Every day we see new stories about the capabilities and sheer momentum of artificial intelligence. AI is as good as experts at detecting eye disease and it is changing education. Research firm IDC expects cognitive and AI spending to reach $52.2 billion in 2021. But what are some of the indicators of successful leaders in organizations using AI?
5 Key Attributes of Artificial Intelligence Leaders
Here are five characteristics that successful AI leaders tend to share:
They commit to clean, integrated, high-quality data
There are many criteria that are necessary for businesses to derive value from AI, but one of the most important, says Ashok Srivastava, SVP and Chief Data Officer at Intuit, is clean, quality data. He says, “One of the most effective ways to reduce risks for AI and machine learning products is to have a high-performing data platform.”
One of the most effective ways to reduce risks for AI and machine learning products is to have a high-performing data platform.
And, according to a recent article in the Harvard Business Review, the potential impact of bad data is cumulative: “The risk is that a minor error at one step will cascade, causing more errors and growing ever larger across an entire process.”
They choose high-impact, clearly defined use cases
Choosing the right use case is a combination of art and science. The art is to choose a problem that is clearly defined and will be viewed as a win for the business. The science is to choose a problem that is rich in data and that includes enough characteristics and patterns that deriving insights is possible.
Problems in the way algorithms are behaving can sneak up on you unless you have things in place to catch them.
The sweet spot lies somewhere between confirming the obvious (the pitfall of many early AI projects) and trying to solve for events of such low probability that they would never emerge as a pattern. At the same time, it is important to scenario-plan for edge cases. Mike Curtis from Airbnb says, “Problems in the way algorithms are behaving can sneak up on you unless you have things in place to catch them, so think through the edge cases and the unintended implications. Don’t just let people only think through the business goal, or you might miss something important.”
They set clear priorities & align on them
Matt Baker, Senior Vice President, Dell EMC Strategy and Planning, points to immature tools and tool sets as one of the biggest challenges to AI adoption in the enterprise. “Today there aren’t a lot of data science workbenches that are enterprise-ready, which has the effect of constraining both innovation and democratization of AI until they mature enough and are cost-effective enough to scale.” There is also a scarcity of AI talent, at least for now, which means that organizations need to be extremely careful with resources they do have.
Data science and AI technologies are growing up, and as part of that phase need to stop being research projects and start producing results.
“Data science and AI technologies are growing up, and as part of that phase need to stop being research projects and start producing results,” says Scott Clark, Co-Founder and Chief Executive Officer, SigOpt. “A huge part of that is defining what success means and what the project will affect. It’s one thing to try to replicate an academic paper in TensorFlow, but, at the end of the day, your shareholders expect business value.” In order to do this, alignment is key. “Every single business needs to define this differently,” he says. “It’s a really hard problem, but if you can get everyone aligned and racing toward the same goal, it’s amazing.”
They approach AI as an opportunity to augment, not replace employees
Some experts (notably the late Stephen Hawking) have argued that AI will replace large sections of the workforce. Others, such as Susan Lund and James Manyika of the McKinsey Global Institute, believe the net effect of AI will be job creation. The likely outcome is that, as with many other technology shifts throughout history, some jobs will disappear, many will change, and new ones will be created. For now, AI requires an appropriate level of supervision and governance to be successful even in the most sophisticated companies. Making the most of AI requires rethinking the capabilities and skills organizations need in the future and allocating human and technological resources with that in mind.
They take ethics seriously
AI challenges many of the expectations that govern the interactions between people and organizations. For example, voice agents and chatbots have created new ways of interacting, while computer vision, such as facial recognition technologies, alters our understanding of privacy. AI also introduces a level of uncertainty to the decision-making process. In predictive algorithms, we can see the inputs and the outputs, but in many cases what happens in the middle is a “black box.”
Machine learning algorithms also encode bias. Common datasets, such as Word2Vec (used to train language-based applications such as translation, predictive text, recommendations, and a host of others) and ImageNet (used to train image-based applications), have been shown to include intrinsic gender and racial biases that can compound inequality and disenfranchise people. The good news is that, while AI does pose unique challenges, it also offers ways to fix them. A few examples:
- Ethical Data Use. Conversational interactions (chatbots and voice agents) are being used to provide context-aware information and transparency around data use, critical in the age of GDPR;
- Marketing and Product Development. Data scientists are working on ways to identify and remediate bias in data models and algorithms, which is not only good for society but also supports business needs such as market segmentation, ad targeting, and, ultimately, customer trust.
- Corporate Brand and Values. Companies like Microsoft, AirBnB and others are using social engineering techniques to embody corporate values such as inclusivity by making customers aware of the impact of their choices (Microsoft has a “bias check” to help users avoid using discriminatory language in their documents); and
- Governance. Researchers are studying ways to provide better explanation and oversight for algorithmic decision-making, critical for legal and audit purposes.
Granted, we’re in the earliest days of exploring the impact of intelligent technologies on business, but, conversation to conversation, these themes keep popping up. What are the indicators of success you’re seeing? Please share in the comments.