More is smart
The more data, the more AI can learn at scale, the smarter the system becomes. China has declared its intention to become the AI superpower for that reason. With more people, and more data about them than any other government, they are potentially able to train and deploy AI systems learning more, ahead of anyone else.
Learning never ceases, for as long as data is coming in. AI systems self-correct, upgrade and change the rules, remodelling themselves to keep improving. Google’s data centre cooling AI system continuously optimises climate conditions, adapting to internal and external factors like room temperature and weather, self-regulating in a positive feedback loop, that has saved energy by 40%.
Data can come from anywhere. Tesla combines data from online customer forums, its cars (fleet-learning), devices like traffic cameras and road sensors, and soon other car brands; Maritime AI first Maersk from surrounding vessels, and the sea. In digital surgery Verb Surgical links robots to learn from each other. Yodlee, US AI platform finds people who have learnt to learn how to achieve financial health.
Each interaction is an opportunity to ultimately create customer value – the more interactions, the more learning, the more intelligence can be fed back into the system to make known and not yet known connections.
Virtual interfaces can interact with customers en masse, simultaneously – China’s Merchant Bank’s WeChat converses with 1.5 to 2 million customers a day. Bots have none of the old constraints, interacting 24/7, from anywhere to everywhere, taking up no space, working always at peak performance.
Predicting with accuracy
With AI, learnings can be more accurate than any human accomplishment. Music streaming giant Spotify predicts what music a person may like, with amazing accuracy. Google’s AI energy innovation can create better learning rules than the researchers who made it!
Increasingly creative inputs are being used to machine-learn. Smart Finance in China knows the likelihood of a customer repaying a microloan, using app typing speed and phone battery percentage.
Prediction can mean prevention. In healthcare, SemanticMD has achieved 95% detection for tuberculosis in Africa, Case Western Reserve University 100% for breast cancer. Through AI application from its 20 million car stronghold, Bilprovningen, Swedish motor vehicle-inspection company predicts faults and service needs, outperforming any car’s on-board computer.
Making the collective work for individuals
Aggregated intelligence can be hyper-personalised. US Syngenta’s enhances crop productivity by recommending which soybean variety to plant where, based on a farmer’s specific piece of land. Amazon’s AI has given it a head start into fashion, generating clothing recommendations for individuals based on their wardrobe and style.
AI knows what customers are doing at any moment in time enabling enterprises to manage micro-moments in people’s lives, e.g. alerting customer if they are too drunk to drive (Uber), or recommending a nearby restaurant based on food preference (Waze).
Bots are also progressively able to contextualise by learning to recognise not just what customers need, but how they feel. Spotify knows what music customers may enjoy, based on mood, taste and occasion. Spanish Bank BBVA’s virtual assistant can detect the caller’s sentiments and reactions and adapt responses. China’s Hema retail disruptor uses AI to get emotional cues from chats to fix each person’s problems, reportedly ten thousand times more efficiently than customers were getting before.
Zero delay action
Actions and reactions can have zero delays because learning is turned into intelligence at lightning speed. Said one financial services executive, “a decision which would have taken us 24 hours or more to do, we now get the answer in seconds”.
Decisions can be made by anything or anyone, because intelligence is decentralised to whatever or whoever whenever it is needed, signalling an end to old structures and protocol.
Impact is felt across industries and professions. Huge document loads to prepare lawyers for trial and help with litigation decisions now can be done in real time. Insurance claims can be made superfast: SA start-up Naked can get a claim approved within minutes, including fulfilment.
Eliminating old trade-offs
Scaled learning produces increasing value at low cost, a winning formula for getting and keeping a competitive advantage. Take Uber – the more riders and drivers it adds, the better able and more cost-effective it becomes.
No trade-off has to be made between value and cost.
An Irish bank executive shared this: “With our AI chatbot deployment, we knew about each person’s habits, intention and situation, so we were able to target and service precisely, making sure each customer got the right response. If customer behaviour changed, the system changed. This pushed the cost of transactions down 65%, and the costs will keep going down over time.
Additionally, because we could assess when customers left a journey and why and fix it we reduced bounce rates by 15-50%, which meant conversion rates increased. We also got jumps in customer satisfaction – recommendations went up 10% and we halved the complaint rate”.
When the cost of expensive physical assets such as call centres or vast branch networks are trimmed, people can be reskilled and redeployed to further add value. Savings can be passed onto customers, attracting more customers and keeping existing ones longer, spending more, further pushing costs down.
Unlike the finite resources of the past, learning has elasticity – it grows as its being reused. It can be leveraged over and over, therefore, getting and giving value without incurring the set-up or marginal costs of the past.
The more time Amazon’s Alexa spends with customers the more data it gets, the smarter it gets, the more pervasive its virtual assistant can become, at minimal added cost. Amazon exponentially not only keeps increasing Alexa’s now over ten thousand skills including home automation, shopping, fashion recommendations, etc, it recently extended its presence from homes to cars.
The point is, cover the AI set-up costs, and expansion to get more to and from customers becomes a virtuous cycle of higher value and decreasing costs. Singapore’s supercharged Smart Nation AI investment is premised on this principle. New services are increasingly being added from on-demand public transportation – buses, shuttles, aerial taxi, rail, all autonomous – to security, assistive health for aged and smart homes, paid for in one cashless, ever-smarter system to improve lives.
Learning investments can also be leveraged by licencing it to other industries, as Google has done with cooler systems. Or to go global. Bringing computer vision to agriculture, Aerobotics, an SA start-up, now in eleven countries, identifies crops or plants with health problems to enhance yield. Said an executive, “Once we have data from one crop in one country we can scale the learning for multiple crops in multiple countries cost-effectively”.
Competitive AI framework
AI is most potent when an enterprise knows what it needs to learn in order to become the expert at doing something better than anyone else. And when it uses the positive force of learning to get and stay ahead.
From our research, case studies and consulting, we believe the approach an enterprise chooses is a function of its business logic – either product or customer-centric – (y axis), and the objective for learning – either to optimise what exists or originate. (x axis). (see Diagram 1).
Four competitive approaches have been identified.
Product logic drives making and moving more stuff. Operations Maximisers do this by deploying AI to optimise internal systems, in order to make and move core products better. Product Innovators compete by making and moving better products. They originate by adding smart new features to their products, based on real-time learning on how the products perform in use.
Customer-focused logic uses AI to deliver better customer outcomes. Category Champions optimise the customer’s journey in a product arena. New Market Makers originate by changing the business model to create better social and business practice. Mainly platform-based, mostly newcomers, they use AI power to connect products, people and places crossing traditional industries, to build new competitive spaces.
The cases below demonstrate.
CASE STUDIES
STRATEGY 1
Operations Maximisers: Makes and moves products better.
Case: Zara (B2C)
Competitive advantage: Wins through systems efficiency.
Strategic question: How do we get fashion products from factory to customer quicker than anyone else?
AI learns from: Internal data and systems
Zara has concentrated AI to optimise its internal systems to build superior supply chain management. It can get fashion from the factory to the store quicker than anyone else. Its consequent growth has given it the ongoing inventory intelligence that predicts customer buying trends so as to get the right products into stores at record speed, minimising distribution lag, keeping production optimised with near zero stock waste, costs down and increased affordable fashion availability to customers.
Case: Maersk (B2B)
Competitive advantage: Manages risk through systems efficiency.
Strategic question: How do we operate our vessels at sea better?
AI learns from: Ship data and environmental surroundings
Maersk uses AI learning to improve the situational awareness of its container ships at sea to maximise efficiency. Its intelligent risk management capability makes them the expert at identifying and tracking objects and potential conflicts to avoid collision, decreasing potential damage and delays, thereby increasing customer reliability and ultimately lowering cost.
STRATEGY 2
Product Innovators: Makes and moves better products.
Case: Tesla
Competitive advantage: Adapts products to context.
Strategic question: How do we design and produce better cars and make them smart in use?
AI learns from: Products in situ
Tesla pioneers in electric cars. It has harnessed the AI embedded within its cars to enable them to do smart things, adjusting to conditions and occurrences in real time. When the car has a problem, like with its suspension, it instantly takes action for that car and feeds the learning back into the system. When Hurricane Irma devastated the Florida coast, Tesla remotely upgraded the battery capacity of its vehicles so that customers could escape harm by being able to drive the necessary distances without having to recharge.
STRATEGY 3
Category Champions: Delivers a better customer journey.
Case study: Hema
Competitive advantage: Offers a better retail journey.
Strategic question: How do we deliver integrated in-store and online food shopping?
AI learns from: Customers' behaviour and preferences
Hema has optimised its system to form smart stores, merging on and offline environments to enhance the customer journey. From billions of searches, customer profiles and behaviour it has built an unrivalled AI capability. As customers choose products or ask questions about say, origin, ingredients, cooking instructions, nutritional information and price, the system learns more about them, and gets increasingly good at translating this into the millions of personalised pages generated in real time on the Hema phone app. As a result, the system is said to have better engagement and 100% better click-through rates than that created by people. Adjusting lists, reminders and personalised route-guiding are next on the ongoing list to elevate the offering based on learnings.
AI heavy, stores act as warehouse and fulfilment centres with staff as packers, enabling a delivery of 30 minutes within a three-kilometre radius.
STRATEGY 4
Market Makers: Creates new competitive spaces and better practices.
Case study: Uber
Competitive advantage: Connectivity across products and industries.
Strategic question: How do we get people and products from A to B in new, better ways?
AI learns from: Crowd to make connections
Uber created a whole new market for never-before-heard-of ridesharing. Using AI at its centre it opened the “urban mobility” space, allowing customers to pay for a service or outcome, instead of purchasing the product.
Then it expanded mobility cross-sector, including helicopters, boats and more recently electric bicycles and air taxis, expanding its footprint at low cost. That emerged into a “product mobility” space – moving things from food to flu vaccines.
Uber uses AI for almost all of its functions. But at its core, Uber knows how to use its crowd to connect people, products and places. It knows which people and cars are travelling to where and when, weather and other conditions. It has learnt how to deliver a real-time personalised experience, matching demand and supply, passenger and driver.
It continues to open up competitive space for itself and its ecosystem, using learnings to originate, making new connections across product and industry. For instance, with Toyota and others, a new space “mobile shop and mobile production” (think workers shopping from a mobile vehicle or pizzas being made on the move), has materialised.
How can I use this framework at my company?
The framework presented can be used as a strategic tool by both legacy or newcomer enterprises that have already embraced AI, to assess where they have aimed their strategy and if that is where they ideally should be positioned. Alternatively, it can be used for planning an AI-led future strategy.
In both cases these questions serve as useful guides:
· What is the strategic question we need to solve?
· What do we need to know better than anyone else that will answer that question so we can get and stay ahead?
· Which of the four AI strategy options is most appropriate for us?
