• Only 4 percent of executives surveyed in this year’s AI Predictions report said they plan to roll out AI across their enterprise in 2020 — a considerable drop from the prior year.
  • Operating more efficiently and increasing productivity are the top two benefits companies hope to gain from their AI investments. 
  • Five priorities can help businesses achieve ROI on their AI investments today, lead the disruption tomorrow, and manage risk along the way.

US business leaders have learned that deploying artificial intelligence (AI) is harder than they thought it would be, but they’re not backing down. Instead, they’re applying lessons learned and moving forward.

PwC’s 2020 AI Predictions Report, which surveyed 1,062 executives at US companies currently investigating or implementing artificial intelligence initiatives, found that just 4 percent are planning to roll out AI enterprise-wide in 2020. It’s a steep drop from last year, when 20 percent reported such plans. 

The reason for this reality check? Executives realise that they need to focus on AI fundamentals before expanding these initiatives across the business. 

This pause is likely to be a short one: 90 percent of the surveyed executives believe AI offers more opportunities than risks. In fact, forward-looking executives are already achieving ROI on their AI investments and laying the groundwork for an AI-enabled enterprise. 

How can the rest catch up? Executives leading the AI charge can prepare their organisations by focusing on the following five priority areas developed by PwC’s AI experts:

1. Get on board
with ‘boring’ AI

‘Boring’ may not seem like an apt adjective to describe AI. But ‘quick wins’ and ‘emerging technology’ don’t seem to go together either.

Despite this seeming disconnect, both pairs do make profitable partners. For one thing, they provide improvements in efficiency and productivity. In fact, 44 percent of survey respondents in the 2020 AI Predictions Report cited “operate more efficiently” and 42 percent cited “increase productivity” as the top benefits they have gained — or hope to gain — from their AI investments.

Many firms are attaining these rewards by focusing on “boring” (meaning practical or commonplace) capabilities, such as managing risk and threats, automating routine tasks, and providing decision support for employees.

AI Predictions Exhibit 1

 

What’s the best way to build the capabilities behind these mundane yet impressive improvements? Set an intake strategy to determine the best use cases for AI, establish centralised oversight to make sure AI tools are used effectively, and create specialised metrics to keep track of progress. That will help you deploy AI enterprise-wide, while achieving ROI along the way.

2. Rethink
upskilling

If you think that upskilling for AI requires simply offering tech courses to your non-tech staff, you need to rethink that narrow view. As 50 percent of the executives in the 2020 AI Predictions report realised, the top upskilling priority is to give staff immediate opportunities and incentives so they can apply what they’ve learned in training, enabling their knowledge to evolve into real-world skills that improve their performance.

Companies also need cross-skilling: training specialists in one area (such as data science) to gain enough basic skills in another area (such as finance or marketing) so they can speak each other’s language and collaborate effectively. Such cross-skilling is critical — not just for collaborating on AI-related challenges, but also for deciding which problems AI can solve.

The best workforce programs create a new culture, in which business leaders set the direction and goals and then stand back. This approach gives employees the tools, platform and incentives (compensation and recognition) to learn skills and then use them in new ways to perform their work more effectively. It also creates “multilingual” teams, with data engineers, data ethicists, data scientists and MLOps engineers as members of the application development and business teams.

Exhibit 2

 

3. Lead on risk
and responsible AI

There is some bad news here: Business and tech leaders often seem far too complacent about AI’s risks. For example, 85 percent of those surveyed in the AI Predictions report said their companies are taking sufficient measures to protect against AI’s risks, yet only one-third have fully tackled risks related to data, AI models, outputs and reporting.

Exhibit 3

The good news is that there is a way to proceed with confidence: Roll out responsible AI by integrating processes, tools and controls to address issues of bias, explainability, cybersecurity and ethics, among other key areas. 

In our survey, the leading area that executives are working on is making AI interpretable and explainable. Half are taking steps around explainability for those building and operating AI systems, while 49 percent are focusing on explainability for those affected by these systems. 

Beyond these efforts, companies should establish governance by creating a multidisciplinary team that covers all AI operations in the enterprise; customising controls and monitoring to cover all risk aspects of AI; and instituting processes to ensure that rigorous risk management doesn’t impact performance.

4. Operationalise AI —
integrated and at scale

Leaders are putting AI to work 24/7 — not in silos, but as part of operational systems that cross functions and business units. To operationalise AI at scale, it’s essential to integrate data with the technology. That’s why it’s unfortunate that — the same as last year — standardising, integrating and labelling data is low on executives’ priority list: Only one-third of respondents said this is a top 2020 priority. 

Exhibit 4

 

One way to meet the data labelling challenge is active learning: Data scientists do their work, and — by labelling and revising algorithms’ decisions and recommendations — they then teach machines to do it for them. Another approach is to take advantage of cloud-based services that include data sets, which can enable companies to quickly capitalise on analytics and AI.

More broadly, the three keys to operationalising AI are to integrate it into your overall IT stack; develop machine learning operations (MLOps), which combine expertise in data science with software engineering and IT operations; and make your data trustworthy.

5. Reinvent your
business model

Getting artificial intelligence technology right is complex, but it’s actually the easiest part of an AI initiative. According to our survey respondents, the top AI-related challenges are business- and people-oriented: measuring AI’s ROI, getting a budget approved, training employees, making the business case and recruiting AI specialists.

Exhibit 5

 

That’s why it’s essential to treat AI as part of your broader automation or business strategy. Depending on the business issue at hand, analytics or simpler forms of automation, such as robotic process automation (RPA), might be the best solution. Or there might be bigger strategic efforts in which AI is a great addition, particularly in looking at how to prepare your company’s workforce to be future-ready.

Going forward, the non-tech challenges of AI will only grow. For a start, as AI helps automate, assist, and augment your workforce and decision-making, you’ll need to plan how to share, use or invest any value generated. 

If your business is built primarily on workers and physical assets (buildings, equipment, products, etc.), you’ll need a new business model that also incorporates AI’s cognitive assets (AI models that encapsulate your company’s experience and expertise in a specific domain). That model should include a strategy that, rather than being refreshed every year or two, is dynamic and works at the speed of ‘AI time.’

Get ready to lead in
an AI-powered future

Almost half (46 percent) of the surveyed executives in the AI Predictions report believe that AI will disrupt their markets, the sectors in which they operate, or both. But only 12 percent said they are planning to disrupt their own or other industries. In other words, nearly four times as many respondents fear disruption as plan to be disrupters themselves.

With AI likely to be worth US$16 trillion over the next decade, that gap represents an opportunity: Businesses that act quickly on these five AI priorities won’t just achieve impressive returns today. They also will position themselves to play a dominant role in the AI-powered economy of tomorrow.