Will AI Replace ERP?
June 19, 2017 Alex Woodie
Programmers in the midrange space are used to coding IBM i applications, like ERP, supply chain, and retail management systems, using RPG, Java, and COBOL. It’s simply what they do. But what happens when machines program themselves and react to events in real time? One impact, according to prominent members of the software industry, is the end of the enterprise software business as we know it.
Led by Web giants like Google and Facebook, the computing industry is making rapid advances in the field of artificial intelligence (AI). The advent of deep learning techniques, such as convolutional neural networks, when combined with a huge corpus of training data (i.e. cat pictures) has pushed the accuracy of some digital sensing systems, such as image detection and speech recognition, very close to what humans can achieve.
In some circumstances, the AI-based systems are already better than humans. It has already been 20 years since IBM‘s Deep Blue beat Russian chess grandmaster Garry Kasparov, but that was a relatively easy challenge. More recently, Google’s DeepMind team bested humans in the ancient Chinese game of Go, which requires more strategic thinking than chess. And this spring, a deep learning system powered by a supercomputer at Carnegie Mellon won a five-day Texas Hold ‘Em poker tournament filled with some of the best players in the world, bringing home $290,000 for its masters.
Sure, AI can be great fun and games. Who can forget what happened when Microsoft released its AI-powered chatbot, Tay, into the Internet wilds last year? Thanks to a handful of Twitter users who exploited her “repeat after me” feature, Tay was soon cussing like a sailor, which forced Microsoft to end the experiment after just 16 hours. LOLs all around.
But here’s the thing: AI is not all fun and games. There is real power being generated in the field. And it’s not all confined to the Web. AI is venturing out into the real world.
One man at the cutting edge of this trend Tom Siebel. The founder of Siebel Systems, the CRM software company that Oracle acquired for $5.8 billion in 2005, is several years into his latest venture, dubbed C3 IoT, which applies machine learning techniques to big data gathered across the Internet of Things to help its clients squeeze more efficiency out of their operations. Currently, most of C3 IoT’s customers are public utilities, but the company is expanding into other markets.
What Siebel told Datanami in a recent Q&A sees should give other IT leaders pause:
“I’ve been in the information technology business now for four decades. I’ve lived through minicomputers and mainframe computers and personal computing and enterprise computing and relational database technology, the Internet, the cloud,” he says. “When I think of IoT and predictive analytics and big data, it looks to us like an entire replacement market for everything that’s taken place in enterprise application software.”
The databases don’t go away, the 64-year-old billionaire says. “But with the CRM example, absolutely, it’s a complete replacement market for CRM. The next generation of CRM is all about the device that you have in your front left pocket or maybe your hand . . . . Masses of data that we can collect and we can provide to use to provide value-added services to people in terms of next-best product or next-best offer.”
“That’s what CRM looks like in the future,” he continues. “I know something about CRM. Candidly, I invented it. And so the next generation of CRM is CRM meets AI, so we’re going to have much more precise revenue forecast, product forecasting, much more persistence and value added customer service capability, both automated and also providing human assistance through customer service personnel. So I think it does replace the CRM market but it doesn’t replace the database market. It leverages it. It replaces the ERP market, it replaces the supply chain software market, and it replaces manufacturing automation, absolutely 100 percent replacement.”
Siebel isn’t saying we’ll just throw all our database programming logic, triggers, procedures, and built-in functions into the ocean. He’s not saying that the trillions of lines of RPG, COBOL, Java, and PHP code will be replaced, line for line, by machine learning algorithms written in R or Python or Scala.
Siebel understands that this programming logic has been honed over the decades to support the actual business processes that companies use to run their daily businesses. There’s not much point in ditching a proven order-to-cash processes in favor of machine learning algorithms. The comparison doesn’t even make sense.
The larger point that Siebel is making is that the instrumentation of the world through sensors – and the potentially valuable insights that we can glean from the data collected by those sensors – will dramatically shift what we can do with computers, and frankly what we need computers to do for us.
It’s something that Dan Magid has been advocating as part of Rocket Software‘s digital transformation roadshow this spring. “Every part of the organization is going to have some kind of software component,” Magid says. “It’s not going to be just ‘Geez we run our accounting on the computer system.’ It’s how we interact with customer and how we work with our employees. It changes the way employees work.”
The days of employees sitting behind a 5250 green screen typing on the keyboard are numbered, according to Lane Nelson, president of HarrisData, the Wisconsin developer of ERP software for IBM i and other platforms.
“The more you automate and imitate human decisions, the more time you save,” Nelson tells IT Jungle. “And the fewer back-office people there are going to be doing repetitive jobs, like entering invoices into the system. It’s not that we want to put our customers out of work. But information robots are going to do to the back office what robots did to the plant.”
Instead of trying to replicate in digital code the real-world business processes that companies follow – and then spend years coding rules to hone those processes to the umpteenth degree to claim that last percentage of profit margin that will get you that gold watch – Siebel says why not just use sensors to plug the real world into the digital world, and then use machine learning to optimize the behavior? It’s a radically simple concept, and the interesting part is that it really hasn’t been done. At least not in any widespread manner, anyway. We’re in the process of instrumenting the world now, which is fundamentally what’s driving all the hype over big data.
We probably won’t get rid of the low-level code that describes the order-to-cash process, or many manufacturing processes. But thanks to the proliferation of sensors up and down the supply chain (or the “value chain” or whatever chain links customers and partners and suppliers), the way humans interact with the software is bound to change dramatically.
“We’re able to solve classes of problems that were previously unsolvable,” Siebel continued. “We’re seeing this sensoring of the value chain that is I think a very significant economic and social phenomenon. At the beginning of this century, there might have been half a billion sensors out there. Today there 19 billion and in five years there will be 50 billion. This is across every industry – healthcare, travel, transportation, financial services, discreet manufacturing, automotive, aerospace – you name it. Value chains are all becoming sensored. I don’t think there’s any hype to it at all. It’s a fact. If you look at the rate of IoT adoption across these value chains, it’s exponential growth.”
All this new data and the new business processes that emerge from it will likely be enough to disintegrate the ERP industry as we know it. It will start with more widespread use of machine learning algorithms to automate low-hanging fruit – such as optimizing widget prices to seasonality, detecting fields to automatically enter invoices, or detecting fraud.
But increasingly, thanks to the growing instrumentation through sensors, the new data that will introduce, and the new ways that we’ll interact with computers and each other, many of the existing business processes that we now know as ERP will likely evolve into something much different. It will take time, of course, but eventually AI will replace ERP. It has to.
Q&A with C3 IoT’s Tom Siebel (Datanami)