by Clint Boulton

7 tips for scaling your AI strategy

Dec 31, 2018
AnalyticsArtificial IntelligenceIT Leadership

Now that your enterprise has experimented in AI it’s time to consider how to expand the efforts. Here’s how, according AI visionary Andrew Ng, as well as experts from PwC and Deloitte.

ai artificial intelligence circuit board circuitry mother board nodes computer chips
Credit: Getty Images

Pilot projects of artificial intelligence (AI) technologies proliferated in 2018, as many enterprises tested machine learning (ML) algorithms and an array of automation tools to cement relationships with customers, improve network operations or augment their cybersecurity postures.

Encouraged by early results, CIOs are preparing for the next challenge: scaling AI throughout the enterprise. Twenty percent of 1,000 U.S. business executives said their companies plan to implement AI across their enterprise in 2019, according to new research from PricewaterhouseCoopers (PwC).

Business aspirations are soaring. Companies are investing more in these emerging technologies, as IDC projects spending on cognitive and AI systems will reach $77.6 billion in 2022 — more than three times the $24 billion forecast for 2018.

But no matter how big the aspirations are, the road to scaling AI is fraught with perils such as warring strategies and shifting business priorities that can stifle cross-departmental collaboration. The dearth of talent to handle the technical work compounds the issues.

Here, AI experts from PwC, Deloitte and startups offer key advice CIOs should consider when scaling their AI initiatives.

Build an in-house AI team

Securing buy-in from the C-suite, enterprises should build an AI team, which will help keep projects within the company, says Andrew Ng, founder and CEO of startup Landing AI. The AI team could sit under the CTO, CIO, or CDO (digital or data), or even a chief AI officer. The formation of such a team will help with recruiting and retention.

“With a new AI unit, you’ll be able to matrix in AI talent to the different divisions to drive cross-functional projects,” Ng says in an AI transformation playbook published in December. “New job descriptions and new team organizations will emerge.” In previous roles leading AI teams at Google and Baidu, Ng has installed machine learning engineers, data engineers, data scientists and AI product managers. However, Ng acknowledged, the current war for AI talent is “zero sum in the short term.” Companies must work with recruiters to fill key positions.

Teach ‘citizen AI’ staff and AI specialists to work together

The paucity of AI talent shouldn’t kill AI initiatives. Rather, companies should leverage tools that democratize AI and data science, including applications with user-friendly interfaces for AI developers, as well as educational programs designed for non-tech specialists.

Companies can group employees into three tiers: citizen users, who will learn how to use AI-enhanced apps; citizen developers, or power users who can identify use cases and data sets and work closely with AI specialists to build new AI apps; and data scientists, who will do the heavy lifting to create, deploy, and manage AI applications, says Scott Likens, new services and emerging technology leader at PwC, and co-author of the report on scaling AI across the enterprise. This will require upskilling efforts in order to close the talent gap.

Establish a center of excellence

One of the best ways to build an AI foundation is to set up an AI center of excellence (CoE), Likens says in the PwC report. This organization, which will determine technology standards, architecture, tools, techniques, vendors and intellectual property management, will figure out how to identify use cases and how to develop accountability and governance.

Energy giant Shell, for example, has set up a data science CoE that leverages AI, ML and analytics to tackle projects such as predictive maintenance for oil-rig machine parts. Twenty-four percent of respondents currently have some form of AI CoE, PwC says.

Increment your AI strategy through experimentation

While it may tempting to craft an AI strategy right away, Ng says that most companies can’t develop a thoughtful AI strategy until they’ve had some experience with the technology.

Ng recommend building several difficult AI assets that are broadly aligned with a coherent strategy, yet tailored to create an advantage in an industry sector, which makes it difficult for a competitor to replicate. This requires a sophisticated data analytics strategy to cultivate business insights.

Real estate firm Keller Williams for example, leaned on thousands of carefully curated datapoints about houses and ML software to improve its listings, says Neil Dholakia, the company’s chief product officer. Real estate agents record footage of a home on their smartphones using a Keller Williams app, which connects to Google’s Cloud AutoML software. The software immediately identifies and tags features such as hardware floors or granite countertops.

“This went from days and cost to minutes and free for our agents,” Dholakia tells Dholakia, who prizes ML for its potential to offer competitive advantage in the sector, says he plans to expand Keller Williams’ use of AI in 2019.

“An AI strategy will guide your company toward creating value while also building defensible moats,” Ng says. “Once teams start to see the success of the initial AI projects and form a deeper understanding of AI, you will be able to identify the places where AI can create the most value and focus resources on those areas.”

Build responsible AI

One of the chief hurdles of AI adoption is explaining how an AI model makes its decisions, a salient concern in regulated markets such as finance. That’s why it’s important to create AI models that are transparent, said Cathy Bessant, chief operations and technology officer for Bank of America, at the recent AI Summit in New York City.

Organizations can address these “black box” concerns by answering: Can an organization ensure those decisions are accurate? Who is accountable for AI systems? Are there proper compliance controls in place?

A successful AI deployment will build in accountability for each of these factors to create “responsible AI.”

Practice participatory or human-centered design

How does one go about building responsible AI? The first step stakeholders should take is taking a hands-on role in designing complex AI implementations, according to a recent Deloitte report on the state of AI in the enterprise.

Participatory design — a form of human-centered design — embeds the needs of a user “community” directly into the design process to develop more sustainable solutions. This enables designers to become aware of and stave off issues they may not have anticipated by a failure of context or imagination.

For example, if a call center implements a chatbot to reduce employee workloads, a participatory process would include the call center employee, a member from the leadership team, and customers that may interact with the chatbot.

To ensure AI is grounded in ethics, enterprises should build on participatory design by “periodically reviewing and assessing the algorithms to ensure that they are doing the right thing,” says Vic Katyal, a principal and the global data risk and analytics leader at Deloitte. Finally, businesses should allow a third party to validate the AI independently, which will help fill in gaps and work through blind spots.

The good news? Katyal says CIOs are far and away the most common senior exec tasked with governing AI adoption in the enterprise by corporate boards.

Craft a communications strategy

Because AI will have a significant impact on the business, enterprises should create a communications program to ensure alignment. This will cover investor relations (explaining a value creation thesis for AI); government relations (if necessary); customers (think strategic marketing); talent (branding is crucial to lure fresh blood); and internal communications.

“Because AI today is still poorly understood and artificial general intelligence specifically has been over-hyped, there is fear, uncertainty and doubt,” Ng says. “Many employees are also concerned about their jobs being automated by AI. Clear internal communications both to explain AI and to address such employees’ concerns will reduce any internal reluctance to adopt AI.”

The bottom line

Most executives are bullish on the promise of AI. Fifty-six percent of 1,100 IT and business executives Deloitte surveyed said that AI will transform their business within three years.

“The AI/analytics arms race will continue as businesses need to get lean, agile and focused on growth,” says Andy Walter, a consultant for CIOs and a strategic advisor to Fractal Analytics. “Leaders that have leveraged AI capabilities across targeted business processes will expand to enterprise-wide value driving opportunities. The ‘intelligent enterprise’ will beat competition on the top and bottom line.”