Technological change has long been associated with hyperbole and inflated expectations. It has created great wealth and sustained change. Artificial Intelligence, Machine Learning, Big Data and Analytics are the latest in a long line of innovations. How do you maximise your chances of success?
The fear of missing out (FOMO) is a powerful motivator. Did you buy shares in Google or Amazon when they had their initial public offering? Would your company survive their moving into your market? The value that these and others have built on the back of AI is colossal. Google held its IPO on 19 August 2004. The company went public at $85, sold 22.5 million shares and raised $1.9 billion.1 At the time of writing, the share price of Alphabet was $1590, a market capitalisation of $1.082 Trillion. Yet the majority of organisations adopting AI do not see a material financial return.
Definitions
A definition of Big Data and Analytics is:2 “The application of statistical, processing, and analytics techniques to big data for advancing business.” Artificial Intelligence (AI)3 “is often used to describe machines (or computers) that mimic cognitive functions that humans associate with the human mind, such as learning and problem-solving“. Traditionally, computers are programmed to take pre-determined steps and logical branches by their creators. Whereas in AI the creator determines an optimisation objective and gives the machine the data from which to determine the approach. This can lead to some unpredictable outcomes.
I here deal with Big Data, Analytics and AI together. The reason is that AI, the actions and service that it informs, feeds upon data. Frequently the data sets are massive, sometimes in the hundreds of millions of transactions. Such volumes permit the observations of small clusters of similarity amongst thousands of parameters. The input includes structured data (typically held in databases) and unstructured (text streams, video, image).
At a recent conference, an attendee described AI as “Big data plus fancy maths.”
Objectives of Investment
A recent survey by MIT reported that few organisations (11%) were gaining significant financial returns from their investment in AI.4 It stated “Overall, a growing number of executives recognize that competitors are using AI, and they believe that AI will generate significant strategic benefits for their organization. Competitive dynamics, ecosystem forces, and financial incentives all motivate the increasing dependence on AI in business strategy.”
The pressure to reduce costs is a common driver to apply new techniques. The MIT survey listed this at the bottom of the reasons cited. I suspect that this is only part of the story, as higher citations include obtaining competitive advantage and defence, there being the potential for overlap. Within IT, tools for monitoring network and data-centre estates are now widespread and chatbots are common in contact centres. Nobody aspires to a wholesale takeover by AI. But if it can automate and bring insight to the mundane, staff are released to concentrate on activity that delivers greater value.
Top-line Growth
Many see Service improvement as a means of gaining competitive advantage, so driving top-line growth. Chatbots supported by knowledge-banks will work 24*7*365 without pay, holidays or caffeine. Millennials have learned first to query YouTube or such tools, only if those fail, calling the service desk. This is to the advantage of both customer and supplier.
As organisations’ maturity in the use of AI rises, they begin to innovate and draw on insights from their data in ways that those lacking the capability could not imagine. Tesco ascribes much of its competitive success to the use of loyalty card data used to drive promotions.5 At the end of a modest 3 month trial of the pilot scheme, the results were reported to the Board. “Tesco’s chairman, Lord Ian MacLaurin, said, ‘Well, this really worries me because you seem to know more about my business in three months than I’ve learned in 30 years’”
One of the most widely cited examples of top-line growth is the deployment by Amazon and others of recommendation engines. These analyse purchasing patterns to suggest other items that the customer may be interested in, based on what they have looked at in the past. News organisations use similar tools to suggest articles of potential interest, so sustaining user page-views through which they can serve adverts.
The Promise of AI and Big Data
Whole new industries have been created that were infeasible before the days of AI. Google was founded on 4 September 1998. It quickly established a strong reputation for web-search. Before it, the likes of Yahoo (January 1994) and Alta Vista (December 1995) had offered directories. They appeared to have owed much to Yellow Pages (1966) in their reliance on industry classifications. They were revolutionary in their day in organising the web. AI allowed Google to break free of prescriptive structures, delivering great value to users in the process. Existing industries have also been significantly changed (Tesco’s use of loyalty data; Tesla’s development of automated guided vehicles and associated data sets).
Academics have researched how AI / Big Data and Analytics have added value to organisations.6 These include:
- Better and more personalised recommendations for future purchases and offers;
- Determination of the root causes of failure, prediction and pro-active fix to sustain service;
- Understanding consumer experience;
- Anomaly detection and fast response, including service interruption, fraud; and
- Fine-tune internal process performance, e.g. logistics.
They note the non-linear effects, the winners (Facebook) gaining hugely greater user volumes and revenues than the second or third in the market. This makes for big, strategic bets. Most wagers will not pay. Some will be upset by the outcome and I foresee that disputes will arise.
Extraordinary returns are seen when a strategic position is built that is valuable, rare, costly to imitate and organisationally embedded. Others cannot quickly replicate such a position to erode margins. The value of data and AI is in proportion to its contribution to the creation of such a state.
The Challenges
The first key to success is obtaining your proprietary and useful data set. The rate at which you can do this is likely to determine your overall progress. In many instances, this will need to be built progressively and requires the solution of multiple small issues along the way. The good news is that this is not all-or-nothing.
Investments in data quality can be expensive and slow to deliver results. Some will be valuable, others will not. It is wise to progress in concert with the development of your AI algorithm and measurement of customer response. It has been said of AI that the winners are not necessarily those with the best algorithm that win, but those with the most good data. There are good mathematical reasons for this. Techniques such as “learning curves” and “error analysis” can be used to ensure that investments are focused where they are likely to deliver the greatest return.
The exploitation of AI is not easy. One cannot generally go out and “buy some AI” and quickly find that it makes a strategic difference. The key, as seen in the earlier Tesco example, is experimentation. Christensen7 noted that transformation was difficult because it relies upon the creation of a different ecosystem: many components need to come together.
Examples of Success
Rolls-Royce has built such an ecosystem.8 It started with a strategic vision of delivering higher engine utilisation to airlines and passengers. It had the engineering knowledge to extend existing instrumentation that informed pilots in-flight, combining local processing with increasingly massive central monitoring and analysis. This central element provided a strategic advantage in tying the airline to the supplier for on-going maintenance and support. Such an interconnected system grew over time. Capabilities were extended. It also adapted its business model in favour of annuity revenue, reducing reliance on initial engine sales.
The development of AI capabilities has significant associated risks. One of the reasons for this is the highly interconnected nature of successful ecosystems. Simple transactions may be predictable and subject to little risk. Where there is such interaction, they are by definition not simple. When it comes to cost and time to realise success, surprises and omissions will be discovered along the way. You cannot eliminate this risk, but you can manage it through proceeding in a series of small steps, each of which is relatively simple and defined. When you find that one has failed, step back, take a different route and try again. You will fail, but this will have a manageable impact.
Getting There
The late Clayton M. Christensen wrote with delightful clarity of thought and expression. He analysed industries in transformation. He advocated:
Managers who don’t bet the farm on their first idea, who leave room to try, fail, learn quickly, and try again, can succeed at developing the understanding of customers, markets, and technology needed to commercialise disruptive innovations.
Clayton Christensen, The Innovator’s Dilemma
Some conjunctions are critical to success. One is building an infrastructure that delivers a data-stream. For Tesco, this included getting Club Cards to customers and persuading them to scan at the check-out. For Rolls-Royce, it included the construction of a network to get engine instrumentation data to their control centre. If this infrastructure gives your organisation a valuable edge that is strategically defensible, that is a good start. Sainsbury could not obtain Tesco checkout data, nor Pratt & Whitney see Rolls-Royce engine performance.
Some elements are easily substituted. There are many AI toolkits available on the market to perform largely similar tasks. A critical step in planning and Intellectual Property definition is the determination of the core against flexible plug-and-play.
Experimenting
A recurrent theme is that of experimentation. A well-designed experiment produces learning by doing. This applies not only to the modification of the parameters in the light of new data, but the adaptation of the data sought, the objectives of optimisation and all aspects of implementation. Even if the product of an iteration was not immediately a market-killer, it should advance you towards a mature solution. If it tells you that there is a gap in your capability, resolving it becomes either a step to include in the next iteration, or conceivably a reason to kill the project.9 It is in this that effective governance and clarity of the business case help to bring unity of purpose to the development team. The avid search for feedback from those who may usefully challenge and develop the approach is natural to some. Others will never manage it and should probably not try.
Obtaining the Data Set
Your greatest challenge would appear to be the construction of the data set and embedding sources within your organisation for updating it. Tesla is valued at a premium in part because of its unsurpassed data from millions of passenger-miles, used to feed automated guided vehicle development. For many, data is fragmented, inconsistent and does not address their needs.
Some will construct data lakes to bring what they have together. Others will start by automating and standardising often boring activity with technologies such as Robotic Process Automation. This brings some advantage in speed and processing cost, but the real prize is often in the reliability of data for later analysis. Such steps take time. They enable you to start with pedestrian improvement of existing processes and move on to more strategic transformation, gaining modest benefit along the way to sustain the journey. A data-driven mindset is increasingly seen as an important enabler.
The delivery of effective AI requires a set of capabilities. Some are commonly hired in, others may be purchased from suppliers, some will be within your workforce.10 The capabilities you need to combine are:
- Governance, leadership and strategic direction. Prioritisation of the portfolio.
- Strategic articulation and formulation of the AI issue, business case and route to realisation. This includes the articulation of customer value and marketing.
- Knowledge of AI algorithms and their respective applicability. Formulation of the AI issue and approach. Refinement, testing.
- Data capture, management, cleansing.
- Operational interactions and embedding in process. Doing the ongoing work of refinement.
- Data architecture, efficient management, processing.
- IT architecture, integration, construction and operation.
The capability will include tools, infrastructure and talent. Supporting activities are likely to include security, data privacy and compliance, legal, commercial and finance.
Conclusion
AI and big data suffer from hyperbole just as much as preceding technological development. Some suppliers will do no more than re-brand existing offerings. This does not detract from the significant value that has arisen from AI and Big Data to date. This is nothing in comparison with what is to come. Building coherent capability is a difficult task. Success is won by the few who assemble the pre-requisite conditions and manage them well. Will you be among them?
…………………..
References
- https://www.investopedia.com/articles/active-trading/081315/if-you-would-have-invested-right-after-googles-ipo.asp . Google split its shares on 27 July 2015. ↩︎
- Varun Grover, Roger H.L. Chiang, Ting-Peng Liang & Dongsong Zhang (2018) Creating Strategic Business Value from Big Data Analytics: A Research Framework, Journal of Management Information Systems, 35:2, 388-423 ↩︎
- https://en.wikipedia.org/wiki/Artificial_intelligence ↩︎
- MIT Sloan Management Review Research Report, Expanding AI’s Impact with Organizational Learning, Ransbotham, Khodabandeh, Kiron, Chandelon, Chu, LaFountain October 2020. ↩︎
- https://www.marketingweek.com/tesco-clubcard-loyalty/ ↩︎
- Varun Grover, Roger H.L. Chiang, Ting-Peng Liang & Dongsong Zhang (2018) Creating Strategic Business Value from Big Data Analytics: A Research Framework, Journal of Management Information Systems, 35:2, 388-423 ↩︎
- The Innovator’s Dilemma, Clayton M Christensen, Harvard Business School Press (1997) ↩︎
- https://www.rolls-royce.com/media/press-releases/2018/06-02-2018-rr-intelligentengine-driven-by-data.aspx ↩︎
- Stealing a March: Using Innovation to Beat the Competition ↩︎
- Building Service Capability ↩︎
