Infor has long invested in data science, with its dedicated Dynamic Science Lab team at MIT, but with the launch of Coleman, the executive team is claiming that it will be looking to ‘rethink ERP’ where it can - taking advantage of ION (Infor’s API/middleware, ensuring data flows through all apps to Coleman), machine learning frameworks, and by more closely integrating the apps with GT Nexus (Infor’s commerce network).
It appears that Infor isn’t attempting to build out its own AI platform, but is rather using these building blocks to tap into a bunch of existing machine learning frameworks that are already available - some provided by Amazon, others open source. Infor hosts its cloud apps on AWS, so it makes sense to tap into AWS’s AI capabilities.
However, it’s the building blocks that make this an interesting proposition. For instance, the ION architecture allows data to flow freely between applications and the supply chain, which can then be stored and called on by Coleman, then surfaced to users, who may use conversational UIs to interact with it, or simply be presented with the information as and when they need it. Add the supply chain to this, where Coleman is able to use third party data within the existing application to make more accurate predictions and recommendations, and things get pretty interesting.
For example, what would be possible for a retailer that could have an intelligent system that knew about HCM, CRM, ERP and the supply chain all in one place? It doesn’t take a great deal to imagine what could be possible in terms of personalisation, inventory, staff management, etc.
That being said, this is very early days for Infor, and although it has laid the groundwork for some of its CloudSuites to work in this way, questions remain and work still needs to be done.
Infor claims that Coleman optimises human work in the following four ways:
• It's conversational - interactions are more efficient and natural, and offer a better user experience
• Augmentation - serves as a partner to amplify one's work
• Automation - manages low value tasks like repetitive tasks to enable the user to focus on more valuable work
• Advise - provides intelligent insights to help the user make decisions
Taking to the stage on the first day of Infor’s annual user event in New York City this week, Infor President Duncan Angove, said:
Traditionally the way to get the computer to do something was to write down an algorithm explaining in painstaking detail how to do it. But machine learning is different. It is composed of algorithms that enable software to learn by itself, to train itself, to perform tasks, by exposing it to vast amounts of data.
Compute power has become cheap and powerful. The Internet happened and knowledge moved online. Mobile happened. We developed better algorithms, better AI and now we can train it with vast amounts of data and do it on the super computer that is the cloud. AI is the new steam engine, the new electricity, and it will transform everything. Society is changing, one learning algorithm at a time. Our job at Infor is to bring it to the enterprise and the industries that we serve - that is Coleman, our industry focused AI that will power our CloudSuites.
Angove said the Coleman has the ability to recommend what applications to go to to do a certain task, can recommend what attachments to add, can recommend what information you need within the context of what a user is doing. And as it gets smarter, it can suggest the next best offer, or when a customer is most likely to leave, or what is the next best lead to follow. In other words, it is learning, understanding and making recommendations based on all aspects of an organisation’s workflows.
Coleman can also connect to machines, products and the commerce network - connecting to suppliers, carriers, shippers, banks, trucks, railway cars - making it very pervasive. However, it’s worth mentioning that this is the aim for Coleman. At present we have not been shown any live use cases and it seems that it’s early days for Infor’s use of it.
Coleman is our cloud-based AI platform, with a collection of industry specific AI services, that harnesses our vast dataset and industry knowledge. And it’s wired into Infor’s CloudSuite to make it infinitely smarter and faster. It’s enterprise grade and it’s industry specific.
With Coleman, we seek to maximise human potential. What if you could make the best decision every time? Do the work of two people and then relieve the repetitive transaction task that allows you to spend more time with customers, or patients, or guests, citizens.
Employees spend almost 2 hours every day, 10 hours per week, searching and gathering information. In other words, you hire 5 employees, and only 4 show up. Coleman can find information and execute tasks for you.
Coleman automates, it frees you up to focus on valuable work, relieves you of mundane, repetitive, low value work that can be performed more efficiently and quickly by an AI, with less errors. Plus Coleman works 24 hours a day, 7 days a week. The AI is the UI. Coleman can observe, tens, if not hundreds of millions of invoices being matched. And learn from that experienced, the decisions that drive the best financial outcomes for your organisation.
In a Q&A with press, Ziad Nejmeldeen, Infor SVP, Chief Scientist, Development, explained that the use of ION as a network for data is “very powerful”, as this allows the company to aggregate data from several applications, which can then be used for optimisation and predictions. This is something that hasn’t been possible previously, he said, without a lot of hard work. Nejmeldeen explained:
If there is an outstanding problem out there that requires HCM data, and CRM, and EAM, we would pull all of that from one single source - ION - and so now we can think about new problems we can solve that we couldn’t before, without a monstrous integration with all these different applications.
Angove elaborated on this during a separate press Q&A, where he said that he believes that the wrong approach to machine learning is to build discrete applications that are bolted on to existing apps - which he claims is what vendors have traditionally done. He believes that AI should be a “pervasive, ubiquitous part of the overall experience”. Angove explained how this would impact pricing and product release. He said:
That’s what happens when you take a more of an AI-first approach to rethinking the way that work should get done. For example, you rethink the way demand forecasting should be done, you’re not bolting on a machine learning app, it’s the same app that’s been rethought with AI. So it’s the same sort of proposition. But there are other areas where we are embedding Coleman, where it’s just optimising the decision, there should be a feature in CloudSuite that people are entitled to as a CloudSuite user. For example, inventory optimisation should be a core feature.
Machine learning is a platform that we fuse into the application. Price optimisation is another one, that should be a core application set. So, in most cases Coleman will just ship as a standard part of the CloudSuite at no incremental cost. In other cases, depending on the size of it, like demand forecasting, that would cost what you’d expect a machine learning demand forecasting module to cost. But it’s one of the reasons people should upgrade, because they get to turn on Coleman.
But it’s this AI-first approach that Infor appears to be talking up. It wants to reconsider how enterprise applications *could* be built using artificial intelligence and machine learning. It said during the day that this has been particularly true of its retail CloudSuite, as this was a new venture for the firm and it had to start from the ground up (and has made good headway, with some high profile customer wins). Angove said:
Our science labs team and our Hook&Loop team sit with the development team and say, for each of these different areas across all of these applications, what does an AI-first approach look like? In retail we are way down the path because we had the luxury of starting with a network native, AI-first approach from scratch, because there was no code. It’s been dramatic.
People don’t just want to replace existing, legacy retail applications with something that just looks better and runs in the cloud, they want us to completely rethink it. And that’s what we’ve done in retail. Depending on the industry, we are further along than others.
You’re replacing those tasks, but you’re doing it in an AI-first way. The way we are doing assortment and promotional optimisation, it isn’t the way it would have been done in the old world. The way we predict orders, the way we predict demand, no one wants a replacement for ERPs that were built 20 years ago. The point I’m making is that the way we come at the market is very, very different.
As I said above, Infor has an advantage in that it has re-engineered its CloudSuite applications in a way that is suited to the application of machine learning frameworks, namely via the ION network. It’s also operating on AWS, which has a strong play in this area.
That being said, questions remain. We weren’t given many real-life examples of Coleman in use. Which applications will go first? If Infor is rethinking ERP, what does that mean for the applications that it has lifted and shifted to the cloud, but not had the luxury of designing from the ground up in recent years? If it is going to be AI-first, what does this mean in reality for redesigning apps? How far will it go and where will it start?
What do these AI-first apps actually look like??
Furthermore, we need more information about which machine learning frameworks Infor is using, how its using them, and how these are integrating with the likes of Birst and Predictix.
Simply put, in recent years Infor has been good at laying down the foundations and executing on a cogent strategy. So I have high hopes for its AI offering. However, I need to see this in practice. We need customers telling us how Infor is rethinking its applications for them, using Coleman. I know it’s still early days, but AI is a great buzzword - everyone’s using it - we now need the evidence.