Artificial intelligence and machine learning have been shaking up many areas of business, from cybersecurity to market analytics, bots and self-driving cars.
But when it comes to core corporate functions, especially those where the risks of making bad decisions are substantial, the use of artificial intelligence is still in its early stages.
Take, for example, AmerisourceBergen, a drug wholesale company based in Conshohocken, Penn. The company has 19,000 employees in 47 countries, and gross revenues of $147 billion a year, good for 11th on the Fortune 500.
Alexander Kugler, the company’s VP of pricing, is well aware of the potential of artificial intelligence to help the company make better decisions when it sets prices for its products. Set prices too high, and customers will go elsewhere. Set prices too low, and the company will lose money.
Previously, the company used spreadsheets to pull in data from various systems to determine production costs, and used past history and their own general knowledge to try to figure out how sensitive customers were to price changes and what the competition was doing.
“It was an archaic pricing methodology that hasn’t kept up with industry trends and dynamics,” says Kugler.
So, 15 months ago, AmerisourceBergen began the move to an integrated system that automatically calculates production costs, analyzes historical transaction data, and pulls in outside data such as weather forecasts to create a foundational layer for future deployment of artificial intelligence.
This anticipation of the need for AI to improve ERP functionality as part of business transformations is growing among first movers, and ERP vendors are weaving machine learning functionality into their offerings to meet the coming demand.
The future of ERP
The platform AmerisourceBergen picked, Vendavo, has intelligent functions built in, but so far, the company hasn’t put those to use.
Instead, the company is using built-in expert algorithms such as alerts for when a price is set below cost. These were created based on the work of data scientists, but are not generated on the fly by machine learning systems.
“We just went live with the system in September, and we’re using it more in terms of executing the black-and-white business rules right now,” he says. “We’re crawling — and we’re about to start walking.”
Calculating whether a price has been set below production costs may be a complicated calculation, but is still just a calculation. Once there’s a formula, and the right data is available, getting a result out is just arithmetic.
But there are other possible alerts, which require judgment rather than simple calculations. For example, a particular series of weather events could trigger more demand for, say, flu vaccines. Or a new competitor may be about to enter a particular market segment, driving prices down.
“Looking forward, I absolutely see the value in business-risk alerts,” says Kugler. “We have tens of thousands of customers, and sell tens of thousands of products. Having a process or framework in place that can alert us of potential issues before they occur would be phenomenal.”
Another use of machine learning is in workflow automation.
The first step is to have a framework where a person can make a decision that automatically triggers a series of actions. Then, the next time a decision needs to be made, the system could suggest a course of actions based on past experience. Finally, once there is sufficient confidence in the recommendations, the system could take action automatically, with human beings just overseeing the process and dealing with exceptions.
“You would get consistent decision making, and enable the pricing team to focus more on the art behind the pricing,” says Kugler. “It would free up the pricing team to focus on real opportunities that truly bring value to the organization.”
Today, for example, a pricing team member might spend three hours on more complex price analysis, and five hours on the more routine tasks involved with price administration. With smart automation, they might be able to spend just one hour on price administration, and the other seven hours on value-added activities.
“This is not a situation where I’m looking to reduce headcount,” he says.
But all that is in the future.
Right now, AmerisourceBergen is still in the process of putting the basis in place, so that it can then move on to figuring out how to use machine learning for business risk alerts, predictive analytics and workflow automation. No decisions have been made yet, Kugler adds.
But making better predictions, setting better prices and even lowering production costs isn’t the only benefit of having smarter systems.
AI technology will also help the company defend itself — and not just against its current direct competitors.
“We’re acutely aware of things like Amazon entering our space,” he says. “We would be fools to think that Amazon isn’t going to leverage everything it can to try to muscle in.”
Those companies that don’t take advantage of AI will suffer, he says. “I think you’re going to see more erosion of their profits and ultimately they’re either going to have to diversify and find other revenue streams, or become candidates to be taken over.”
Getting ready for AI
Like AmerisourceBergen, many companies are still in the process of getting ready for smart ERP systems that use machine learning and advanced analytics, intelligent interfaces, and workflow automation.
AI-based ERP products and features have only recently hit the market and companies are taking a slow and careful approach to adoption, says Josh Sutton, head of AI at New York-based SapientRazorfish, which provides consulting services to help companies add AI-powered capabilities to their ERP systems.
“AI is a very significant part of business transformation today,” he says, “and the rate of transformation is faster than it has ever been.”
Sutton is seeing a lot of companies starting narrowly-targeted pilot projects related to adding AI to ERP, but it’s still very early.
“We’re really at the beginning of this path,” he says. “And the companies that are having the most success are those that are taking the small, bite-sized pieces rather than trying to boil the ocean. These companies have real results, and sooner rather than later.”
Another company that can see the potential of artificial intelligence and machine learning to improve operations is Home Depot.
Today, the company has data scientists on staff analyzing sales, weather trends and other data to help anticipate customer needs.
Take hurricanes, for example.
“We’re able to quickly respond and stage emergency supplies like water, plywood and generators at strategically-located distribution centers,” says Paul Mayer, Home Depot’s manager of communications. “We definitely are the hurricane headquarters.”
Today, this is all done with human brain power. But the company is evaluating and testing the use of artificial intelligence and machine learning, especially when it comes to supply chain and inventory management.
“Our goal is to make sure we have the right products at the right time when customers need them,” he says.
According to a survey released this month by analytics company LevaData, 69 percent of companies are extremely interested in how AI can help improve their supply chain.
So what’s holding up the move to smarter ERP? There are several factors at play, including cultural issues, the slower pace of cloud adoption for ERP compared to other aspects of business, and the fact that the technology is only just now emerging.
ERP’s people problem
According to the LevaData survey, 49 percent of respondents said that their in-house talent is not yet ready for radical digital transformation of these core business processes.
Take, for example, the procurement process. Large companies may be dealing with thousands of different products, from thousands of suppliers, but the people in charge are generally still using Excel spreadsheets and gut feelings when negotiating prices, says Rajesh Kalidindi, founder and CEO of LevaData.
They want to pound on the table, and use their years of business experience and well-honed instincts, he says.
Now they need to move to a data-driven approach.
“And people think, ‘It’s going to take away my job,'” he says. “‘Are they going to rely on the decision of the machine, or mine?'”
According to a survey conducted this summer by SAS, the cultural challenge is the biggest barrier to AI adoption, with 49 percent of respondents saying that they had a lack of trust in the technology.
In addition to being reluctant to take a machine’s advice, employees may also be reluctant to turn over the information that the machine needs to make better decisions. In particular, when it comes to, say, price negotiations, the AI needs to know not just the final results, but also what strategies didn’t work.
“The challenge is getting all the data, not just the data that people want to be managed by,” says SapientRazorfish’s Sutton. “This is a behavioral challenge that companies are struggling with. AI works better with all the data, not just the data that people want to share. Many times, people will only put things in the system that reflect well on them.”
Cloud-based and SaaS deployments make it easy for ERP vendors to implement the latest technologies and integrate data feeds and analytics tools from outside partners.
However, 67 percent of companies still use on-premises ERP, while only 33 percent use cloud or use cloud-based vendors, according to a report released earlier this year by Panorama Consulting.
By comparison, according to IBM, 87 percent of CRM systems are now cloud-powered.
“Most companies have stayed on top of the pace and technology enhancements on the front office but are only beginning to look at their back-office applications,” says Mickey North Rizza, an analyst at IDC.
Companies that do use cloud-based ERP or SaaS products have a head start when it comes to artificial intelligence, she says.
“Unfortunately many large enterprises still have traditional on-premises ERP systems and have not moved to the cloud yet,” she says. “These large enterprises are missing out on the innovation.”
Rizza expects more companies to move operations to the cloud to help drive business transformation, and to do more with machine learning.
ERP vendors rolling out AI capabilities
Every major ERP vendor either has artificial intelligence on their near-term road map, or is already rolling out features and tools.
For example, in October, Oracle announced several AI-powered add-ons for its cloud-based ERP products.
More products are coming soon, says Steve Cox, group vice president for Oracle ERP and EPM product marketing.
So far, companies are still at the very beginning of adoption, he says, though some organizations are already using AI on the analytics side.
“The NHS uses artificial intelligence to spot fraudulent claims in Britain,” he says. “That’s a good example of a use case.”
But AI can do more.
“Imagine that you get a weather warning indicating that one of your factories won’t be able to make its production quota on a certain day, and you’re going to find yourself with major problems with customers,” he says. “Based on what happened previously, it suggests to you that there are six possible solutions, and it shows you the financial implications of each solution. When you select the one you want, it shows the steps you need to take. But the most important thing is that next time something like this happens, the system remembers.”
That’s the future, Cox says, and that’s why he expects AI to be as transformative for business as cloud computing.
“In my view, artificial intelligence and machine learning will change everything,” he adds.
Cox predicts that 2018 will be the year when AI will start seeing major adoption in the ERP space.
“I think the interest is huge,” says Paul Farrell, VP of product marketing for Oracle NetSuite. “There are some businesses utilizing it right now, but looking at our base of customers, everyone is interested, but they’re waiting on the practical applications.”
Another ERP vendor with an artificial intelligence product is Infor, which announced its Coleman AI bot this past summer. The bot uses the Amazon Lex deep learning and natural language interface.
The product is currently in beta, and will go live next spring, says Rick Rider, senior product director for Infor Coleman AI.
It’s a fantastic set of features, says Billy Blackerby, solution architect for supply chain and merchandising operations at Whole Foods.
“You are bringing consumer-grade functionality into a business setting,” he says.
There are also predictive analytics and machine learning capabilities being developed that help the system predict what the user wants.
“A particular use case that comes to mind for me is in the realm of assortment planning,” says Blackerby. “‘Coleman, I need a product that can fit on a 3-foot-wide shelf in the X category that has parity with other dried goods to be carried in the Y holiday season.’”
“Large corporations want to ease into this,” says Mitchell Lee, profit evangelist at Vendavo, which sells pricing analytics tools to large global companies in the B2B space, and began offering machine learning tools in 2016.
“Most of those tend to be very conservative in any endeavor,” he says.
In particular, they’re hesitant about “black box” AI systems, where the reason behind a recommendation isn’t clear.
“It’s important to business leaders to be look investors in the eye and say, ‘I understand the process by which we are making these decisions,'” he says.
For example, Vendavo can automatically sort customers into market segments, but customers need to be able to see why the system has created the clusters it did, and why a particular customer is in one cluster and not another.
“You might know something that’s not in the data, your business knowledge, your general knowledge that isn’t incorporated in the system,” he says.
About 10 percent of customers are already using the machine learning technology to automatically identify market segments and calculate pricing power, Lee says, and those recommendations are then reviewed by humans.
“It’s monitored over time to see if the recommendations are working, or if the new pricing is either not winning the business, but winning business but at a lower price,” he says.
“But corporations are hesitant to flip the switch and have it go automatic,” he adds. “People are used to making these decisions themselves, and the consequences of a machine-produced error, the consequences of some of those recommendations could be consequential.”
“We are looking to applying machine learning to many domain-specific areas in ERP,” says Ajoy Krishnamoorthy, vice president for platform strategy at Acumatica, a cloud ERP vendor.
For example, users will be able to ask, “Alexa, ask Acumatica how many laptops I have in stock,” he says.
There are companies already piloting some of the new features, but they aren’t in production yet.
In the case of the Alexa integration, the company is close to rolling it out, but security concerns remain. For example, you don’t want random people asking for and getting company data.
“We need the voice authentication piece done,” he says, “and we’ll have that soon.”
Another company with a cloud-based ERP product is VAI, which serves mostly midsize companies. It started working on AI about a year ago.
The company has the IBM Cognos line of business intelligence products built into its applications, and is also integrating with IBM’s Watson artificial intelligence platform.
“A lot of our customers in the traditional pillars of industry are looking at what AI can do for them,” says Kevin Beasley, CIO of VAI. “In the future, that is going to expand as we develop more and more AI capability.”
“We’re just getting started,” says Aaron Harris, SVP and head of engineering and technology at Sage Intacct. “This is all pretty new stuff. We’re just building out the underlying technology, but we’re not quite ready for customers to start using this yet.”
Sage Intacct plans to completely eliminate the close, so that corporate books are always up-to-date, and problems are spotted and addressed immediately, instead of at the end of the quarter. Plus, instead of creating reports, users will simply ask a natural-language question, and the platform will pull in data not just from the financial system, but from a multiple sources.
“We’re getting a lot of excitement from customers,” Harris says.
Nintex, a vendor that focuses on workflow automation, is working on adding machine learning and natural language processing to help its customers move away from rules to smarter, more flexible workflows.
The technology is in the testing stages, says Matt Fleckenstein, CMO at Nintex, and will be launched in early 2018.
“We’ve got a bunch of customers in the advanced preview stage right now,” he says.
Some companies, for example, have more than 100,000 different workflows, he says. Intelligence can suggest actions to employees, or even automatically perform some of the actions.
“First, it’s saying, ‘You always tend to approve contracts below a certain amount from a certain person. Do you want to approve these?'” he says.
Then, in two or three years, once companies have gained confidence in the recommendations, the system can skip the recommendation step and just go ahead and take the action.
“As I build more trust over time, and actually see the value of it, and see that there’s limited downside to it, I’ll give it more power,” he says. “It’s not all that different than if you had a new employee who joined your team — as you get your confident in them you give them more responsibilities.”
Learning by doing
To get started with AI and machine learning, doing is the best first step, says Helio Mosquim, IT innovation manager at Brasil’s Vale, one of the world’s largest mining companies.
Vale has been experimenting with machine learning using SAP’s Leonardo platform by building prototype AI-powered services.
For example, employees trying to order replacement parts currently have to go through supplier catalogs, find part numbers, then enter those numbers into the system.
“It’s a complex process, with a lot of mistakes,” Mosquim said in a conference presentation last month.
The company considered using voice recognition, but that turned out not to work in practice. “The equipment is out there in the maintenance area, where it is so loud, so noisy,” he says. So Vale decided to go with image recognition, and used SAP Leonardo’s machine learning capabilities to learn to identify parts by sight.
“Now a guy in the field can take a picture with an iPad and create the request right out there in the field,” he says.
The potential for AI in ERP is incredible, says Patrick Bakey, president of SAP Industries.
“In the next few years, repetitive, boring tasks that can be automated, will be automated, which will increase productivity and allow companies to reallocate jobs and create new roles,” he says. “Companies will be able to dedicate more talent for strategic and creative projects.”
Plus, employees will have a much easier time interfacing with enterprise technology.
“Today, you are using an AI-powered bot in your home like Alexa or Siri to look up pizza delivery restaurants nearby, with recommendations, reviews and coupons,” he says. “We will bring the same level of convenience and intelligence to enterprise. applications.”
Related ERP articles: