Artificial intelligence (AI) imposes a layer of discipline that’s unexpected for many companies. That’s good news, because the discipline requires us to rethink the collaboration, metrics and talent necessary to develop the focus that AI needs to succeed.
A successful AI adoption is distinct from big data, its predecessor. Big data was a solution in search of a problem: Enterprises were encouraged to gather their data and see what it could turn up. AI requires the opposite approach. First, identify goals and underlying drivers. How might your organization become more competitive? How can it deliver more business value?
AI is also different in that it views data as a strategic asset. What exactly is your organization going to do with the new data? With AI, you need to know the answer.
Step by Step: Creating a Focus
Taking the following steps before your organization launches AI initiatives will ensure that it has sufficient focus to succeed.
1. Reimagine your data foundation. Finding the answers in AI begins with a strategy for a new data foundation. Making sure the foundation has the computing, storage and analytical capabilities engineered for purpose is only part of the equation. Perhaps more important is the shift in perspective that’s needed: Instead of using data to track and measure how business functions perform, the new data foundation lets your organization perform with data. Big difference.
It’s at this point in conversations about data that ride-sharing giant Uber typically comes up. And with good reason: Everything Uber does is all about its data. Its data foundation processes trillions of Apache Kafka messages per day. It stores hundreds of petabytes of data across multiple data centers and supports millions of weekly analytical queries. To gain context for autonomous decisions, Uber’s in-house platform Michelangelo surfaces and manages metadata and ontology essential for data-driven performance. All that horsepower lets them know just where we want to go at any given moment and matches us with willing drivers.
It’s not just born-digital businesses that understand the concept of performing with data. A few leading businesses such as farm-equipment maker John Deere are similarly data driven: The 179-year-old company created an open platform for small agricultural start-ups to leverage data and analytics. AI presents a new way of thinking about business with a new data strategy, and the data foundation represents your new building blocks.
2. Encourage collaboration between IT and business functions. Most enterprises continue to silo their technology efforts, with the data organization reporting to IT and analytical teams reporting to business functions. The result is a chasm that both sides need to cross, yet often don’t.
The two functions typically operate independently. For its part, IT drives data and analytics modernization within its own smaller spheres. There’s no joint definition of purpose with business functions, and no effort to lay out a clear vision for business value creation.
Analytics teams focus on the individual functions they support. The teams often bring in third-party or new data they’re unable to get through the IT-managed data pipeline. While the efforts address small problems at the departmental level, organizations miss out on the opportunity to tackle larger, more strategic challenges such as improving operational efficiency at local process levels.
By finding ways to more tightly couple IT and analytics, organizations can begin to develop the integrated data foundation they need for enterprise-wide AI adoption.
3. Examine your organization’s data quality and how you measure data success. Is the data you have ready to support your goals and intended business outcomes? For example, if your goal is to generate AI-driven recommendations that help customers make the right product choices, then the data brought in to train and test the AI component has to have high relevancy, meaning that in addition to having the requisite information it should be highly correlated to business outcomes without noise due to operational or system errors. The data should also be diverse and complex enough to demonstrate consistent value over time. (For more on data’s role in your AI effort, see Part Two, “Measure Your Organization’s Data IQ.”)
4. Rethink the talent assigned to AI projects. Within data organizations, team members with deep business knowledge are often in short supply. Yet the design thinking and outcome-driven approaches that business leaders bring to the table are essential for successful data and AI initiatives. The most successful AI programs understand how they impact the business. It’s common for organizations to fill CIO and CDO roles with individuals whose resumes include data experience rather than the business knowhow. A better option is to bring in the head of sales or head of marketing to run the CIO/CDO/CAO function for two or three years. You can always hire a strong CTO who can work closely with the CDO/CAO and assist in making the right technology choices.
Choosing which programs to run, however, and where to focus investment and budgets is best served by someone who knows the business and will return to run it — and take advantage of the AI systems being built into the process.