Unsere Erkenntnisse vom 9. jährlichen Big Data & Analytics Summit in Kanada

Good decisions based on bad data are just bad decisions you don’t know about yet

SCOTT TAYLOR – The Data Whisperer brings a unique perspective on mastering the language of business data, which is essential to ensuring your executives are delivering results from their D&A investments. He advocates pursuing “truth before meaning” when it comes to data – in other words, you need to do your hard work on data quality, master data, and related disciplines first before moving on to the more glamorous disciplines around machine learning and AI.

And when your CEO or CFO is unable to fund the “Truth,” you need to get inspiration from Scott on how to make your case so they get it! It provides linguistic tools and guidance to connect any organization’s strategy to its data management strategy. You may want to explore his book for the wisdom he has gleaned from his experience working with organizations like Nielsen, D&B and WPP/Kantar. Or even better: let him do the talking in your company!

If your CEO is unhappy with the results of the CDAO role, your operating model may be wrong

Despite the widespread adoption of the CDO/CDAO role across various industries, a 2023 Data & Analytics Executive Survey finds that only 35.5% of organizations report success in the CDAO role and only 23.8% report that they are doing enough to ensure responsible and ethical use of data in their organization and industry.

So it was refreshing for me to hear a success story from Sandeep Kumar, MMA, CFA , CDAO of Scotiabank’s Retail & Wealth business. Close alignment between data and analytics, technology and business teams; focus on practical analytics; and working with the finance department were key to their success. Dealing with outdated technology, being proactive with data and ethics, and integrating talent were her biggest challenges.

A key observation was that the role of CDAO reports to the bank’s Chief Risk Officer and is clearly differentiated from the CTO. Obviously, banks’ job is to manage risk, so it made sense. However, an even more compelling model would be for the CDAO to report to Enterprise Strategy/COO or directly to the CEO. Unless you’re in the technology business, keep CDAO reporting separate from the CIO/CTO to bring data, adoption, and value closer to the business.

Introduce responsible AI to your organization to prepare for the inevitable regulations

The rapid advancement of generative AI has attracted the attention of regulators around the world. Stanford University AI Index 2023 shows that in 2022, 37 AI-related bills were passed worldwide. Other regulations such as the EU AI Act and Canadian Bill C-27 are forthcoming. These regulations aim to define AI in general, protect individual rights, classify risks and prescribe governance measures including internal processes, impact assessments and penalties.

Several speakers – from government to the private sector to technology companies – emphasized the need for business leaders to introduce responsible AI into their organization. Responsible AI helps organizations design and operate systems that conform to their values ​​and commonly accepted standards of right and wrong. It means embracing the principles of fairness, privacy and security, ethics, accountability and transparency throughout the lifecycle of an AI project, as shown in Scotiabank . It showed that the financial and insurance sectors in Canada were early adopters, while the public sector may be in the late majority.

Recommendations include the creation and adoption of a data ethics framework; explicit review of decisions where people are on-loop, in-loop, or off-loop to balance accountability, safety, and speed; and use of various consortia. IBM for example has an explainability toolkit & recently released by the Singapore Monetary Authority Veritas Toolkit 2.0 for financial institutions. Done right, responsible AI can help companies scale their AI efforts, identify and fix problems early, and make it easier for businesses to innovate.

Design explainability in AI models to simplify business conversations and gain adoption

It was very interesting to listen to Eric Lanoix, FRM , an applied mathematician and problem solver, on the topic of increasing customer, corporate and government confidence in the use of AI in highly effective and regulated applications such as lending. Approving or denying someone’s loan application using an AI recommendation gives auditors and regulators a better understanding of why — and that requires your model to be fully interpretable.

High performance models such as neural networks are low interpretability & will not stand up to the new regulations. He shared an example formulation of a fully interpretable basic ExNN model that resembles a logistic regression and uses three factors – creditworthiness, asset price, income – that influence credit decisions. I am constantly reminded of the importance of business translation skills to ensure stakeholders understand AI. Eric’s Quantitative Risk Analytics team provides innovative quantitative finance solutions to both internal and external partners, including an innovative Model-as-a-Service. It’s worth exploring further.

Innovation comes in different forms – you can innovate with approach, design, tool or technology

For me, innovation means finding valuable ways to solve new and existing problems. I’ve often heard that innovation requires new technology, more money, or a laboratory – while it’s true that some innovations require these resources, it’s often more important to be imaginative and persistent. Here are four useful ideas I picked up at the event that could lead to innovation:

• Olga Tsubiks , Director, Strategic Analytics & Data Science, RBC, shared her thoughts on driving strategic workforce planning using Big Data. One of their key ideas was to look at the state of your business (rapid growth, steady growth, decline, reinvention) and then use relevant data (market leads, customer opportunities, workforce data, and acquisition data) to gain planning insights. Infocepts Employee360 supports such an analysis.
• Paul Moxon , Chief Evangelist, Denodo shared his thoughts on adopting a logical data architecture design due to the fast growing data and rapid technology evolution. Organizations should accept that the future is distributed, diverse, and the business view of data is separate from physical data storage. This can be achieved through the design of data structures in organizations that connect to various systems to receive real-time data.
• Almost everyone faces data quality issues. Vicky Andonova Anomalo, shared her thoughts on why outdated data quality approaches no longer work when scaling data quality. I saw a demo of the tool and could understand why rules and metrics-based approaches are no longer sufficient. You need to leverage automated machine learning models, and Anomalo has a remarkable tool for unattended data monitoring.
• Microsoft introduced Cloth and its capabilities . With OneLake, OpenAI Service, and an integrated ecosystem, Microsoft enables data and AI innovation to help companies achieve results faster while reducing costs. The speed at which these technologies are evolving should prompt IT leaders to reconsider their decisions to invest in and innovate in technology. The questions in the room made me feel like several companies were wondering if their data was being used beyond their own needs and limitations.

The purpose is important

I would like to conclude with observations from two thought-provoking sessions:

Teresa D’Andrea , General Manager, Service and Data Modernization, Transport Canada – Transports Canada shared his approach to service design and modernization. It is based on the concept of design for humanity, where everything we build is designed to serve an end-user goal, while rules, tools, and data are a means to achieve those goals. I appreciate their focus on “consistent user experience” and giving the internal teams the time and space to evolve the backend according to user needs. How many leaders think like that?

Dan Kershaw & Vanitha Lucas shared the partnership story between Namely for the Social Good & The Furniture Bank . This is an excellent example of how data and the power of storytelling are used to influence human behavior – to create awareness and provide means – to make an impact. What is amazing to me is that working here has been done by volunteers – including Hemal Sheth an Infoceptian – from 7 countries, each possibly driven by the goal of doing good.