Source – https://www.forbes.com/
Chief Technology Officer at Unit4, overseeing development of intelligent software for service organizations.
There’s a lot of hype around machine learning, but what does it really mean in the context of enterprise software? How does it work, where is it adding business value today, and what should we expect from it in the future?
Let’s start with some definitions. Artificial intelligence (AI) is an umbrella term that includes machine learning (ML), deep learning and cognitive learning. The part most relevant to enterprise software is ML, which in this context is the ability to create automation through AI algorithms.
A lot of what ML does is really just statistical analysis: crunching numbers, measuring parameters, identifying patterns and projecting future outcomes based on past results. You don’t actually need fancy ML algorithms to do this; you can do it with standard logical programming.
The degree to which the ML itself improves business outcomes is currently marginal. The accuracy of a financial forecast, for example, is sensitive to far greater factors than whether the algorithm can refine itself slightly over time. If you haven’t got harmonized, accurate and complete data to start with, simply applying ML to it isn’t in itself going to result in better business decisions.
Realizing the Growth Potential of AIArtificial Intelligence Is Learning To Categorize And Talk About Art
A Solution Looking For A Problem?
In terms of Gartner’s hype cycle, ML is currently at the peak of inflated expectations. You cannot simply throw ML at a bucket of big data and expect it to magically come up with a perfect business plan.
As so often in business, you shouldn’t start with the technology itself. Before you think about where to apply ML, you need to step back and ask: What is it we’re trying to achieve?
Look for points in your business processes where some sort of judgment or prediction is required and where any small improvement in accuracy would have a disproportionate benefit to the business. These are the potential use cases for ML. Otherwise, ML is at risk of becoming a solution looking for a problem.
For example, if you apply ML instead of conventional statistics — and you have good underlying data — you should be able to continuously enhance the accuracy of the predictions to improve, say, operational efficiency and customer experience.
Where Is ML Adding Value Today?
ML is currently being used to good effect in enterprise software to automate routine business processes.
Receipt recognition: In this use case, an ML algorithm examines a scan of a receipt and deduces what type of receipt it is, then automatically matches it against an expense record in the ledger.
Smart invoice processing: Here, the ML algorithm examines a scanned paper invoice or electronic invoice and identifies the key elements: invoice number, customer number, amount, payment terms and line items, then matches them against the relevant purchase order or delivery note.
Time sheet completion: Typically, there are around five dimensions to completing a time sheet — project, task, level of resource, type of work and time spent — all of which, until now, had to be input manually. An ML algorithm can auto-populate them based on previous patterns. This can free up a lot of time for people and make work easier for them.
The Human Intelligence In AI
A great deal of human intelligence is required to get AI to work. To get predictable, reliable results, you have to decide the use case and make sure the data itself is of a high enough quality before setting the algorithm a task. Then you have to train it.
In its simplest form, training an algorithm involves a person checking the results and providing feedback on their accuracy. This is called supervised learning.
The human mind is by far the best pattern-matching machine in the universe. The average 2-year-old can probably correctly identify a cat after it’s seen two or three, while an ML algorithm might need to see 2,000 before it can be sure. But, once trained, ML excels at dealing with huge volumes of data and processing it very quickly, while never getting bored performing repetitive, tedious tasks day in, day out.
What Can We Expect Next?
Based on my experience, typically less than 20% of business processes are automated in enterprise software. I believe that in as little as two to three years we could see up to 80% of routine business processes automated by ML.
The current frontier is about how we interact with software, and there’s an ongoing paradigm shift around user experience. As in the book, The Best Interface is No Interface, which calls for an end to “screen-based solutions,” software should be able to recognize human speech through natural language recognition; ML is now making this a reality.
The next big leap forward will be to eliminate humans from some business processes entirely. In a workflow, at the point where a person has to approve something, an ML algorithm will examine an approver’s behavior and learn what falls within the usual tolerances. By copying the person’s judgments, ML can carry out the work itself and simply inform the user when it’s done.
Focus On What Matters
The promise of ML in enterprise software is to make it pervasive but not visible. Will it put us out of our jobs? No, but it will allow us to offload the hum-drum, low-value tasks and focus on what really matters — adding value to the business.