AI is arriving for your business
Yes, artificial intelligence is a buzzword, but dismiss it at your peril: it’s already becoming a useful tool for businesses, automating small tasks that eat up employee time
AI is automating small tasks that eat up time.
Artificial intelligence isn’t limited to cutting-edge healthcare and driverless cars – it can power your business, too.
There’s more to AI, be it machinelearning or neural networks, than chatbots of questionable use. Instead, the software you already use is slowly letting AI-powered tools and features trickle into their systems, be it Microsoft, Box or Slack. Businesses know this: a report from Boston Consulting Group and MIT at the end of last year revealed that eight in ten respondents said AI will help their company gain a competitive advantage and is a strategic priority for their businesses. An EY survey suggested one in three businesses are already piloting AI tools, with the aim of improving or developing new products or services, cutting costs and accelerating decision-making.
So what is this AI technology doing? Plotting a future war against humanity? Coming over here to take our jobs? The reality is rather more mundane: it’s often automating basic tasks, says Angela Eager, research director at TechMarketView. “It’s tempting to look for the most headline-grabbing developments but some of the most useful AI – and specifically machine learning – uses within the business arena are more down to earth,” she said.
Eager points to invoice processing and filing expenses, such as Abacus and CumulusPro, along with service desks and customer services where automated tools can handily sift through content. Sales departments
are already using machine learning to predict and forecast fraud detection, via AI-powered services such as Kount. And HR and recruitment are being helped with intelligent matching that far surpasses keyword association – Textio even examines ads for gender bias.
Behind the scenes
“In the business environment, many of the use cases are almost behind the scenes in that only the output is seen by the user, be that a recommendation for action, a decision point or a completed expenses claim – and that’s precisely how it should be, now and in the future,” said Eager. “Algorithms don’t help users do their jobs but the output does. These technologies are increasingly being embedded within applications and some sort of capability is rapidly becoming a baseline requirement – whether the business has immediate plans to deploy it or not.”
Such technologies aren’t going to replace office workers, or at least not anytime soon. Instead they will take over or assist with repetitive tasks, which will make them welcomed by staff tired of such dull jobs. “Today’s business use cases are also driven by levels of capability – machine learning is at home with repeatable patterns, narrowly defined tasks and large amounts of data, but is not capable of general purpose machine learning,” she said. “‘Comfortable’ use cases – ones that help rather than replace the user such as invoice processing – are finding traction.”
For that reason, chatbots aren’t necessarily a silly piece of technology – they could help your business make contact with customers and gather data on what they’re looking for, as well as help staff run queries and research. “Voice-based chatbots in an office or a noisy environment are unlikely to be appropriate but can be well suited to front line customer service triage,” Eager said. “Textbased chatbots potentially have more widespread potential within business – it’s about the right technology in the right environment.”
Growing complexity
So far, then, the use of AI in business is limited to chatbots, automating expenses and other simple tasks. However, as the Microsoft and other announcements reveal ( see “Three ways you can use AI at your company now”, below), there are more complex jobs that such technologies can help with, in particular those that help sift through data quickly, such as for sentiment analysis, data discovery and image recognition and transcription, as in the OneDrive features. Such technologies will be embedded into existing software, as with Microsoft and Box, but it will also be available as machine-learning-asa-service for more specific use cases, Eager said.
While AI is becoming easier to use and addressing more in-office tasks, challenges remain. No machinelearning system works without quality data – as the IT adage goes: garbage in, garbage out. So none of us should necessarily trust AI output – instead, it must be auditable.
“Businesses need to have a specific and well-defined task in mind, a known output and ways of measuring the impact,” said Eager. “They need to be able to select the ‘right’ data for the task – which assumes they know what the ‘right’ data is. They also need to know what a ‘good’ result looks like and be ready to question output to prove its validity and build trust. Beyond that, the issue of auditable AI/machine learning is becoming more important – understanding the how and why behind a recommendation, particularly in highly regulated industries – and that’s not easy to determine.” In other words, AI can work for business, but let’s take the effort to make sure it works.
“The tech will take over or assist with repetitive tasks, which will mean it’s welcomed by staff tired of such dull jobs”