Executives less confident in AI at scale in 2020, going back to basics

Profile picture for user ddpreez By Derek du Preez February 25, 2020
Summary:
The annual AI Priorities report from PwC provides some interesting insights into where companies will be focusing their AI attention in 2020.

Image of a business executive using AI and machine learning

Executives are less confident that they will achieve adopting AI at scale in their companies this year, according to PwC’s annual AI predictions report. Some 20% of respondents last year said that they planned to deploy AI enterprise-wide - that figure is down to just 4% this year. 

As we’ve highlighted time and time again on diginomica, AI in the enterprise comes with a whole host of risks and opportunities. Not to mention that a lot of the vendors that are pushing ‘AI’ as a revolutionary technology option for buyers are simply repurposing automation (albeit, perhaps more sophisticated automation). 

That being said, AI and ML will reshape a lot of business activity, bringing with it interesting choices for how companies decide to use it. For example, will cost savings from the new technologies be reinvested in employees, enabling them to spend more time on ‘value add’ activities? Or will those savings simply be counted as a win for the balance sheet? 

Equally, there are a range of ethical questions to address, where companies grapple with the complexities of bias and ensuring algorithms are deemed ‘ethical’. Those questions don’t always go hand in hand with providing transparency around how AI is used, as the tools are often perceived to be an industry competitive advantage. 

It’s also worth noting that AI varies in sophistication. Some robotic process automation, for instance, will likely be bought out of the box, whilst other systems will be developed in-house using a team of data scientists and proprietary data. 

Which is why PwC’s survey is somewhat interesting for buyers considering the AI landscape - it highlights what the top AI priorities are for companies over the next 12 months. 

This year’s survey, for what it’s worth, includes 1,062 survey respondents, 54% of which hold C-suite titles, with more than half working in IT and Technology functions, and 36% were from companies with revenues of $5 billion and up.

On the perceived lack of enthusiasm for AI at scale this year, the report notes: 

The annual survey reveals the primary reason for this retrenchment: the need to focus on fundamentals before enlarging AI projects.

Looking forward, these insights provide lessons in how business and technology executives are overcoming challenges, what their priorities are for AI progress, and how they expect to reap rewards. UK businesses are still focused on capturing the expected £232 billion in AI gains in the next decade, the vast majority of executives surveyed believe that AI offers more opportunities than risks, and nearly half are expecting AI to disrupt either their geographical markets, the sectors in which they operate, or both.

Some of the more interesting takeaways (which will be outlined below) include that companies are fairly confident on the ethical challenges around AI, but are less confident about their ability to use the new technologies to disrupt their own industries. 

Back to basics

Interestingly, this year, respondents appeared to be looking to AI as a tool to realise further productivity gains, rather than reinvent the wheel. The specifics provide a bit more detail, but broadly companies are either looking to AI to ‘operate more efficiently’ (44%) or ‘increase productivity’ (42%). 

Digging deeper, most said that AI would this year help their companies to manage risk, fraud and cyber threats (38%), automate routines (35%), or help employees make better decisions (31%). You can assume from these responses that those surveyed expect AI will help them do what they do today, but more efficiently. 

When it comes to rethinking the ‘art of the possible’, it appears that companies are less clear on this. The report notes: 

When asked about the AI-related scenarios respondents considered among the top-three threats over the next five years, the answer was clear: a full 46% cited disruption, whether of their geographical markets or of the sectors in which they operate. Yet, when we asked which benefits they were hoping to achieve with AI, only 12% said they were 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.

Considering the size of the AI prize, the disruption of markets and industries is simply a matter of time – and the clock is ticking.

Skills continues to be an issue

When looking at the impact of AI on skills, there’s a lot to consider. For example, for employees no longer doing the routine tasks being carried out by AI tools, what is their role now? Will they be retrained? What about companies needing to hire data science teams and people capable of understanding the complexities behind AI? Also, what about the C-Suite that needs to be educated on the role of AI and the consequences of its use? 

PwC’s report pulled on some of these themes, noting that traditional upskilling - where companies offer learning opportunities focused on a siloed technology - will not be enough to get employees or a business ready for AI at scale. 

Some 50% of respondents noted that companies need to give immediate opportunities and incentives for people to apply what AI skills they’ve learned, so that knowledge turns into real-world skills that improve performance. The report adds: 

Companies also need cross-skilling: giving specialists in one area (such as data science) enough basic skills in another (such as corporate strategy) so they can speak each other’s language. Such cross-skilling is critical not just for collaborating on AI-related challenges, but also for deciding which problems AI can solve. Your teams should be ‘multilingual,’ integrating multiple tech and non-tech skills.

While the trend of democratising AI – making it accessible to your entire workforce – is a positive one, data scientists and AI specialists will still need to maintain expert supervision and control in areas such as AI model development and training, data and model governance, and engineering for production. Thirty-eight percent of survey respondents said they are implementing credentialing programmes for data scientists and more advanced AI skills.

Ethics

As noted above, companies and industry are having to face up to the realities of using AI ethically. Consumers are apprehensive of machines knowing too much about them and interacting with them in ways that don’t feel ‘natural’, whilst employees are also cautious of the role AI will play in their day to day work life. The media is also getting more clued up on the risks of AI - particularly around the ‘black box’ nature of algorithms. 

However, according to PwC’s survey, business leaders appear (worryingly) unphased. Some 85% of those surveyed (executives actively working with AI) said their companies are taking sufficient measures to protect against AI’s risks. PwC notes that this “may suggest an under appreciation for the true level of effort needed to responsibly capitalise on AI”. 

In fact, only one third of respondents have fully tackled risks related to data, AI models, outputs and reporting. On specific AI risks, 70% of respondents said that they we either ‘still exploring’ or ‘exploring/no activity’ as it relates to models and algorithms. 

It is somewhat concerning - although perhaps not all that surprising - that executives are chasing the benefits of AI, whilst lacking in concrete examples of how they’re protecting against the risks. 

My take

There’s a lot of meat in this survey, which isn’t something that I often say about a survey. And given (I’m sure) PwC will be turning around to companies saying ‘we can help you solve all these challenges’, some of it needs to be taken with a pinch of salt. However, a lot of this rings true with what I’m seeing when I talk to vendors and customers. The hype around AI is subsiding (a bit) and the realities of how it can be used in the short to medium term are setting in. However, there is still a zealousness within the industry to pursue it at all costs given the potential cost savings, whilst perhaps not always thinking about the consequences - the need to put effective governance and controls in place. I can guarantee that in years to come, companies will suffer because of those over ambitious plans, if they don’t think this through carefully.