On November 6th, 2018, on the third day of the Gartner Symposium/ITXpo, which took place in Barcelona, Magnus Revang (Senior Director Analyst, Gartner) shared his knowledge on Artificial Intelligence (AI) and focused on the strategic decisions around it. His presentation was entitled "CxO perspective on Artificial Intelligence".
"AI is on top of everybody's mind. All executives and top managers are now looking at it as an enabler to reach success," started Magnus Revang, who's an expert in customer experience. As a matter of fact, he started his presentation by drawing a parallel between CX and AI in terms of evolution, moving from a trend to strategic decisions. Yet, several questions remain: how does AI translate to what I do? How can it be used for success?
According to the latest CIO Survey, just 1 in 7 CIOs says they have already deployed Artificial Intelligence. "Sure, they are loads of Proofs of Concept (PoC) and planning, but only 1 out of 7 are in production. Therefore, many CIOs think they are behind on AI, but actually they're not," added the Gartner Analyst. On the other hand, CEOs see it as the most important advanced tech for the future and expect a lot from it. They also see the market and business change because of it: they are clearly asking CIOs to work on it and find the best use cases. It is also important to highlight that the real time to get an AI project running is close to 3 or 4 years. "The first year is about experiment, the second one about production. In year 3, companies will probably still be piloting and the project will only be deployed the year later. Therefore, AI suggests a strategy as it means a long effort," explained Magnus Revang.
Myths of AI and how they will trouble you
One of the first myths deals with the fact that "AI is learning without training". To highlight the fact that AI actually does need training, Magnus Revang shared the example of a predictive model used for a call center where AI needs to be fed with loads of phone calls data and their outcomes. Some of the other myths are, as follows: "AI learns as it goes", "you turn AI loose on things and it works," "the same AI algorithms work for everything, "you can buy 'an AI' from a big vendor", etc. "A huge amount of work is requested, and it is actually a lot of manual work, feeding data into the system. AI and the solutions sold by venders is not a silver bullet," added the AI expert.
Necessarily, working and eventually deploying AI is not risk-free. As it redefines the way people work and even creates new business models, there are several operational risks. Moreover, the cost of implementing AI needs to be anticipated. As it is powered by data, several challenges rise around information in terms of privacy, compliance, geopolitics, communication, etc. According to Magnus Revang, "it is not enough to have a strategy for AI and companies need to keep investing in it but also to target the people responsible for all of these risks. Training them is also crucial".
He then listed some of the challenges CIOs face in the adoption of AI: 79% of IT experts actually fear the unknown (notably 37% fear security and privacy issues, 32% risk and liability), 40% do not think their company is mature enough (20% admit to have governance issues, 23% explain they lack skills, etc) and 48% believe it is due to vendor strategy (is the product/service available in the market?). Moreover, 63% need to find their starting point (use case, strategy definition, funding), but actually, most CIOs will have to go through these 4 steps before being able to scale AI.
Benefits and strategy
"The benefits of AI do not rely on tech. They depend on the practitioners who need to take care of it. CIOs care about the strategy, but also on how to improve the business and therefore money," added the Gartner Analyst. First of all, the expert advocated the creation of a center of excellence where data scientists can focus on operations and interoperability by feeding data directly in and gathering insights. Magnus Revang also insisted on the need to concentrate on augmenting Human decisions and therefore developing Augmented Intelligence rather than Artificial Intelligence: "every single AI project takes on some of the heavy tasks Humans are doing. For instance, a chatbot takes 70% of all inquiries, allowing the employees to spend more time on value-added tasks. Tech need to augment Humans". Still talking about chatbots, he explained that there are currently more than 1000 vendors dealing with them. "There's not strategic vendor in AI. The only thing that's strategic is your data. Focus on the outcomes, be tactical, think about an exit strategy, while always keeping your data strategic," he underlined.
Magnus Revang then advocated the launch of multiples small AI projects and PoCs at the very same time, keeping in mind that all won't deliver their promises. "Things add up. The same skillsets and approaches are needed to only multiple PoCs can make it happen. There's no single project that will save your world. Learn first and see through those expectations," he added.
When it comes to employees and their perception of AI, most of them seem to prefer having AI as an assistant, rather than as a manager or even as a collaborator. More precisely, 52% of them would rather have it as an "assistant on demand", when only 32% wish to have it as a proactive assistant. In this respect, HR and management need to directly communicate and exchange with all the members of the IT department to reassure people in the use of such techs, but also in its outcomes.
"If you do AI to pretend that innovation is happening, you're doing it wrong. Approach it because it will make a difference and pick a use case where AI is better, with clear outcomes," he added, before listing some of its benefits: improving prediction (with automating analytics), more accurate decisions (using predictions at scale to test hypotheses more aggressively), enhanced understanding through conversational interface and augmented analytics, all with the mission to allow the company and its employees to move faster and therefore provide better solutions and services.
Uses cases and actions will get you along
According to the Gartner Analyst, AI will mature and spread through delivery models, including smart "swarms". Currently, CIOs and companies can already pick use cases and implemented AI in a lot of ways, even by embedding it in products. He also highlighted the fact that finding a use case that is relevant can be tough as most of them do not have vendor support. Therefore, IT teams often have to do custom development. With data and outcomes, such projects can start, but CIOs need to keep in mind that AI may not be better than other technologies, depending on the outcomes and results you want to reach. Magnus Revang then shared a use case developed by Georgia State University: the private college noticed that many students didn’t show-up on the first day, after getting the scholarship. They simply didn’t attend. A team conducted many interviews and discovered that the fear of the unknown was one of the main reasons for not showing up. A simple chatbot was created with mappings, questions and answers, "nothing advanced at all", as explained by the Gartner Analyst. People started using it and 4% more showed up, consisting of more than 300 students. "It's not a huge project which aimed at providing a miraculous solution. But developing several small AI solutions will clearly help the university become better. These targeted solutions are only possible using new tech and make a tangible difference," he explained.
As a conclusion, Magnus Revang explained that it was key for companies to get a center of excellence in order to "conquer the unknown, hire the right people and get ready to start", but also to develop a strategy of data stewardship and data gathering. Finally, they should set up priorities for many use cases and proceed…test, adapt: a virtuous circle.