Looking beyond the hype with Deloitte’s Beena Ammanath
3 December 2019
Over the last two years, the ethical deployment of AI has been a growing concern for enterprises and technologists alike. But 2020 could be the year we really see it reach the forefront of AI development.
According to Deloitte’s second State of AI in the Enterprise survey, 32 percent of businesses rank ethical issues as one of the top three risks of AI, and most are yet to develop a specific approach to this problem. Meanwhile, 45 percent have appointed senior executives as AI champions – showing that AI is a growing strategic concern for enterprise leaders.
What isn’t clear is how this is going to affect the wider business landscape in 2020. Beena Ammanath is AI managing director for Deloitte Consulting LLP. Having previously worked at General Electric and Hewlett-Packard Enterprise, she has been interacting with AI for years – and crucially, has been on the frontlines of the technology through its entire hype cycle.
To better understand what AI will look like in 2020, versus two-four years ago, we sat down with Ammanath to discuss her expectations for key trends and new challenges in AI across both enterprise and industry.
Q: How did you get started with Deloitte?
My experiences at GE and HPE taught me that, as with any other technology, you need a trusted partner when you’re working with AI. But right now, AI is a swamp – there’s so much hype and news that my team are spending a lot of energy finding the right AI product and working out what to believe.
There is so much fluidity around topics like AI ethics and policy that you need a company which can look across industry lines, to not only help policymakers but be able to take those lessons to clients. Consulting companies can really help provide that cross-functional cross-industry perspective and help us leverage AI in the best way we can. I strongly believe this is the best time to be a consultant especially if you’re in the AI space.
Q: What should enterprises be looking for in an AI solution in 2020, compared to a couple of years ago?
There’s been a big shift. Three or four years ago, it was all about finding use cases that can bring value. People were asking questions like: how do we work on the right AI projects? Which are the use cases that AI can solve and can’t solve? Their priorities were around value addition with AI and how it could fit in with the business strategy.
Now, I do see organizations shifting a little more from value generation, to staging it all – especially on the ethics side. Those newspaper articles and headlines have moved them to begin asking asking: are we doing the right thing? What does a strong governance structure look like? What are the policies around AI and how do they differ between geographies and cultures? This is especially becoming prevalent at the board and C-Suite level.
At the IT and infrastructure level, it’s still more about the value, but ethics is a growing concern. The importance of hiring is much higher than it was four years ago, when it was okay to hire PhDs in machine learning and let them take control. Now, as we’ve seen, all it takes is one algorithm to go wrong and the CEO has to face the law. Being mindful and aware of how the algorithms are being built, and educating the workforce, is becoming much more important compared to four years ago.
Q: After years of hype, is disillusionment setting in around AI?
There are fewer delusions. The conversation is much more grounded in reality now than it was earlier. Earlier there was definitely more hype than real practical value of an AI project.
The good thing is that AI has brought more value to almost all industries. Fifty percent of these may be value added projects and 50 percent are not. This may not be the fault of AI but a lack of education. In legacy industries that have started machine learning projects, for example, there is a growing realization that they are not capturing the right data points and that they need to backtrack and collect new data.
This is because they are not sole tech companies and their data is 20-30 years old. So there is recognition of the need to invest a lot more in data prep and management to really get the full value of AI – it’s not a magic bullet that’s going to automatically add value.
Q: So where are we seeing the most growth in value right now?
For legacy companies, it is in business process optimization and automation. There are two ways to deliver value with AI – cost savings and increased productivity, and new revenue opportunities. Right now, most of the value is emerging in cost savings and increased productivity for these companies.
For companies that have been created in the past 10-20 years, there is definitely much more value because they’ve been built in a way where data has been captured upfront, and it is much easier to drive insights because of that.
Q: What kind of challenges and stumbling blocks for AI do you see emerging in the next year?
As I’ve said, ethics is going to be a big one along with different aspects of privacy, including data transparency and security. These are conversations that we should all have been having five or six years ago, but we didn’t, because we were all chasing the value AI can bring, rather than thinking about what happens when it goes wrong.
There’s now a need for privacy, security and clear lines of accountability if a machine learning model doesn’t work the way it is supposed to. Enterprises need to establish who is responsible. Is it the CEO, or the CIO, or the CTO whose team developed it, or is it the data scientist that built the model? These conversations need to be had and I don’t see it happening enough.
One other thing we need to think about with AI solutions is around traditional software systems, particularly in reliability and robustness. In the past, when you developed a software solution and deployed it you knew it would be consistent in the results it produced and in the way it behaved.
But machine learning keeps learning, right? It is continuously training and learning from the data it is given. So the software that you deploy now can look completely different two months from now depending on what data it receives.
Q: What are going to be the major technology developments in 2020 then?
There are three developments that I see emerging in 2020:
Explainable AI / traceable AI: As AI continues to find its place in operating models and within business strategy, it will be very important to ensure that those that are working with machines understand how the AI works as well as consumers of organizations that utilize the technology. Transparent or explainable AI allows humans to see the models that machines are using to make decisions. This fosters more trust in AI, as well as gives people more insight into if decisions are being made responsibly and ethically.
Voice and Video bots > Chatbots: While chatbots will continue to have their place, I believe we will start to see more voice and video assistants / voice bots and video bots start to emerge as technology continues to mature. Every user interface will be on the path to be voice-enabled – including online.
Hyper-personalization: As businesses mature their data processes, they will gain deeper knowledge of their customers’ behaviors and preferences. By merging analytics and insights with AI, companies will be able to hyper personalize every experience, including IT support and services, to build customer loyalty and engagement.
Beena Ammanath is AI Managing Director for Deloitte Consulting LLP. Catch her and the Deloitte team at The AI Summit New York, December 11 – 12. Click here to find out more.