Where is artificial intelligence going in 2020? According to a recent Forrester research report, many companies feel an urgency to reap the benefits of AI. Indeed, artificial intelligence is seen as a propulsive driver of competitive success. If you’re not on the AI train, your competitors are leaving you at the station.
And yet there’s also a growing need to grapple with adopting AI in a pragmatic manner. AI can be wildly expensive, and companies have gotten burned doing “moonshot” projects. The mood is, “okay, we’ve heard that AI is big magic — now prove it to me.”
This discussion covers:
- What is the current state of AI adoption?
- Specific predictions for AI’s future.
- How companies are purchasing artificial intelligence solutions – from AI companies?
- Expectations for AI in 2020
Scroll down to see an edited transcript of highlights from”Data Analytics 2020.”
Download the podcast:
Edited Highlights – AI in 2020 video/podcast:
Tech trends come and go, and they have their 15 minutes of fame. Yet it feels like artificial intelligence is more foundational – it’s the Mega trend that will eat all other trends. Agree or is that just hyperbole?
“I think that the future of AI is always bright, and that’s one of the problems we’ve had with AI. In the dawn of the computing, computer science era in the 1940s and ’50s, before we even had proper hardware, some computer scientists thought that we would solve the general AI problem by the 1970s. And of course, that didn’t happen.
“We look forward and then there was what was called an AI winter, which was a period of disillusionment when people realized that practically speaking, certain problems could not be overcome.
“But in the last three or four years, we’ve entered this new phase of AI development where not only have hardware and software become more capable to move to the cloud that you mentioned and other factors making us feel like AI could be done effectively. However, when we come right down to it, we need AI to be in service of something, hopefully it’s in service of improved efficiency or customer obsession or operational effectiveness. And we have fairly mixed record at the moment on that.
“Finally, I want to say, AI is such a plethora of different technologies that it’s all over the place. In some areas, AI will very quickly become table stakes. If you look at what Alibaba has done in the retail sector, or Amazon in terms of personalization and choice and predictive analytics around what people want to buy. Well look, that is becoming really powerful and important, but other areas of AI are quite lagging and are definitely hyperbole today.”
The Forrester report says “the urgency to reap the benefits of AI is real.” So, are we still closer to infancy, or are we approaching mainstream in terms of AI adoption?
“So our data also shows that about 53% of data and analytics decision makers say that their company is in the implementing phase or they’ve already implemented some AI. But that is to say that within those organizations, that could be a small project.
“So 53% of companies are doing something, but that might be a Tensor Flow model running on one workstation for a data scientist. So it is again, rather variable.
“And I would say that in the grand scheme of things, we remain at an early stage of this, and there’s reasons for that, this is not easy to do well. We don’t necessarily have data hygiene that’s allowed us to tap into the right kind of data, we have these data silos and stuff. So, the foundation of AI being data, that’s a problem.
“We have a lack of governance, most companies don’t know from ethics to explainability to exactly who to participate in the process of overseeing governance. Very few companies have gone deeply down that road.”
How are companies actually purchasing AI solutions?
“So there’s a wide range. Again, with AI being such a broad area, a couple of important factors here. Number one, it can be very challenging even for a large company to hire AI talent. So I was talking to an insurance company that’s global in span. It’s a huge company, based in the Midwest, however. This is not an area where there’s a lot of local talent. They could choose to hire someone who sits in San Francisco, but it might be a bit of an inhibitor. And it’s also the cost of that talent can grow to the millions of dollars.
“Secondarily, you may not even have the basic organizational capability to set something like this up. And so you may choose to go outside. But there could be other cases where you do have a data science team and what you’re doing is more incremental. You’re building using open-source kind of solutions like Tensor Flow, which is common in the ML and deep learning spaces. And maybe you can start internally.
“So what we’re finding is a distribution, but many companies do turn to external experts, companies that are able to offer data hygiene, data engineering services that are offering a variety of different kinds of analytical techniques. Or they’re working off of a big platform like Microsoft Azure or AWS, which have their own AI tools from which you can build applications.
“Finally, there’s also what you could think of as everyday or embedded AI, which is to say, ‘I am already using software and the vendor of that software has decided to add AI features to make my experience better.’
“Now this is good because it means that I am not doing a bespoke AI project. It means I’m getting certain benefits of AI within an existing work stream. And sometimes, that might be provided by the people. Like you look at Salesforce Einstein which is, ‘Hey, we’re gonna apply AI to your sales leads within the Salesforce ecosystem.’ Or it could be done by someone like Microsoft Office 365.
“When you log into the latest version of PowerPoint, you start building a slide, it actually tells you, ‘Here are some things you can do. Do you wanna do these things?’ And that’s actually powered by AI. So there’s a large span of different ways to do this from systems integrators, who are gonna do big projects, to existing software providers to building off of a platform to maybe a little bit of internal work.”
In closing, I’d like to get your thoughts about where we are in the bigger picture – AI going into 2020 – and if there’s any way for companies to get ahead of that wave as you see it happening.
We’re going to continue to see a move toward, ‘Prove it to me. What are the business results measurement?’ I think that some of the moonshot projects in a year in which we don’t necessarily expect recession, but reasonably slow economic growth, companies are turning pragmatic. That was the theme for 2019. We think that will continue into 2020.
“We also see that AI will be applied to, I think, more specific business problems rather than these broader ones. We will also see an element of AI riding in on certain other technologies. Principally or in the leading case, it’ll be robotic process automation.
“So RPA has been the fastest growing technology area over the last two years in the enterprise, we’ve been looking very closely at that. And what makes an RPA bot – which is effectively relatively dumb and deterministic – what makes it effective is adding layers of artificial intelligence, whether that is machine learning to capture successes and failures, a better decision engine that takes advantage of those successes and failures, and something like maybe a natural language understanding. So the bot can source its directions from unstructured data, that sort of thing.
“So those are some big trends that we see next year. AI will be a critical part of the conversation for enterprise tech, but it will become a pragmatic part of that conversation.”