Delivered by DHL

Over the last five years, artificial intelligence (AI) has exploded out of the research laboratory and is thriving in the outside world. AI technologies have become so successful, so rapidly, that their presence is now ubiquitous in many sectors. AI algorithms drive the systems that choose the content displayed on social media feeds, rank product recommendations on e-commerce sites, and allow smart speakers and digital personal assistants to parse and respond to spoken questions. So far, the most significant impact of AI has been felt in the consumer space, but according to a new report from DHL, AI technologies are poised to transform the B2B economy too.

To understand the potential of AI, say the report’s authors, it’s important to recognise that the term refers broadly to human intelligence exhibited by machines. These systems make use of data and learning frameworks to solve the kinds of problems humans solve, interact with humans and the world as a human does, and create ideas like humans. It’s vital to understand the capabilities and limitations of today’s AI techniques. Computing systems of the past were, by and large, deterministic in nature. They were designed to achieve specific objectives and operated using explicit rules created by their developers, where a set of inputs generated a fixed set of results. While they excelled in speed and accuracy, conventional computer systems were unable to learn from experience or find new, creative solutions to previously unseen challenges. AI shifts this paradigm by demonstrating the ability to perceive, translate and understand the content of vast amounts of data generated by individuals, systems and businesses every day.

Narrow but powerful

While AI practitioners debate the feasibility of building a machine with the flexibility, reasoning and imagination necessary to rival the human mind, this “general artificial intelligence” remains a purely theoretical concept today. Real-world applications are commonly referred to as “narrow AI”; these systems can cope with noise and uncertainty, produce novel solutions and improve their own performance over time. Unlike a person, however, narrow AI systems can’t switch easily between tasks; AI programs that beat world champion board game players like Lee Sedol in Go, or Garry Kasparov in chess, could not perform lifestyle services conversationally the way Amazon’s Alexa could.

AI technologies have some characteristics that have hampered their adoption in business and industrial applications, however. AI systems are stochastic in nature: like a person, they get things right most of the time, but not every time. In some contexts, that doesn’t matter too much. A website that serves up irrelevant recommendations is only irritating. A system that makes business- or safety-critical decisions needs to consistently perform better than the approach it replaces. That’s happening today. Virtual assistants now achieve word recognition rates of more than 94%, exceeding average human comprehension in normal conversation. Similarly, while autonomous vehicles have been involved in a handful of high-profile collisions, human drivers cause comparatively more traffic accidents, without the advantage of computational memory that never forgets a driving manoeuvre or scenario.

The performance of AI technologies isn’t the only factor reaching a tipping point. They are also becoming much more accessible. AI practitioners are placing many of their technologies into the public domain via open source tools. Established companies and startups are packaging those technologies into robust products or cloud-based services. Finally, as more industries embrace digital products and services, they can use their own historical data to train various AI technologies for specific use cases.

A new industrial revolution

AI technologies are already being applied to do everything from sorting and grading vegetable produce to medical diagnoses. “AI will eventually be contained within every software product on the market. This will bring us a new generation of tools that fundamentally augment and enhance the quality, reach and velocity of human expertise,” says Ben Gesing, one of the report’s authors. In logistics, for example, DHL and its partners are already exploring the use of AI techniques to provide better responses to customer queries, predict service delays and even determine delivery requirements based on customer profiles.

Ultimately, says Gesing, one of the biggest risks for companies lies in long-term non-competitiveness from not investing in AI technologies soon enough. “Artificial intelligence platforms are fundamentally different from other investments,” he says. “Because these systems can learn and improve their own performance, the business case for their use is very different from conventional assets. Rather than diminishing over time, the benefits tend to grow over time.”

This article was originally published in DHL’s Delivered magazine.