Earlier this week, the Cutter Consortium released an in-depth AI review reporting that most companies were deploying AI badly. Like any new—or potentially amazing new—tool, few people initially know how to use it effectively, and most are oversold on capabilities that may not be achievable with the current generation of technology.
What is needed is a way to determine what AI can do, where an enterprise most needs it and how you can learn from others’ mistakes to avoid the expensive path of learning while doing. To that end this week, Dell EMC had a briefing on its approach to AI.
Go here to see a listing of eWEEK’s Top Predictive Analytics Companies.
Dell traditionally has been one of the most pragmatic companies in the segment, so its approach was more about the practical and effective application of AI and not the more common unreasonable promises of some future, although admittedly more amazing, potential offering.
Let’s talk about that.
Pain Points
As with any new technology, there are a ton of pain points that most vendors don’t seem to want to talk about, and it is to Dell’s credit that its reps started by identifying them. This approach is important because by focusing on the pain points, their solution should evolve less by what can be done and more by what the customer wants. Often these emerging technologies are crippled because engineering wants to showcase imagination rather than listening skills, with the result often drifting painfully away from where the customer would like it to go.
The pain points that Dell EMC identified were grouped in two classes: pain points by the data scientists who use the AI offerings, and pain points by the poor IT folks who must deploy them.
Data Scientist Pain Points: The big issue is that data scientists are frustrated with current AI solutions because they can’t yet take full advantage of the massive (and often wrongheaded) big data repositories they have and, even when they can properly form a query and get a response, it comes too late to be useful. If you read between the lines, a solution that either doesn’t work or doesn’t work timely isn’t worth much, and these AI tools aren’t cheap. It makes you think that a lot of firms that have deployed AI secretly now would like their money back.
IT Pain Points: This dovetails with the data scientists’ pain points because IT complains it lacks the needed skills to deploy the technology effectively and don’t have the ability to evolve IT operations to embrace this technology set. No wonder the data scientists are complaining; it would be like talking to race car drivers using cars built by people who didn’t know how to build them in the first place. These drivers certainly wouldn’t be happy (I expect a lot of them would be injured or no longer living).
Dell EMC Approach/Solution
Taking these pain points into account, Dell EMC has focused on simplifying its AI offering and creating training so that IT can not only deploy more quickly but also more effectively. Working with AI partners NVIDIA and Intel to both optimize their solution for the available technology and ensure its successful deployment, Dell is differentiating its offering by better assuring the success of the result.
Because NVIDIA and Intel aren’t known for working well together, Dell EMC is bridging the technologies with its IP to optimize the result and provide a stronger ROI as a result. And, talking about the result, this has reduced the deployment time, according to Dell EMC, from months to weeks and significantly improved the ease of use.
Results
Examples of where this approach worked ranged from retail and large enterprise to health care and banking. Here are some specific examples highlighted:
- A retailer complained that customers were taking pricing stickers for self-checkout off inexpensive items and putting them on expensive items, so they were taking out watches but paying for bananas. The solution was to introduce image recognition and matching so that cameras would flag a disparity between what the POS (Point of Sale) system was reporting and what was being scanned, significantly reducing fraud.
- A casino was able to increase sales by better segmenting customers, and thus better-formulating offers based on the unique needs of defined customer groups. This example is one of the most common successful uses of AI, but it is very difficult to deploy making the improvements that Dell EMC is making critical to this positive outcome.
- In health care, one of the problems regarding replacement surgery using prosthetics is poor matching between the manufactured item and the person that will get it. In this case, the Dell EMC solution provided much better matching between hip replacement parts and patents, increasing the percentage of positive operation outcomes significantly. (I suddenly have something I want to ask about if I ever must replace any body parts).
Other solutions included improvements in fraud protection for credit cards (MasterCard), fraud detection on insurance claims and better analysis of risk for mortgages and other loans.
There were a couple of interesting solutions that caught my eye as well. For instance, Caterpillar is using the technology for autonomous driving and mine safety, which, given some of the physical tools used in that industry are bigger than houses, is critical to avoid a catastrophic outcome. Simon Fraser University is using the solution to analyze DNA and microbes to stop pandemics. MIT Lincoln Labs is using the technology with one petaflop of data to advance robotic vehicle viability, and Zenuity can use this tech to run 50 simulations an hour to advance autonomous driving.
Finally, AeroFarms has used the technology to increase farm production 390X over more traditional methods; this could be critical to our survival as global warming destroys existing farm resources.
Wrapping Up
What Dell EMC and the Cutter Consortium reminded me is that with any new technology, there are limitations that massively offset the hype early on. Understanding these limitations, however, can massively increase the effectiveness of the result and not only ensure that expectations are reasonably set but that those expectations are themselves met.
AI can do some amazing things, but only if it deployed properly, IT and the data scientists’ needs met and the solution crafted against the specific problem that drove the purchase decision. Fortunately, that is what Dell EMC is advocating and doing, and its example should help move the industry and AI solutions forward.
Rob Enderle is a principal at Enderle Group. He is a nationally recognized analyst and a longtime contributor to QuinStreet publications and Pund-IT.