For this Present Value post, ESI’s thought leadership team sat down with experts participating in our Driving ROI Through AI research program to gain deeper insight into how companies around the world are currently using artificial intelligence, where the technology is headed, and how it will impact the future of business.
Driving ROI Through AI is a multi-client initiative launched by ESI ThoughtLab aimed at providing executives with the market intelligence, strategic insights, and data needed to help them use AI to drive corporate performance.
In your opinion, what are some of the most challenging issues organizations currently face in adopting artificial intelligence?
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: One of the biggest challenges is articulating the business problem they want to solve. Often, companies start with the technology, saying “I want to use AI” and then look for the business problem. Even after defining the problem and how to solve it, they need to understand the data they have access to, as AI systems are only as good as the training data we put into them.
Kurt Muehmel, Chief Customer Officer, Dataiku: At Dataiku, one of the biggest things we see organizations struggling with today is really scaling AI; that is, going from executing on one or a handful of use cases successfully to tens or hundreds. More and more, companies are discovering that leveraging AI at scale is the only way to make it profitable in the long run.
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: More often the biggest obstacle is the inescapable fact that stakeholders are not ready for a new way of thinking. Without a cultural transformation preceding the newly executed digital strategy, the “AI is a new technology, and let’s adopt it like any other new technology deck” approach does not work. This costs companies precious resources without clear ROI, causing some of the justified skepticism.
Peter V. Henstock, Machine Learning & AI Lead, Pfizer: Adopting AI is a major transformation for companies. It’s asking corporations to make their data not only accessible but also findable and connected, perhaps for the first time. The mere processing of this data may require specialized hardware that may not even be used in all companies but may be available through the cloud, so the infrastructure must change. It then calls for a group with a different set of skills to find patterns, predict outcomes, and help make decisions from this data.
Mihir Sharma, Head of AI & Cognitive Initiatives, Financial Services, North America, Publicis Sapient: There is a general scarcity of data scientists within the market. Firms are facing a challenge in understanding how they can restructure and transform their business to meet the needs of a digital world. There are very few employees within firms today who understand the possibilities of leveraging AI to transform their business.
How can these challenges be addressed?
Kurt Muehmel, Chief Customer Officer, Dataiku: There are two components to addressing AI scalability. One is organizational; it’s critical to have the right setup that both standardizes processes and enables people across the business to leverage data for more and more diverse use cases. For a lot of companies, that means having a center of excellence combined with robust programs for training, gamification, or whatever it takes to ensure data initiatives don’t stay siloed in this central team. The other component is reusability. Common sense and economics tell us not to start from scratch every time, and that is exactly the principal behind reducing costs associated with data cleaning, preparation, operationalizing, model maintenance, and even hiring woes.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: Many of these challenges can be addressed by creating a cross-functional team—a product owner who is leading the direction of the investment paired with team members from product, engineering, and data science, who can collectively focus on a business problem and attempt to discover a solution. It may be that the business problem does not need AI as the solution, but leaders need to give the team the chance to solve for a narrow, specific, and important business challenge. In the case that the business problem warrants an AI solution, having gone through this exercise enables the team to have plans in place to solve for the data challenges, or look for partners that can help them launch world-class AI.
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: Culturally evolving, continual re-skilling—starting at the very top—and redefining the “job description” of all roles across the organization.
Peter V. Henstock, Machine Learning & AI Lead, Pfizer: The challenges exist for companies of all sizes but the volume of data of large companies represents a larger stumbling block. Newer and smaller companies can transform their data much faster than the larger companies. For companies with a legacy record, the approach is either incremental or may require a larger digital transformation.
Mihir Sharma, Head of AI & Cognitive Initiatives, Financial Services, North America, Publicis Sapient: Senior leaders within the firm need to agree on a clear AI and quantitative strategy with focus towards improving data quality. The key lessons learned from our experience is that there must be a healthy balance between AI engineers focused on evaluating the latest academic research and ML engineers adopting and applying it to use cases to showcase incremental value.
For organizations just beginning to consider the use of AI, what advice would you offer to help them succeed in their efforts?
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: Hire AI domain experts who have demonstrated impact in multiple domains. Do not limit your leadership search to senior experts who have decades of expertise in that specific domain. Disruption and transformation can happen with a fresh perspective.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: Focus on the business problem you want to solve and narrow down the first use case for AI. Make sure the team working on your first AI application is empowered to justify why AI is well suited to solve the problem at hand, and articulate what data will be used in service of solving the problem.
Kurt Muehmel, Chief Customer Officer, Dataiku: We always recommend carefully choosing not one but several use cases with which to start. With dozens of potential use cases but limited resources, that means prioritizing projects that have both high business value and a high likelihood of success. And why several use cases? Well, the reality is that AI isn’t always the answer; that is, applying machine learning to a problem won’t always provide results that are more effective than what the business was doing before. But one failure doesn’t mean that the organization should abandon AI efforts entirely, it just means that they haven’t found the right use case.
AI covers a range of technologies that allow machines to imitate human behavior (robotic process automation, machine learning, deep learning, computer vision, natural language processing, etc.) Of these technologies, where do you see the most investment currently being made and what is the underlying reason for that investment?
David Donovan, Executive Vice President, Financial Services Lead, North America, Publicis Sapient: The answer to this question is really dependent on where firms are with their maturity and adoption of AI within their processes and their business strategy and transformation. We are seeing investments made on all of the above. Most investment by financial firms now is on NLP and RPA focused mainly on reducing operational cost, while there is some experimentation towards alpha generation as well, but not comparable to the focus towards the former. We are also seeing big investments in customer centricity and targeting when it comes to sales and marketing.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: I see the most investment currently being made in tools to make AI more approachable and easier to use and deploy. There is a high demand for data science talent that has the training, experience, and sophistication to build and deploy world-class AI with confidence. There is always a journey from pilot to production, and smart tools can help shorten the time to market.
Where are you seeing organizations place responsibility for AI management and implementation? Which executives are in charge? Is AI management handled in a centralized or decentralized manner in organizations?
Peter V. Henstock, Machine Learning & AI Lead, Pfizer: The success of AI is governed by two factors: the amount of clean data readily available, and the maturity of the industry in its acceptance of data-driven analyses. Older companies can have a wealth of data but may also have a disadvantage in organizing that data compared with younger companies like Uber that have created their data platforms with analytics in mind. The healthcare field has stricter rules for leveraging data and making decisions from them.
Kurt Muehmel, Chief Customer Officer, Dataiku: I don’t think there is just one answer to this; we’ve found that it depends a lot on the size of the company, what industry they’re in, and how far along they are on their path to Enterprise AI. However, we do see in very large organizations that AI management handled centrally with a center-of-excellence model combined with initiatives to ensure infiltration throughout the lines of business is the most common recipe for success.
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: The size of the organization, and culture, play a role in making that decision. With decentralization comes a lot of waste and misalignment. Centralization is difficult. Each strategy requires a specific type of leader in the driver’s seat.
David Donovan, Executive Vice President, Financial Services Lead, North America, Publicis Sapient: In some cases, there is a data organization with a softly appointed Enterprise CDO and a separate Enterprise analytics/AI function, which is obviously heavily data driven. Clarity of responsibilities across both functions is so important to avoid constant friction over who is accountable for what.
What industries will see the most benefits from AI adoption and why?
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: Industries that are more standardized and rely on repetitive work will see the most benefit. The impact of AI is industry agnostic, but it tends to benefit back-office operations first, often not directly interacting with customers. Use cases range from business process automation (invoicing, for example), automation of inventory management, or claims processing in the insurance industry.
Kurt Muehmel, Chief Customer Officer, Dataiku: I’m of the belief that when it comes to seeing returns from AI, it matters less what the organization’s industry is and more what the predominant mindset is on AI from both the top down (leadership) as well as the bottom up (i.e., are individual contributors empowered to use data, and are they excited about it?).
What are some of the most important use cases for AI across enterprises?
David Donovan, Executive Vice President, Financial Services Lead, North America, Publicis Sapient: This world is already seeing a massive disruption, with companies not adopting digital business transformation either being already extinct or running the risk of not being relevant 5-10 years from now. Consumers and the enterprise are demanding a much more immersive experience when dealing with a company. AI helps to create a deeper understanding of and more personal relationship with the end client.
Kurt Muehmel, Chief Customer Officer, Dataiku: Every industry has its own unique use cases, so it’s very difficult to generalize across all enterprises. Of course, some examples of common use cases are marketing or finance related, such as churn or attribution and forecasting, respectively. But those aren’t necessarily the most interesting use cases or the ones that will really give organizations a leg up in the race to AI. In fact, at Dataiku, we believe that applying machine learning in these use cases will become status quo in the very near future. It’s the use cases that are very specific to the business’s operations that will become the most important and bring the most value.
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: While companies are beginning to transform, they are looking for short-term ROI. Marketing is one of these areas where AI brings in measurable ROI. This is the obvious. Companies that will differentiate themselves in the long run will be the ones that can plan for the long term.
In general, how knowledgeable do you feel the general public is about AI?
Mihir Sharma, Head of AI & Cognitive Initiatives, Financial Services, North America, Publicis Sapient: In our opinion, the general public is extremely green field with very limited knowledge of AI, although they do consume AI on a daily basis over their phones, or through augmentation of AI within their day-to-day activities through AI-powered companies, appliances, etc.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: While the general public interacts with AI every day through their search engines, social networks, virtual assistants, maps, and other smartphone apps, they are not that knowledgeable about how it works. Often, they don’t know they are interacting with a product or service that has AI embedded in it, nor that those technologies use training data curated by people to make that AI work in the real world.
Kurt Muehmel, Chief Customer Officer, Dataiku: Unfortunately, much of the media coverage around AI doesn’t have anything to do with Enterprise AI–it’s much more focused around consumer-facing AI (e.g., self-driving cars, smart home systems, etc.). In some ways, that makes sense, because those are the devices that are tangible manifestations of AI, so they’re somewhat easier to understand. But it’s naïve to think that Enterprise AI doesn’t affect the general public. Think about algorithms banks are developing to determine whether a client is eligible for a loan, or even AI systems determining pricing used by e-commerce systems. I think people are much less educated about how these systems are developed, how they work, and how they affect their everyday lives. Again, instead of focusing on these things, unfortunately, much of the media coverage around Enterprise AI has been more focused on fearmongering around AI taking peoples’ jobs. This is unproductive and prevents people from really taking the time to understand Enterprise AI at a deeper level.
Peter V. Henstock, Machine Learning & AI Lead, Pfizer: AI has been lurking in the background for generations and the public has many misconceptions. In addition, the recent wave of AI has caused major transformation within just a few years. A related question is who has the knowledge within corporations and who is driving the strategy? It requires a different skillset and a different vision. Finding the experts who also understand the opportunities and space is very challenging.
Do you feel there is any current or foreseeable backlash/fears on the use of AI?
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: There has been a lot of conversation about jobs and how AI will impact them, as well as the “black box” decision making around deep learning technologies.
Kurt Muehmel, Chief Customer Officer, Dataiku: I do think there are fears and that perhaps that will lead to backlash, but ultimately at Dataiku, we believe in human-centered AI. That means not sitting back and letting AI systems make decisions for us, but instead using AI to enhance—not replace—people. If everyone follows this principal, I think people will see the positive and embrace instead of condemn AI.
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: It’s inescapable to hit a point when the market behavior will push companies to prioritize for short-term ROI. AI operations will be affected. But I do not see any evidence for a global backlash. We will manage to regulate and learn as we reap the benefits along the way.
Peter V. Henstock, Machine Learning & AI Lead, Pfizer: There is a buzz and excitement about using AI. It offers an opportunity to leverage data more fully than it has been utilized before and in new ways. We hear about backlash in areas such as Robotic Process Automation, but it mostly offers opportunities to do better science, better analysis, and achieve better results.
David Donovan, Executive Vice President, Financial Services Lead, North America, Publicis Sapient: There has been fear around cannibalization of jobs, but over time as AI tech matures it will open up greater opportunity for human involvement, as in order for AI to be most successful it has to co-exist with humans.
In the spirit of Building Ethical, Responsible, Reproducible and Explainable AI models, model governance is going to be in the forefront with regulations mandating the same sooner than later. Bias detection and fairness are among the examples that are to be injected through the modelling process, which is going to be absolutely key.
More broadly, how important will AI be in the future for both business and society?
Mihir Sharma, Head of AI & Cognitive Initiatives, Financial Services, North America, Publicis Sapient: Absolutely crucial without an ounce of doubt. Our lives have already seen AI starting to play a huge role in our day to day, be it our phones in using speech to text, or using Alexa/Google Home and other AI-powered devices, to autonomous and driverless cars coming in near future, to AI-enabled home appliances, and the list can go on and on. Because of big tech like Apple, and Amazon, consumers’ demand bigger experience, and AI will help create more personalized, real-time experiences which can unlock unique information and lead to a great experience with consumers.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: AI is just a collection of technologies—nothing more, nothing less. It will be important to use them responsibly and effectively. The technologies have already permeated everyday life and disrupted most industries—I think this trend will continue. AI is becoming less optional for companies. Statistics suggest that it can improve customer experience, productivity, etc. making it a smart business decision when done correctly.
Kurt Muehmel, Chief Customer Officer, Dataiku: It will be such an important part of life in the relatively near future that, like electricity, most of us will start to take it for granted and get to a place where we don’t even realize it’s there.
Dr. Bulent Kiziltan, Chief Data and Analytics Officer: Anything I say will be an understatement!
In what ways are you seeing AI being implemented to combat the outbreak of COVID-19?
Bill Hobbib, Senior Vice President, Marketing, DataRobot: There are a tremendous number of AI use cases to try and combat the outbreak of COVID-19. In healthcare, life sciences, and public health, the categories and use cases include the following:
- Detection, pandemic trends, and analysis: early predicting; herd immunity forecasting; alerting of towns/counties of spread and expected extent of outbreaks; actively predicting deaths, peaks, and diminishing of pandemic, to better predict medical resource needs and safe reopening timeframes
- Virus containment: predictive enforcement, behavior analytics, detection and correction of “fake news”
- Hospital/healthcare operations: AI-assisted patient chatbots and virtual assistants, hospital staffing, augmented triage, image recognition, pneumonia detection and diagnosis, hospital supplies and equipment forecasting and supply chain, patient throughput and capacity management, remote patient monitoring and alerting
- Life sciences vaccine or treatment research and development: information retrieval, predictive analytics, R&D coordination & collaboration
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: Technology platforms, government, academic research institutions and healthcare organizations are partnering together to try to diagnose COVID-19 more accurately, predict and model the spread and impact of the disease, and find solutions for medical treatment as well as economic impact. As an example, there is a project using computer vision to diagnosing CT scans from patients suspected of having COVID-19. Another example, AWS used its extensive modeling platform to help scale and improve the models that were predicting how the disease was going to spread in communities. The project successfully reduced the model run-time from a week to 12 hours with many variations.
Mihir Sharma, Head of AI & Cognitive Initiatives, Financial Services, North America, Publicis Sapient: Publicis Sapient is currently engaged with a large global investment management firm helping them build out their AI & Investment Research platform. Specifically, in response to COVID-19 we have enabled access to varied sets of alternate data as part of the AI platform, for the macroeconomists and front-office investment professionals to be able to consume. This includes COVID-19 tracking (Recovered vs. Confirmed vs. Deaths by County/Country/Region/State), testing results, travel datasets (Canceled flights, Traffic Congestion levels, Airbnb, Hotel datasets), hospital beds, ICU availability, and more.
Are you seeing any new trends in the adoption of AI use given that many businesses are closed, remote, or socially distancing their interactions with clients and customers?
Ari Kaplan, Director, Industry Marketing, DataRobot: Across the board, companies need to rapidly find signals in the unprecedented noise of the changing world. Models used for predictions and other business decisions are based on data that is limited or no longer applicable. Businesses are looking to AI now to more rapidly gain insights based on the changing economy, shutdown of physical businesses, the move to online purchasing, and changing interactions with their clients and customers.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: Yes, we are seeing new trends in adoption and addition of AI given the shift to remote work and disruption to the normal economy. Those trends include the increased demand for remote work—on our platform alone we saw a 31% increase in user activity in March. Additionally, we are seeing our customers double down on their investment in AI and invest further in use cases such as combatting disinformation, combatting fraud using, and content moderation using AI. Customers are looking for ways to automate processes and reduce cost in their businesses or respond to changing demand needs. One social media company is looking to further personalize content. Advertisers are looking to understand the content they are promoting to users, so they don’t strike the wrong tone given the pandemic.
Mihir Sharma, Head of AI & Cognitive Initiatives, Financial Services, North America, Publicis Sapient: Yes definitely. We are seeing added leveraging of call center AI solutions, using virtual assistants powered by NLP and voice to help manage the sudden increase in call volumes that most firms are facing, and better manage workforce planning.
What role will AI play in preventing and combatting future natural disasters, pandemics, and global crises?
David Donovan, Executive Vice President, Financial Services Lead, North America, Publicis Sapient: AI will be instrumental in future prevention and combatting of natural disaster, pandemics, and global crises. There will be huge advancement and focus on medical research in order to prevent this kind of an occurrence in future, and AI is supposed at the core of that research.
Alyssa Simpson Rochwerger, VP, AI and Data, Appen: AI can play a role by modeling future crisis and recommending mitigation strategies such as predicting the impact of climate change on food crops and recommending different seeds or agricultural strategies which use less resources and provide greater output. Furthermore, AI applications can calculate the data around the number of damaged homes and structures or high flood levels to provide that information for first responders to make quicker, more accurate decisions to coordinate response and recovery efforts.
Ari Kaplan, Director, Industry Marketing, DataRobot: A bright spot in the COVID-19 pandemic has been how people and companies across the AI spectrum have come together to collaborate with healthcare, government, and academia. As a result, mankind has been better equipped to understand and minimize the crisis, with AI as a centerpiece of this effort.
Global crises come in many forms, and AI can help predict when and how they arise, and once they do, how to minimize their harm. These crises can come in many forms: financial collapse, political instability, biological/chemical/electronic/terrorist or conventional warfare, famine and water shortages caused by global climate change. AI can help model scenarios so that governments, businesses, and populations can make more informed decisions to mitigate the risks. And if a global crisis does indeed arise, AI can help inform with supply chain for supply distributions, optimizing staffing levels, and running scenarios for appropriate financial assistance for optimal resolutions.