim for ‘built to adapt’ versus ‘built to last.’

Bugerai went back to analyzing operations at Snowflake as an example.

“You have the HR as a function, finance as a function, but some of the problems each business function faces may not be immediately a priority for the business at scale. So we'll try to prioritize, what is the priority for that function, and how can we bring data to the end user in a way that they can move faster?”

Ruiz talked about some notes on how to maximize the use of AI in a business, and again, this is a principle that is common in the AI world.

“AI should be used to automate some of the boring decision making, and then that frees up humans to actually make the decisions that can move the needle,” he said.

Those are some concrete suggestions about making a business “intelligent,” or, in other words, harnessing AI (people are also using that word a lot in reference to hardware/software setups) to really make things work in a corporate context. Stay tuned for more.