2019 was when Artificial Intelligence (AI) started living up to its hype. Do you think 2020 is the year when AI will seep deep into the enterprise DNA?
John Gikopoulos (JG): When a new technology is still at the hype stage, everyone is so awed by its potential that no one talks about its problems. At the peak of these inflated expectations, the technology is hailed as something almost miraculous. The discussion around adoption obstacles or the challenges that the new technology will bring is almost non-existent.
Last year this was the case with AI. The hallowed initials AI were appended to every conceivable offering, much like it happened with IoT and Cloud in past years. However, people are now beginning to think much more critically about AI and are recognizing and talking about the challenges.
So, in my view, 2020 will focus on debating and solving the critical challenges that the widespread adoption of AI has revealed. This is the time when Artificial Intelligence needs to move beyond being a trend, a hype, and needs to become a reality and start delivering value.
Process-heavy, data-intensive sectors such as financial services, healthcare, and manufacturing will likely see the most impact from AI. There is enormous scope for automating manual processes such as algorithmic trading; delivering insights for areas such as drug discovery; and identifying efficiencies from the shop floor to the supply chain.
What are some of the key developments you envision for the year ahead? How can enterprises cut through the clutter and apply AI for impact?
JG: I believe the most important developments in the field of AI in the year ahead will be the following:
We have quite often seen enterprises getting stuck in AI PoCs. How can enterprises move ahead and do things differently to nail and scale AI?
JG: There are two key issues companies face when it comes to using AI:
1. Considering it as a technology and not as something that requires organizational change
2. Lack of know-how to scale the usage of AI to an enterprise-wide service.
These pitfalls can be avoided by the ongoing collaboration across teams and by considering the broader cultural and organizational changes necessary to become a mature AI business. The following four-step approach should be employed:
Define your AI specific vision and aspiration.
When defining your vision, think cross functionally and across the organization, think holistically and have the end game in sight. Do not follow trends or hype and do not accept the limitations and directives of competition and vendors.
Calculate the value creation potential and identify the right enablers.
Spend time upfront doing the math, identifying obvious and hidden costs, talk to organizations that have undergone similar journeys and comprehend what impeded progress.
Choose the right use cases by prioritizing.
Identifying the starting point can make the difference between failure and success. Being able to tackle big and high worth projects is important, but implementation obstacles and challenges need to be surpass-able. Being able to deliver value to your organization within a 3-6-month window will build trust capital that you can cash in later on to overcome soft hurdles (organizational mindsets and change management).
Set yourself up for success.
Create a joint IT-Business Team that seamlessly coexists to translate business to IT requirements and obstacles back and forth. Ensure senior sponsorship to resolve bottlenecks at the highest level.
Across organizational change and AI implementation, organizations should also consider AI governance. The governance layer needs to suit the firm’s level of AI maturity and the level of new risk introduced. At the same time, it must be built to be rapidly scalable as needed. Building a solid foundation for AI governance will help organizations manage programs and spend effectively and deliver high-quality services to customers.
And finally, organizations must have a roadmap in place for AI-related skills. Skilled and experienced AI resources are still scarce so expanding awareness of AI at both business and technical levels is vital to taking AI from experiment to broad scale production usage. AI skills and expertise should not just be in the domain of technology teams; instead, the opportunities, implications, and responsibilities should be shared across an organization.
The question that any business must ask about their AI deployment is whether the technology is helping us make better-informed decisions. That, ultimately, is what AI means for business. The process Automation, the ability to learn, the data crunching-these are all the functions rather than the goal of AI.
The debate around ethical AI has garnered a lot of eyeballs. What are your views on this topic?
JG: One of the most important ethical questions is about intention – for example, where, when, and how AI interacts with the analogue reality that we have all been accustomed to living in? Who is going to decide what is the right playing field and what constitutes acceptable or unacceptable uses of AI? Regulators and legislators are trying to define
and control the degrees of freedom, but they are working at analog speeds. We need faster answers to questions that we’re already facing today.
What is your advice for companies who are struggling with their AI Initiatives?
JG: There are four things enterprises will need to do to ensure that their organization can take advantage of all that AI has to offer in the future:
Acknowledge that AI will be an end-to-end journey
Many firms are trying to implement AI and Automation solutions in chunks, which makes sense as there is a tremendous amount of risk involved for most companies. The problem is that these chunks tend to be built in layers, meaning they replace one part of a process instead of end-to-end, resulting in limited value. For instance, customer service ChatBots in most companies today can
only offer very rudimentary, FAQ-type support. When the customer actually speaks to a human agent, they need to repeat their question, and this leads to frustration and makes the customer experience worse. Organizations need to ensure they implement AI in vertical chunks rather than horizontal chunks-covering one process end-to-end.
Work with HR to reskill employees
Companies are seeking developers, data scientists, and solution architects to build in AI processes. However, demand for these professionals far outpaces supply. There’s only so many people that you can get straight out of university and into these heavy-duty processes to manage end-to-end customer journeys. Hence, retraining and repurposing individuals working in IT and the business side to become data scientists and solutions architects is important in the near term. The next generation of data scientists and solution architects are going to be 40-year-olds, tenured people who are retrained and repurposed to carry out this type of activity.
Spearhead smart technology investments
Messaging about the path forward toward AI and Automation in any organization needs to come from the CIO. Technology is at a level right now that it can support and realize Artificial Intelligence and Automation solutions. The need of the hour is to choose the right process and the right tools to deliver on this potential.
Lead the ethical discussion
There is a lot of talk about how humans can ensure that machines, systems, and algorithms do not expose or take advantage of the information they are supplied with. We’re looking at this the wrong way. We’re trying to manage the aftermath that the bullet might cause once it has left the gun. Organizations need to be extremely careful about what they plan to do with AI before they begin using it, which means taking into account potential issues with cybersecurity and biased algorithms.
Could you please share some real- life use-cases of companies who are thriving with their AI Adoption and what would your advice be as they go through their journeys? (Things they did right that other companies can benchmark against).
JG: Two shining examples of AI at work are a British educational institution employing AI to predict and proactively manage student attrition and a global cosmetics manufacturer who has deployed a B2B2C assistant avatar that learns as it assists the company’s retail partners.
The educational institution, one of the world’s first remote schooling establishments, was facing substantial student attrition to online learning. By looking at the challenge holistically, the institution managed to tap into all its different data sources. Consistent cohort analysis helped them predict the chances of a student failing or dropping out of a course pretty much at the moment of sign up. By adopting different communication channels to convey the message as well as offer alternatives, the institution has managed to drop attrition rates by at least 30%, thus allowing it to re-establish a viable business model in the way it delivers remote learning.
For the cosmetics manufacturer, the focus was on deploying an advanced chatbot solution that not only interfaced with / reacted toward retailers but also had a real ability to learn and adapt. After a short gestation period, a human-like avatar has been deployed as the main interfacing channel with the company’s affiliated retailers. The avatar acts as a (product – services) knowledge management as well as a sales assistant / advisor interface. The affiliated network received the avatar so enthusiastically that it is now used as the internal ‘face’ for company events and communications.
In both these cases, and despite early successes, senior management commitment has remained very high, and business impact expectations have, actually, heightened. My advice to both organizations has consistently been to create wide-reaching but straightforward internal communication campaigns to ensure that AI successes help shape their respective culture and DNA, thus making identification and implementation of future use cases much quicker and effective.
What are your views on the convergence expected to happen between AI and Automation in translating insights to action and creating an exponential impact?
JG: The most critical challenge in the transition between Automation and AI-focused initiatives is the level of understanding, the tangible impact if you like, between the results achieved with solutions within the two domains. In most companies, users, SMEs, product owners, and also senior executives are much more accustomed to the outcomes of an RPA or even I(intelligent) PA focused enablement vs. what can be expected from the advent of AI-related solutions.
Going forward, organizations will start realizing that the boundaries between Automation and AI are blurring, just because outcomes from AI enablement initiatives will start becoming more tangible on a day-to-day level.
The likes of IoT enabled ML solutions, advanced Digital Twin types of modeling, self-cleansing / self-healing databases will end up becoming the expectation rather than the surprise outcome or exotic demo that they tend to be today.
The key to mastering this transition will have to be the faith in AI enablement outcomes with the notion of “Safe being Risky” becoming part of the advanced enterprise culture.