Business Leaders Make the Case for Data Science

Dataiku Product, Scaling AI, Featured Marie Merveilleux du Vignaux

Led by Conor Jensen of Dataiku, a panel of esteemed industry leaders discussed the case for strong data science foundations in their respective spaces. Jensen was joined at the Toronto session of the 2023 Dataiku Everyday AI Conferences by Amy Korosi, VP of Data Enablement and Automation at Hudson’s Bay Company, Kelly Chambers, Business Intelligence Analyst at the Alberta Energy Regulator, and Herve Riboulet, Director of Cargo Analytics and CRM at Air Canada.

→ Watch the Full Panel Session

Building Strong Data Science Foundations

Herve Riboulet led off, describing how Air Canada has been able to deliver data science capabilities in their organization. “What we do is for the business users,” he said. “Being close to these clients that already back data science in AI or analytics practices that you might already have in an organization helps make sure that disciplines and departments are really working closely together.” Riboulet believes that data scientists, as they produce more data more easily, partner with data visualization experts, eventually bringing it back to the business. He added, “It’s also very important that the data scientists get to understand more of the business side and what they’re actually solving.”

everyday ai toronto talk

Kelly Chambers continued in complete agreement, on behalf of the Alberta Energy Regulator. “We provide self-service analytics and self service data prep and are moving into AI and machine learning,” she began. We also build products for people who can’t.” She stressed the importance of having experts throughout the organization. “To use the tools, and create the data, and do the analysis themselves, the data analysts have the subject matter expertise. They can look at the outcomes of the analysis to see if it’s really on the money or not.” Chambers and her team as IT do the best they can to understand the business side but acknowledge they aren’t experts, showing the real opportunity in front of them.

The collaboration between IT and business users is key to getting traction in building the business case for data science.

-Kelly Chambers, Business Intelligence Analyst, the Alberta Energy Regulator

Amy Korosi of Hudson’s Bay Company added, “The reason it’s so important to collaborate with business is that trust is earned. You’re there as a service, and if you don’t show you know the business, it’s hard for them to trust you to start that journey. You have to come to the table prepared and understanding the business, so that they trust you’re going to help them solve the problem and not make it worse.”

Jensen pressed her for more detail. Korosi responded, “One of the challenges is that people worry that you don’t understand the business, but also that you’re going to take their jobs.” She gave a real-world example. “One of the projects we’ve been working on is looking at product descriptions and extracting the keywords so that it’s easier to find the products you’re looking for. The team that works on the descriptions and does the tagging was worried about losing their jobs. But someone still needs those through. So we assured them that we could speed up that work to save time to do the work they actually like doing.”

Gaining Trust With Partners

“I’d love to hear examples of where collaboration helped build trust,” Jensen said.

Chambers spoke up first. “We’ve been with Dataiku for about six or seven months now. We worked very closely with key stakeholders in our business and so they were involved from defining requirements, to evaluating responses, and doing the trials for the product,” she said. Her team involved them in the entire implementation cycle of the project, and she mentioned that in doing so, her team was able to get a better sense and knowledge about what her client values, and where they might be struggling.

She continued with more industry-specific knowledge, specifically about data governance. “We’re a public sector organization,” she said, “we have internal audits that go through our processes. We get audited by the office of the attorney general in Alberta. We need to pass our audits, but working hand-in-hand with the business throughout has been invaluable and helped build that trust.”

Riboulet’s team is trying to make data science and analytics commonplace so his clients won’t find it so abstract. “It’s an integration into their day to day; it helps a lot to build that trust and that relationship.” Indeed, he summed up the very nature of the conference in which the panel was speaking.

AI and data science become something you don’t even notice. That’s what you want. It’s seamless. 

-Herve Riboulet, Director of Cargo Analytics and CRM, Air Canada

Jensen echoed the importance of data scientists spending time observing business team processes to fully understand their needs — something he refers to as a research trip. “I encourage my data science team to spend a week or two sitting with that team watching what they do all day, because you don’t know how they’re going to use your model if you don’t understand the process.”

Managing a Variety of Data Projects

Riboulet explained that true partnership only happens if data scientists gain a deep understanding of what their teams need. This understanding comes from asking the right kinds of questions. “We don’t go out and ask, “Do you have a data science problem you need to solve?” We’ll ask, “What kind of decisions do you need to make today that will affect the future?” This line of questioning swiftly unlocks client needs and helps uncover problems. Riboulet firmly believes that tangible answers to these questions generate creativity. 

When the panelists were asked about specific project management techniques or agile frameworks, Korosi was the first to chime in with her thoughts on being disciplined with data. “My heart is in data,” she began, “Data people should work with product people. We have hundreds of customers in our own business using the sites that we’re creating.” She has seen that for many industries, internal products tend to be of lesser quality than the external ones — a trend that she hopes will change in the near term.

I hope data becomes so disciplined internally that we serve our employees just as well as we serve our customers.

-Amy Korosi, VP of Data Enablement and Automation at Hudson’s Bay Company

Chambers agreed. As a nonprofit public sector organization, Alberta Energy Regulator is responsible for regulating the oil and gas activities in the province of Alberta. “Our data landscape and our users are very technical, from economists, geophysicists, geologists, water scientists, land and soil scientists, and other domains of expertise. We end up with a data environment of about 875 production databases that we use as a regulator,” she said. With data fully in the cloud and initiatives to build a more modern architecture to break down silos, it’s a substantial lift to create strong governance frameworks to manage that data.

What we’ve been able to do with Dataiku is give people the ability to pull all that data together in a more simplified way and generate the insights they need. 

-Kelly Chambers, Business Intelligence Analyst at the Alberta Energy Regulator

Understanding Strong Governance Strategies

“Do you find that your stakeholders are more receptive when you mention governance?” Jensen asked the group.

“Not necessarily,” Chambers responded. “The biggest thing is finding balance. There’s a tension in governance between losing ability and flexibility to pivot, while also having control over the data assets. It’s trying to find an adaptive framework where we don’t sacrifice the agility people need to make turnaround times for analytical asks, but we also can’t change the definitions.” Her team needs consistent structure around what business subjects are understood and calculated, because the insights they generate later goes to other industries, the government, and the public. “We need to know the truth about the information we’re disseminating,” she said.

Korosi compared good governance to something that made the audience smile: the DMV. “You get in your car, you know where you have to go, you know what you have to do. You know the rules,” she said. To her, this is what strong governance looks like: clear patterns of behavior that everyone inherently understands.

Governance is where Dataiku is powerful. When people open it up, they know how to use it. They know the rules.

-Amy Korosi, VP of Data Enablement and Automation at Hudson’s Bay Company

Chambers sought to remind guests what the current state of affairs is surrounding governance. “We keep forgetting that people keep using Excel. There’s no control with it. We have an opportunity to eliminate that mess by having something that is more governed but still accessible. That’s where we have to change our mindset,” she said.

On the topic of spreadsheets, Jensen mentioned the concept of cloud-based spreadsheets. It seemed as though their usefulness was up for debate. Korosi responded, “Sometimes you need speed, sometimes you need to prototype something, but once it has to live on and be supported long term, you have to move to something more disciplined.” Having a set of critical spreadsheets in the cloud can be complicated because only a handful of individuals tend to know how they were created and how they track data — if something breaks later on, it can be in Korosi’s words, “a disaster.” It’s critical to agree on when it’s time to migrate before things get too far.

Chambers agreed. “A cloud-based spreadsheet can lose its value very quickly. From an IT perspective, we want to build a new cloud-based data architecture where everything’s integrated, but we’ve also got all this mission-critical data being developed by business users,” presumably outside of the architecture. This creates a worrisome scenario in which the data might not be properly backed up or handled properly. 

The Challenge of ROI

“We haven’t touched on cost yet,” Jensen said. “How are you capturing the ROI of investments in a platform, which is hard to tie to discrete business things?”

Riboulet responded, “Not everything is quantifiable. Sometimes it’s a business decision that is made. We also don’t have to justify everything — we just need to make it easy. If we’re going to use a tool and start building things, then people come back telling you they use it everyday, that’s ROI.”

Korosi feels that ROI is best expressed in partnerships with others. “I’ve found success in being part of other projects,” she began. “We sell big solutions, but if you just go out and look at projects and deliver small pieces of them, you start to build those relationships and that trust. You don’t have to come up with an end-to-end solution. It’s easier to approach those people and ask how you can help.”

If you spread out and do a lot in little ways, you can become trusted partners. 

-Amy Korosi, VP of Data Enablement and Automation at Hudson’s Bay Company

Chambers closed out the discussion with an excellent example of the efficiency her organization gained by implementing an AI solution. “Two of our business users were given an analysis project to complete in a two-week timeframe. They rebuilt the project in Dataiku, and between the two of them replaced thousands of lines of code, discovered and fixed bugs, and added new features in two days. The rest of the time they were able to do the actual analysis they needed to do,” she said. For more detail, check out some of the data visualization work your team can do with Dataiku.

Through topics surrounding designing capable data science foundations, building effective partnerships, evangelizing strong data governance, and ensuring a good return on investment, the panel made an effective case for data science in any organization. Even though the initial lift might be heavy, time-consuming, and require a slower pace, the upside is dramatic and can perhaps lay the groundwork for a more data-driven future for all industries.

You May Also Like

How Dataiku Turns GenAI Into Business Gold

Read More

Riding the AI Wave: OpenAI’s GPT-4o and Beyond

Read More

Build a Generative AI Chatbot With Snowflake and Dataiku

Read More

Beyond Text: Taking Advantage of Rich Information Sources With Multimodal RAG

Read More