By: Phil Wainewright
t’s obligatory these days for enterprise application vendors to add artificial intelligence and machine learning to their products. For many, the first step is to select an evocative name from the history of computer science — Salesforce has Einstein, SAP has Leonardo, Infor has Coleman. In contrast, Oracle takes a more utilitarian approach. Although its NetSuite business unit made a fanfare of adding AI to its cloud ERP platform yesterday at this week’s SuiteWorld conference, the focus is on practical outcomes, as EVP of Development Evan Goldberg explains:
We’re not boiling the ocean, we’re not trying to do all the general purpose AI …
We’re trying to find the problems that we think we’re going to be able to solve quickly with a high degree of accuracy and provide great value to SMBs.
Applying machine intelligence to midmarket ERP
Goldberg ended his product keynote today with a 15-minute rundown of how NetSuite is applying machine learning technology to its midmarket cloud ERP platform, across functions such as ecommerce, supply chain, analytics and finance. It draws on Oracle’s machine learning services and applies the AI algorithms to the rich datasets in each company’s NetSuite instance across ERP, e-commerce, HR and CRM, as well as anonymized data from all the different industries in the NetSuite customer ecosystem and the broader customer experience data held in the Oracle Data Cloud.
The intelligence is applied to NetSuite in three main ways, he says:
- Insights — predicting what might happen based on correlations found in the data, such as which customers are most likely to pay invoices late, or which projects are most likely to go over budget.
- Intelligent automation — learning a task that users do repeatedly and doing it for them, only presenting the exceptions for their attention.
- Intelligent interaction — make the time that users spend interacting with NetSuite as effective and productive as possible, for example by suggesting useful content to add to their dashboard or pre-selecting the most likely responses in forms.
Goldberg went on to demonstrate several AI-enabled functions that are set to become available in the coming months.
Bringing searchandizing to e-commerce
In e-commerce, ‘searchandizing’ is the use of artificial intelligence to tune search results to perform a merchandizing role. Rather than simply returning keyword matches, this learns to present the results most likely to result in a sale, based on past performance recorded in the NetSuite system. Or it improves relevance by automatically mapping keywords to related terms that other shoppers have used in the past.
One of the real advantages of these machine intelligence systems is that they keep on learning, he explains:
Together all these signals can be put together to weight search results, displaying results to the shopper so that they’re most likely to buy the [product] that’s best for you. But the story doesn’t stop there, because that’s where we start. Everything this shopper does becomes a signal back into the system.
The system uses this feedback loop to learn and improve for the next search by the next shopper. And these machines, they never get tired. We can take great advantage of that to improve your site search, 24/7, 365.
More examples of the intelligent suite
A second example comes from supply chain management, where the intelligent system can analyze past performance and calculate a risk factor for a given delivery. Goldberg showed a supply chain control tower dashboard, where a crucial delivery is flagged as having a high risk of being late based on recent shipments. The system can recommend an alternative supplier for the same item with a much higher predicted confidence in on-time delivery, based on a better track record.
Intelligent form interaction is the third example. This is currently in prototype, said Goldberg, but the principle is that a sidebar next to the form shows the fields the user is most likely to want to fill in, based on their role and other contextual information. Then when they go to fill in certain fields, the system suggests a preferred answer and provides an explanation why this was suggested. The user can accept the suggestion or input another variable, and once again the system will learn from each interaction.
The final example showed several different capabilities coming together in the NetSuite dashboard, such as suggested content, shortcuts, and alerts. For example, a project management dashboard might give key metrics for ongoing projects and could surface potential problems, such as a project that’s heading for a budget overrun based on comparisons to past projects.
Rather than personifying its machine learning ambitions with a cute name, NetSuite has gone directly to showing off no-name capabilities with real-world relevance. That’s good news for NetSuite’s midmarket customer base, because these businesses are only interested in the latest technology if they can see it solving pain points and delivering value for them.