As countries and organizations of all sizes flock to invest in their AI capabilities, there is a clear consensus that we are entering an age where success will hinge on the intelligent use of data. The leading organizations will be those who can identify actionable insights from the large data stores they hold about the way their customers behave.

A recent survey of 2,300 executives conducted by MIT Tech Review and Pure Storage found a C-suite that was enthusiastic about the prospects, but with little really to show but enthusiasm.

I wrote recently about Prediction Machines, the latest book by Ajay Agrawal and colleagues, and some 90% of executives agree that generating data-based predictions will be the very foundation of business success. This is especially true when it comes to creating a personalized customer experience.

Such a future is far from certain however, with roughly 80% of respondents voicing concern that they lacked the volume and quality of data within their organizations, and the resources to analyze that data sufficiently quickly for it to be valuable.

Barriers to progress

Whilst most executives are confident that AI will eventually impact both their company and their industry however, the skills shortage is significantly delaying their attempts to actually extend AI beyond small-scale pilot projects.

“The key here is to gain a greater understanding of how humans and machines will collaborate in an AI-driven world—how humans will support the work of machines, and how machines will support the work of humans,” the authors say.

Similar findings emerged from a similar survey published recently by EY.  The majority of executive respondents felt it would have a positive impact, not only on their organizations but on job creation more generally, with fully 20% going as far as to say it will result in a surge of new jobs.

“Our employees don’t feel their jobs are jeopardized by AI. In fact, they demand intelligent automation that enables them to redirect their time towards more complex work that drives greater employee engagement and adds real value. We estimate that we will save approximately 2.1 million hours of people’s time on repetitive tasks in fiscal year 2018 due to automation. Those are hours that can be repurposed and reinvested into the business,” EY say.

New types of job

In their latest book, Accenture’s Paul Daugherty and James Wilson outline three classes of job that are likely to emerge as humans begin to work more closely with AI.

1. Training

We’re moving on from an age in which humans adapt to how computers work, to one in which computers adapt to us.  In order for this transition to occur however, we will need people to effectively train machines to work alongside us.

This could involve ensuring the data that machines are trained with is suitable, or correcting errors and reinforcing successes in machine behavior.

2. Explaining

AI systems are taking on ever more important roles, and as they do so the ability to explain their working becomes ever more important.  Being able therefore to shed light on the black box workings of an algorithm will be key.

Humans may therefore test, observe and explain the algorithms, or augment their interface to make them more explainable.  It also seems likely that there will be a sizeable role for interpreting machine outputs, especially in areas such as healthcare.

3. Sustaining

There is also likely to be a significant role for people who will ensure that AI-systems operate as they should do.  It’s crucial that AI serves us rather than the other way round, and people in this category will fulfil that role.

Tasks might include setting limits for AI-systems or flagging errors in machine judgement.  It might even include a ‘machine resources’ department in the same that we have human resources departments now, who will be tasked with assessing and evaluating performance levels.

It’s a conclusion that is increasingly common, and certainly seems to be the perception of the executives quizzed by Pure/MIT.

“As AI adoption grows, balancing the human-machine equation will become a key driver of success for these initiatives,” the paper says. “A likely scenario will be that—rather than automating entire jobs—AI will enhance human work by automating manual and rote tasks, handling unmanageable data volumes and varieties, and accelerating discovery of insights.”

The challenge now is in developing the learning environment to ensure that these new skills are developed as quickly and efficiently as possible.