The Chinese firm Ping An Insurance says it has 30 CEO‑backed AI initiatives underway, supported by 110 data scientists. Meanwhile, a senior executive at a major Western insurer, when asked to describe what the company was doing with AI, answered: “Experimenting with chatbots.”
Despite the contrast between their efforts, both companies are ahead of the field when it comes to adopting artificial intelligence. That’s according to a global survey of business executives conducted by Boston Consulting Group and MIT Sloan Management Review, which included these two examples.
Just one in five of the companies surveyed has begun to incorporate AI into its processes, and a mere 5% are using AI extensively. Even among industry titans with more than 100,000 employees, roughly half are still sitting on the sidelines.
Despite the lack of current projects, 84% of executives surveyed said they expect AI to create long‑term competitive advantage. Most expect to see major impacts from AI within five years.
The big disconnect is that just 39% of all companies surveyed have an AI strategy. This mismatch between expectations and planning may remind IT execs of the move to the web in the 1990s, the mobile revolution of the early 2000s, and the race to the cloud earlier this decade. The billions of dollars wasted in these earlier episodes—not to mention all the careers derailed when hurried investments went sideways—only illustrate how important it is to be prepared this time.
Businesses shouldn’t run headfirst into AI projects, experts say. Instead, they should take a more measured approach that identifies a problem or problems that AI can address, and accounts for prerequisites like clean data and a team with the right skills to oversee AI projects.
A good AI strategy doesn’t overlook the impact—both real and perceived—on employees and their outside partners. And it recognizes that AI is changing quickly.
“AI can be very impenetrable for some executives,” says Thomas Davenport, professor at Babson College and senior advisor to Deloitte Analytics. “The greatest risk is often the inability to explain your strategy and how you made the decision.”
Don’t call it AI > Variety of AI tasks
AI isn’t one thing. It’s an umbrella term for an array of technologies designed to perform tasks that normally require human intelligence. These include machine learning, deep learning, natural language processing, and voice recognition.
To get started, business leaders need to educate themselves on the basics. “Investigate what specific technologies under the umbrella of AI make the most sense for your company,” says Davenport.
Let’s say your company’s primary goal is mining data to improve business processes—for example, for data center operations, customer service interactions, employee onboarding, or vendor approvals, to name a few.
Companies can start by exploring big‑data analytics and figuring out what machine learning algorithms can bring to the table. For example, many CRM platforms use machine learning to mine contact data with the goal of creating more accurate audience personas and boosting lead volume.
You should only plot strategy or commit resources once you understand what technology you need and how it works. “It’s more research than development in the beginning,” says Sam Ransbotham, associate professor of information systems at Boston College and a co‑author of the survey. “Then you can start by making limited investments and progressing your way in.”
Find the right problem
Even for early pilots, it’s important not to invest in AI just because it’s new and shiny. “Companies we see that are successful with AI have a specific business problem they want to solve,” says Ransbotham.
He points to financial services giant Fidelity, which invested in AI recently to create a rapid‑response customer call center.
“Most customer service is a gauntlet that customers have to run just to prove you are who you are,” says Ransbotham. “Fidelity uses AI‑enabled voice recognition to simplify that process. They can recognize your voice pattern and get you to help mode in 10 seconds.”
Energy producer BP, meanwhile, uses AI tech to analyze sensor data from drilling systems, making it easier for field technicians to detect potential breakdowns. The time saved translates into substantial economic benefits.
What functional areas of business hold the biggest opportunities for AI? It varies by industry. In the tech sector, IT, customer service, and marketing rank the highest for AI deployment, according to the MIT Sloan survey. In the healthcare industry, R&D, operations/manufacturing, and IT top the list.
The more specific you are in identifying where value can be added, the more successful your AI strategy will be, says Roy Singh, a member of Bain & Company’s advanced analytics and IT practice.
It’s the data, stupid
It’s a common misconception: All you need to do is build the right algorithm, plug in your data and start reaping the benefits of AI.
The reality is that algorithms need to be trained over time to become intelligent. For that training to succeed, companies generally require large volumes of well‑organized data.
“Data is a huge part of any AI initiative,” says Ransbotham. “Companies that find their data is a giant mess will have problems.”
Chief among those problems is fragmented data that’s dispersed across divisions, gathered on diverse (and at times outmoded) software, or formatted inconsistently. The process of reformatting data can be timely and expensive.
“All companies have a problem with fragmented data,” says Philipp Gerbert, a senior partner and AI expert at Boston Consulting Group. “If you want to synthesize all your data, it can take a five‑year program—and still end in disaster.”
One example he cites: Several years ago, Morgan Stanley invested in machine learning to automate drafting of research reports. The effort faced immediate hurdles when managers realized that each department formatted documents differently. This forced the company to ship all the data to an outside firm for reformatting, which delayed the project and added costs.
Hire from within
Companies often hire outside consultants to help them deploy new technologies. It’s not that simple with AI, in part because of the dependence on internal data. The smarter strategy is often to develop the expertise you need in‑house, either through training, hiring, or both.
“Let’s say you find an awesome data scientist who’s going to change everything,” says Ransbotham. “They can give you a ‘gee whiz’ complicated model, but you won’t know what it means. You’re going to need someone in your company who knows what to do with the findings.”
Educate your team
No one knows how many jobs will be lost to AI, automation and robots worldwide in the next decade. The predictions range from relatively modest (1.8 million, according to Gartner) to apocalyptic (one billion, according to futurist Thomas Frey).
A 2017 McKinsey study predicts that very few jobs will be eliminated by AI in the next decade, but that roughly 60 percent of all workers could see 30 percent of more of their tasks automated in order to increase productivity.
While 70% of the MIT Sloan survey respondents said they don’t expect AI to replace their jobs—and most are eager to have AI take over boring or tedious tasks—leaders still need to educate employees about the type of AI they plan to implement and how it might impact jobs, roles, and teams.
Data privacy regulations are getting more complex as AI becomes more ubiquitous. For example, the European Union’s General Data Protection Regulation, which takes effect in May 2018, enforces strict new standards for companies when it comes to consumer rights and use of personal data. It also gives consumers an implied “right to an explanation” when they are subject to automated decisions.
Transparency is the rule of thumb for companies plotting any AI strategy. “Be upfront and explicit with customers about how you collect and use their data,” says Gerbert. “And ask for consent before you gather any unusual data.”
Plan for uncertainty
The final step of a smart AI strategy may be the trickiest of all: Planning for an uncertain future.
Consider the insurance industry, where the adoption of driverless cars is expected to shrink the auto‑insurance sector by more than 70% by 2050, according to research by KPMG. Yet the industry as a whole is expected to stay profitable in the years ahead, aided by the emergence of such innovations as AI tools to predict storms and detect fraudulent claims.
To find a competitive edge with AI, companies need to think more expansively than ever before, to model future scenarios and test their strategies against a range of possibilities. The crucial questions remain: Who will make money from AI, and how?
The bad news, says Ransbotham, is that your competitors are all working hard to find a killer AI strategy. The upside: “There’s no company out there that has 100 years of experience with AI, so you can’t be too far behind.”
Don’t get comfortable, though. The pace of change over the last decade suggests the window is closing fast.