During the James M. Kilts Center’s annual Ann Mukherjee Marketing Summit in April, Professor Bradley Shapiro sat down with Murli Buluswar, MBA ’01, and Suzanne El-Moursi, MBA ’12, to discuss the opportunities and pitfalls of using AI to accelerate business growth. Below is an edited version of their conversation.
Bradley Shapiro: Where have you seen AI help drive growth in your businesses? Is it being used to increase worker productivity, eliminate inefficiencies, create entirely new markets, or something else?
Murli Buluswar: We should think about three levels of impact that AI could have in all of our organizations. Level 1 is the fundamental level of fluency in AI. Level 2 is microproductivity, which in the collective has a sizable impact on how much more freedom people have to do more higher-order thinking and problem-solving.
And then there’s level 3, which is the “big rocks”—your eight to 10 initiatives that collectively, in a highly finite period of time, will improve the return on equity of your firm in a material way.
The conversation gets lost when we focus on just the infrastructure, or when we think of it just as tools and things that people are doing for fun that have marginal but interesting value. In that third level, we need to take a systems-thinking approach to be able to fundamentally reimagine how a particular critical workflow should be different tomorrow than it has been historically.
Achieving large-scale innovation through AI requires imagination, technical aptitude, and operating knowledge, all grounded in cross-functional orchestration. We’re redesigning how decisions are made and even what decisions are made—in essence, redefining people’s roles and their sense of professional identity. Understanding and incorporating this perspective is critical to achieving sustained outcomes.
As an example, in financial services, the concept of risk and controls is a very big issue: The industry is fined $2 billion–$4 billion annually for mistakes. The cost of remediation is significantly more. CEOs lose their jobs and market caps are heavily discounted when regulators take a dim view of banks’ ability to manage processes and comply with regulations.
With generative and agentic AI, banks are able to achieve fully automated, near-real-time detection of 100 percent of errors. This is achieved by triangulating customer claims with what the regulations say and with what objectively happened.
The result is a fit-for-purpose process that is AI-first in how it has been envisioned. The regulators are happy because banks such as Citi commit far fewer missteps—or certainly detect errors before they scale to the point of affecting a much broader swath of customers. The detections happen in near-real time and have full coverage, thus also dramatically shrinking the path to remediation.
Suzanne El-Moursi: My perspective’s a bit different because my company, Brighthive, is essentially a data team in a box. What I’ve observed is a problem with AI adoption today. While 97 percent of enterprises are deploying AI agents, only 10 percent can scale them successfully.
The reason is not their models. It’s the data engineering work required to get the data AI-ready. The root of the problem is that data work is manual, fragmented, and undocumented. The processes that used to ensure data quality are outdated, and the SaaS tools themselves are part of the fragmentation process.
And the amount of data only keeps growing. If you look at upper-middle-market to enterprise companies, the workload of data to people is 100:1, meaning the time it takes to execute a full data-management life cycle far outweighs the number of staff available to perform these tasks.
Getting to clean data—data that is governed, accessible, usable, and capable of powering clean-data pipelines—is the grunt work of data engineering. At Brighthive, we’re able to automate the entire data life cycle inside a company’s own infrastructure at less cost than employing a human data engineer.
The goal is not to put data engineers out of work but to remove this grunt work to allow for a focus on the higher-order thinking and problem-solving that Murli mentioned.
My cofounder—a Chicago alum in data science—and I started the company in 2018. We believed in the productivity of an agentic workforce before ChatGPT arrived on the scene.
We’ve mimicked data engineering and analysis, creating seven AI agents that work in unison, just like a data team. Everything we’re touching is data that already exists. The old processes were cumbersome and slow and had to be 27 steps long because we didn’t have generative AI. At some point, they just became broken.
Companies don’t want to be broken. In every sales call I’m on, the leader is telling me they have a clean-data problem. AI needs clean data. Regulatory needs clean data. What we’re talking about here is beyond chatbots. We’re talking about liberation from broken, long processes that are costing us money and blocking successful AI adoption.
Take, as an example, a big CPG brand we just started working with. They get data every hour. At this point, they want to push more of the product. The pricing information that’s coming in from each channel, the promotions at the shelf level, and the performance of the paid ad campaigns—you’re talking about a model that has about 10 dimensions on a daily basis from each vendor and wholesale and grocery stores for this brand.
How are we going to reduce time to insight? For me, that’s always the question, whatever you’re trying to get to and whatever industry you’re in. Insight is the starting point for innovation; if we can liberate human beings from grunt work, we allow for high-value, high-human-intelligence work to happen in abundance.
I’m a technologist. I like solving human problems with tech. My goal is to get you to time to insight so that you, as the business leader, can take the necessary action, make the next $2 billion, or whatever the case is.
If an AI company can show that they can get their clients to time to insight faster with trustworthy data, that’s quality AI.
