Artificial Intelligence: Why you should think before you act
Tags
Artificial Intelligence (AI) is advancing and the media is flooded with stories of AI exceeding human capabilities. Google’s Deepmind beating the Go world champion, Oxford University’s LipNet lip-reading with 93.4% accuracy (where humans can only manage 20% - 60%) and AI detecting cancer on scans quickly and more accurately than doctors.
We are also seeing AI – and robotics – at play in the world around us. Netflix uses AI to recommend content to its subscribers, Amazon uses Kiva robots in its warehouses and an increasing number of organisations are using chatbots to interact with their customers. And this is just the beginning.
Investment in Artificial Intelligence is growing faster than ever. According to CBInsights, global investment in private companies focused on AI has increased from £445 million in 2012 to over £3.8 billion in 2016.
Despite the investment and interest in AI, it’s falling short of expectations. While much of this is undoubtedly down to inflated expectations, AI is simply failing to meet its potential. In fact, in July 2017, Erik Brynjolfsson and Andrew McAfee, Director and Co-director of the MIT Initiative on the Digital Economy, were quoted as saying: “Although AI is already in use in thousands of companies around the world, most big opportunities have not yet been tapped.”
It’s not the first time that the promise of a new technology has failed to meet industry expectations. In the 1970s and 1980s we saw it with the productivity paradox, where productivity growth in the US declined despite significant investment in IT. At the time, the economist Robert Solow commented: "You can see the computer age everywhere but in the productivity statistics.”
We saw the same story play out in the late 19th Century. Almost 20 years after the establishment of electrical power stations in London and New York, a mere five per cent of mechanical power came from electricity (the rest remaining steam-powered). Economist Paul David claimed in his 1989 paper The Dynamo and The Computer it was a further 20 years before there was a measurable impact on the economy.
The same is happening with AI today. Unless we adopt a very different approach, we shouldn’t be surprised to see history repeating itself. For industry as a whole, the alternative is a significant loss in productivity. For individual enterprises, the alternative is a loss of competitive advantage or – worse still – irrelevance.
So what should business leaders of today do to make things different this time? What is needed to unlock the productivity gains that AI allows without waiting 20 years?
Leaders need to stop, take a deep breath and look at the big picture before stepping into the fray
Let’s explore this through a very similar case – the introduction of electricity to steam-powered factories. This is something economists such as Tim Hartford[1] have explored for some time.
Before the advent of electricity, these factories had large central steam generators that were linked to drive shafts extending down the length of the factory. These drive shafts in turn were linked to belts, gears and other mechanical instruments that drove the machines, like looms, mills and saws. The whole system depended on further cumbersome support systems including fire-prevention, lubrication and safety measures. This was a system crafted around the requirements and limitations of steam power.
Along came electricity and the excitement and promise that went with it. Many factory owners didn’t know how to respond. Some simply replaced their steam generators with electrical ones without changing anything else. What was clear was that the steam generator itself was only one part of a much more complex configuration. Replacing just one part of the system was not going to deliver benefits. For the potential of electricity to be exploited properly, it called for the whole factory to be re-configured under the new rules that electricity allowed.
This has seen the transformation of the factory since the 1880s, and the ability to transmit electricity via wires has allowed power to be delivered where it is needed rather than where it was generated. Individual workstations can now have their own power, making the production line possible. In short, electricity changed the landscape and the assumptions upon which business was built. It was not just a tool to be deployed, but a game-changer.
We are already seeing some of the prolonged teething problems with AI and automation. Worse still, the hype and market buzz is leading to hasty and ill-conceived applications of this emerging technology – applications that often amount to little more than replacing a human with an AI. This is equivalent to replacing the horse in a horse-and-cart with a robotic horse, rather than creating a car. This erratic, bolt-on mentality leads to enterprises that grow in an uncontrolled way to the point of becoming wildly complex and often dysfunctional. These organisations are destined to be little more than extensions of their outdated ancestors but with expensive, high-tech bells and whistles attached.
Just as the factory in our example had systems that were crafted around the requirements and limitations of steam, the systems in the modern enterprise are crafted around the requirements and limitations of humans, such as:
- Line management and audits to compensate for human error, bias and misconduct.
- Large HR functions to handle disputes, stress, sickness, holidays, recruitment, training and more.
- Ergonomic physical infrastructure and equipment.
The fact is, as soon as pen touches paper with strategy, assumptions are made based on the rules that humans allow. Armed with AI, we must revisit our enterprises and challenge them – starting with the customer and the outcome that customer seeks. And we must redesign our business and operating models wherever practically possible to best exploit the resources we have.
In some cases this may mean doing away with the human-centred infrastructure altogether, in others it will mean creating infrastructure that facilitates human-AI collaboration. In no cases should it be blindly substituting AI for humans.
There is no silver bullet, but by applying the principles below, we believe organisations will see better outcomes and stand a greater chance of survival in the short and long term.
1. Be focused
We know that good strategy comes down to choice and focus. Organisations that try to be all things to all people or those with broad-brush strategies ultimately fail. Resist the temptation of getting a quick buck from chasing the latest AI and Automation fad. Instead focus your valuable attention and resources in reinforcing specific areas of competitive advantage.
There are many lenses that can be applied here. One way is to determine your focus for AI and Automation based on your value discipline – product leadership, customer intimacy or operational excellence. Another way would be to take a core competency of your organisation, like product development, and use AI to make it less susceptible to competitive threat. Considering desired customer outcomes is another critical way to prioritise AI initiatives. Will AI really create value that’s recognisable to customers?
2. Drop the ‘AI supplanting humans’ ethos
There is a prevalent and dangerous mentality today – business leaders are transfixed on using AI to supplant human roles. This is not only short-sighted and likely to create a dysfunctional organisation, but it creates an ethos that devalues the human workforce. In the best case, the potential of the human workforce is not achieved, in the worst case it creates mass demotivation and attrition.
Focus on the outcome required and not on the role. Consider that existing roles are intrinsically human-shaped and forcing AI into these gaps is not the answer. Give your employees the opportunity and incentive to be imaginative about how AI can enhance their output. Identify higher order work in areas such as innovation, creation and strategy for your human workforce to begin to focus their efforts on as AI takes the more repetitive work off their hands.
3. Don’t let automation calcify your business
The Robotic Process Automation (RPA) market is expanding rapidly and organisations are making savings – largely in their back office. The danger here is that cost savings are achieved at the expense of flexibility. As bots do tasks previously done by humans, the processes can become more rigid and less transparent. There is a risk that your organisation can lose agility and be less able to react to changes in the external environment.
To mitigate this:
- Focus on processes that are least likely to change.
- Improve processes rather than simply automating them.
- Use ‘intelligent’ RPA to exploit the possibilities of AI (e.g. image recognition and natural language processing) and not simply rules-based automation.
4. Experiment, but always with the end in mind
The future is uncertain and any rigid plan is doomed to fail. The answer has to be the right balance of bold vision and curious exploration. You cannot embark on any journey by only looking at the horizon or the ground at your feet.
Develop an AI and Automation roadmap that extends five or even 10 years into the future. Then take the first step. Start experimenting with small bets in line with your roadmap to test your thinking. Experimentation needs to become business as usual, and low-risk test environments should be established.
The roadmap should be a dynamic, living and breathing document that is continually reviewed and adapted as you learn from experimentation and as the external environment evolves and shifts.
Establish AI forums to ensure your portfolio of initiatives continues to align with your vision. Establish a design authority to preserve the integrity of your organisation and protect it from piecemeal, bolt-on tinkering.
The journey to adopting AI will be uncertain but also exciting and transformative. With a little foresight, courage and tenacity you can achieve a lot.
This article was first published on Industry Europe.