AI Strategy

AI strategy is about having a clear view of how to enable the organization to utilize AI capabilities. An AI strategy should encompass the following subsets of strategies.

Data Strategy Data strategy covers data management, accessibility, and utilization of data.
Machine Learning Strategy Machine learning strategy covers the implementation of infrastructure, the serving of models, inferencing, and monitoring of machine learning models, and implementation of machine learning operations and platforms.
Data Science and Analytics Strategy. The Data Science and Analytics strategy covers the implementation of a guideline for data science and integration with business. Data Science should bridge business use cases and the implementation of analytical and statistical methods.

Note: The use of AI strategy instead of machine learning strategy is because of the more technical nature of the term machine-learning versus the more ambiguous term AI

Links

Driving Digital Strategy - This is related to the AI strategy discussions- The Weather Company is a case in point. When consumers moved away from TV to mobile phones but did not stay long enough to generate ad revenues, the company execs pivoted and created a service called WeatherFX, which uses data from its app to help retailers predict how the weather will affect consumer purchasing behavior.
Creating an independent unit is like launching a speedboat to turn around a large ship. Often the speedboat takes off but does little to move the ship.
In 2012, Kasper Rorsted, CEO of Henkel at the time, realized this when he asked a team of senior leaders to catalog all of Henkel's digital initiatives. He was shocked to find over 150 separate digital initiatives throughout the company.
In 1960, Theodore Levitt, a Harvard Business School professor, published a provocative paper in Harvard Business Review in which he argued that companies were too focused on products and not enough on customer needs. To help managers address this problem, he asked, "What business are you really in?"
In 1979, Michael Porter, one of my colleagues at Harvard Business School, published a landmark paper in which he argued that a company could follow one of two potential strategies for competitive advantage: either by being cheaper (that is, as a low-cost producer) or by being different (with differentiated products that command higher prices).
Although Amazon started as a low-cost player without the fixed cost of stores, it is not product-centric knowledge that gives it an advantage of differentiation or low cost. Instead, Amazon has mastered three skills: Deep knowledge of customers obtained from mining customer data. This is embedded in the recommendation system for books and movies as well as in the introduction of new products and services. Back-end logistics for warehousing and shipping that could rival the logistics systems of FedEx and UPS. With its investment in drones and now its own trucking business, Amazon is further strengthening this part of its competency. Knowledge of and ability to manage technology infrastructure. This has allowed it to become not only one of the largest online retailers but also a dominant player in cloud computing.
In many cases the value of a product (e.g., WhatsApp) increases as more consumers use it, without any change in the product's features or functionality. This is the direct network effect. In addition, as a product becomes a platform that connects, say, buyers and sellers (e.g., eBay), it gains from indirect network effect.
The two most valuable assets of a company today are its data and its customer base, yet they don't show up in the balance sheet.
Bezos explained how Amazon shifted from selling electronics to manufacturing them: "There are two ways to extend a business ... Take inventory of what you're good at and extend out from your skills. Or determine what your customers need and work backward, even if it requires learning new skills.
In other words, concerts were the razors to sell music albums--the blades.
Music studios and artists had traditionally used concerts to generate awareness and excitement among fans in order to sell and make money on music albums. In other words, concerts were the razors to sell music albums--the blades.
Companies have used the razor-blade strategy for a long time: sell razors cheap to make money on the blades.
Hecht and his team found many surprising correlations. Some were hard to explain, such as that strawberries and raspberries sold far better when it was humid outside! Others made a lot of sense after some reflection. For instance, at a particular dew point, people in Dallas rush to buy bug spray . . . because insects' eggs hatch at that dew point.
Yet other insights were intriguing. During bad weather, for example, women worry about logistics, such as getting kids from school or groceries from the market. For men, by contrast, the same inclement weather seems an occasion for watching sports, hosting parties, and buying beer.
Levitt, a Harvard Business School professor, once said that people don't buy drills; they buy holes.
While a product-focused company would continue to make its drills better, a customer-focused company would think of new technologies, such as laser, which could be used to achieve the outcome (in this case, creating a hole) that the customer is looking for.
Accenture found that payment models such as pay-per-use and power-by-the hour were preferred by an overwhelming 70 percent of consumers to the full up-front product purchase
Coase argued that firms exist because of transaction costs. Simply put, it would be too difficult and costly for you to get up every morning and find a day's work that was suitable to your skill.
GE's analysis showed that a 1 percent efficiency gain could lead to billions of incremental dollars for its customers.
Deloitte, a consulting firm, defines business ecosystems as "dynamic and co-evolving communities of diverse actors who create and capture new value through both collaboration and competition."
He went on to explain three things that are needed for markets to function properly: Markets or platforms need to provide thickness, which brings large numbers of buyers and sellers together. They need to make it safe for participants to reveal and act on confidential information they may hold. And they need to manage congestion, or competition and complexity, that arises from thickness.
In one of his early studies, Eric von Hippel found that 77 percent of the most important innovations in scientific instruments over four decades were developed by scientists using these instruments, and not by the scientific instrument companies.
Dietmar Harhoff confirmed von Hippel's claim that users are the most important source of knowledge for innovations across all major technologies.18 User innovators are by definition close to their
Dietmar Harhoff confirmed von Hippel's claim that users are the most important source of knowledge for innovations across all major technologies.
Open innovation is best suited for well-defined problems. You won't get very useful insights by organizing a challenge around a broad and vague question such as "What is the future of banks?"
US Congress in response to the 1989 Exxon Valdez oil spill in Alaska: During cleanup, the oil-water mixture they extract from the ocean and put on barges becomes very thick and viscous in subarctic temperatures. When the barges try to unload this oil-water mixture at the shore, the process becomes very difficult and slow. As one
Chains are able to reduce inventory by 15 percent, increase order fill rates by 20 percent or more, increase revenues by 2 percent on average,
In the four years since its roll out, ORION has eliminated 1.6 million hours of truck idling time, and has produced an annual savings of 85 million miles in driving and 8.5 million gallons in fuel consumption.
As Jeff Bezos, the founder and CEO of Amazon, has often said, the goal of a company should be to remove friction for consumers, and technology should be used to do exactly that.
Growth is a key priority for every business, and acquiring new customers is a major driver of growth.
Yet most companies track a host of short-term metrics to assess their marketing campaigns--impressions, number of clicks, click-through rate (CTR), conversion rate (from clicks to purchase), and customer acquisition cost (CAC). Of all these, CAC often becomes the key metric for managers when evaluating the effectiveness of their marketing efforts and when allocating budgets.
According to the familiar 80-20 rule, 20 percent of the customers provide 80 percent of the revenue. However, research shows that if we focus on profitability instead of revenues, the rule would be 200-20, where 20 percent of the customers provide almost 200 percent of the profit!
"The best brands consistently win two moments of truth. The first moment occurs at the store shelf, when a consumer decides whether to buy one brand or another. The second occurs at home, when she uses the brand--and is delighted, or isn't."
In 2011, Google coined the term "zero moment of truth" (ZMOT) to reflect the importance of this period of online searching before consumers show up in a store or make an online purchase.
In a Google study 84 percent of shoppers claimed that ZMOT shaped their decisions of which brand to buy.
For decades marketing textbooks have talked about the four Ps (or 4Ps): product, price, place (distribution), and promotion (or advertising).
Advertising people love advertising, but everyone else hates advertising.
"Half the money I spend on advertising is wasted. The trouble is, I don't know which half." This quote, attributed to John Wanamaker, a department-store magnate in the nineteenth century, highlights a constant challenge for marketing executives.
The CEO of Coke in China told me a few years ago that he created a teen advisory board where he invited a few teenagers every quarter to help him get a deep understanding of their media consumption and buying behavior.
First, it is very time consuming. Deloitte found that it spends over two million hours per year to do performance evaluations for its 65,000 employees.

Thoughts

It is more important for business stakeholders to focus on viable business use cases and make measurable KPIs than finding out what machine learning can be utilised for. Machine Learning is just a means to an end, not the end goal itself. Let the machine learning engineers figure out the way there, that is what they are paid to do.
Machine learning projects fail because there are no actionable effects from the output. The output of an AI project in itself is worthless; it is the business value that matters.