BOSTON, Dec. 17, 2018 (GLOBE NEWSWIRE) -- As more and more enterprises experiment with AI to automate back-office operations, an interesting paradox is developing regarding the accuracy of AI and its relative value. Indico, a provider of enterprise AI solutions for intelligent process automation, has observed some common and costly misconceptions in many enterprises pursuing AI initiatives.
As enterprises look for new ways to automate existing processes with AI, their goals tend to focus on either streamlining existing workflows to free up staff to focus on higher-value activities, or analyzing large volumes of existing enterprise content to gain new insights about market trends, sales opportunities, customer sentiment, etc.
In both cases, users are typically targeting processes that are highly manual – where humans are making decisions based on their subject-matter expertise and any additional information they are able to consume. Whether reviewing market research data, analyzing contracts, or responding to RFPs, the accuracy and consistency of these manual decisions is never objectively measured. Instead there is a heavy reliance on subjective judgement and gut instinct.
When companies introduce AI to these processes, a lot of attention and rigor is focused on the accuracy and consistency of the decisions driven by machine learning models. Users can see immediately how well these models perform against the target outcome as the model is trained, and very often it’s the first time anyone has actually tried to measure the efficacy of the given process. Yet, it’s these detailed measurements that make many users pause. Given the choice to go with their gut sense vs. a machine learning model that performs at 95% accuracy, users often opt to stick with the manual process because the machine learning model is ‘not accurate enough’ in their view.
“Many users simply have unrealistic expectations for AI,” said Tom Wilde, CEO of Indico. “They think it’s a magic solution to their data problem and are disappointed when it is not 100% perfect. When this initial disappointment exceeds their enthusiasm, projects tend to lose momentum quickly.”
In Indico’s experience, this problem is exacerbated because many users are not focused enough on a business outcome. Users look at the accuracy of their AI project in a vacuum vs. how much more accurate it is, or how much time it will save, compared to what they are currently doing manually. But, according to Wilde, this misconception can also be a good forcing mechanism for people to align their understanding of the problem at hand.
“Often times the problem they think they are solving and the problem they ultimately solve end up being quite different,” said Wilde. “Successful AI projects require a careful defining of inputs and outputs. If either are poorly defined, the results tend to be disappointing.”
Another challenge observed by Indico is the idea of AI accuracy itself. There isn’t a single metric whereby the efficacy of a machine learning algorithm is measured. Accuracy is a uniquely poor metric of an algorithm’s efficacy. In its work with enterprise customers, Indico has observed that more successful projects calculate their solutions by its impact on the problem being solved; e.g., how much true efficiency can be gained by applying AI to the specific business process in question.
Indico suggests users approach their AI projects as follows to create a better environment for success:
- Treat AI like other enterprise technology projects. They require business context to solve business problems and produce business results.
- Identify a business outcome up front. Keep the focus on how AI can truly help advance the desired business goal.
- Involve business Subject Matter Experts (SMEs) more, and Innovation Labs less. At its most useful, AI is augmenting what SMEs are already doing – by making them more productive and making their gut instinct smarter and more accurate.
- Measure the efficacy of existing processes. Understand what a reasonable target for the machine learning algorithm’s accuracy is. If your existing process is 90% accurate, 85% accuracy could be a reasonable target given the significant efficiencies to be gained. But if an existing process is only 20% accurate, expecting 85% accuracy from AI is unreasonable.
- Ensure model ‘explain-ability.’ Don’t settle for a black box. Users need to understand the efficacy of their models in a very detailed way; e.g., what kinds of errors the model is likely to experience -- and be able to explain the ‘thought process’ behind them.
“By taking into account some of these common misconceptions about AI and its accuracy, and taking steps to address them up front, organizations can turn their attention to the real value and productivity improvements AI can deliver and start to put it to work for their benefit much more quickly,” added Wilde.
Indico is a provider of Enterprise AI solutions for intelligent process automation. Our focus is on helping to automate tedious back-office tasks, improving the efficiency of labor-intensive document-based workflows, and extracting valuable insights from unstructured content, including text and images. Our breakthrough in solving these challenges is an approach known as transfer learning, which allows us to train machine learning models with orders of magnitude less data than required by traditional content analysis techniques. With Indico, enterprises are now able to benefit from the dramatic advantages of machine learning in a fraction of the time. For more information, visit. https://indico.io/.