Every quarter, I can count on Oracle to issue a boatload of Oracle Cloud news, along with a parade of customer stories. Each quarter, I go behind the news and ask "why"? Steve Miranda is the lucky
victim recipient of my Q/A - but this quarter, there is a twist.
Oracle's latest cloud news, including supply chain and HCM, is being released in conjunction with Oracle CloudWorld Tour events in London and Austin. My diginomica colleagues Phil Wainewright and Derek du Preez are on the ground in London; you can expect their updates/analysis shortly.
Meanwhile, I dove into the Oracle Fusion Cloud news with Miranda, who is Oracle's EVP of Application Development. This paragraph from the press release jumped out:
New AI-powered lead time estimates in Oracle Supply Chain Planning: Help customers improve the accuracy of lead time assumptions by using machine learning to highlight variances based on actual performance. Embedded in Oracle Supply Chain Planning's Planning Advisor, the new feature can improve planning efficiency and results by identifying lead time trends, anomalies, and their potential impact with prioritized actions and resolution suggestions.
Why? Because in the midst of this generative AI hype tsunami, the comparatively mature integration of other forms of AI into industry are now underplayed. And, in a disruptive macro-economy, achieving better planning results has become mission critical. But how do you get there?
Want better predictive supply chains? Start with the data silos
My AI question to Miranda starts with a fundamental issue: data quality. There's still a problem with siloed data across the supply chain, and the visibility problems that result. I don't care how sophisticated your predictive algorithms are, your AI can only be as good as the data you're feeding into it.
Example: if you order a package on Amazon, and you try to track it closely, you realize there's still a bunch of black holes - even for one of the most deep-pocketed retail logistics companies in the world. It's not like your product is communicating all the time and saying, 'Here's where I am.' So what is Miranda hearing from Oracle customers? First up: customers need (and expect) more resiliency than ever before. He explains:
Since COVID, our customers started to rethink their supply chain, and rethink it for resiliency purposes. Not to say that that wasn't always a factor, but I think that's certainly heightened. They realized that a single source of a single critical component could be a potential issue to them. And so it's gone beyond a typical, you know, cost-to-fulfill and time-to-fulfill It's also a resiliency element. And so that's put the focus on supply chain.
As for the logistics data silo problem, Oracle thinks it has a viable answer:
I'd say that that is one of the points you made as far as the data. It's not just supply chain planning, but supply chain planning that sits on top of what we think is the most complete suite, which includes transportation management, global trade management for import/export regulations, and the supply chain execution system.
Miranda sees another factor, which I would personally attribute to the advantages of true cloud ERP in general: better connectivity to third party logistics services, like Oracle partner/customer FedEx.
The next thing where we feel we have an advantage - and you also alluded to it - which is [dealing with] those black holes. That's precisely what we get with our expanding partnership FedEx. What we heard over and over from our customers is: you're in your ERP, but when there's shipments, you go out to the FedEx portal.
When we talked to FedEx about this partnership, their first reaction was, 'Oh, yeah, we've got those API's, all of our customers use them.' And we're like, 'What are you talking about?' 'Well, they use them because they build their own portal to track their shipping notifications... So it's offline. Our B2B integration is really aimed towards eliminating those empty spots in the equation, and bringing that data together more precisely.
But let's be fair - most enterprises are dealing with heterogeneous environments. If a supply chain platform can't handle diverse data sources, it won't deliver. Miranda's response?
I don't want to sound naive about this whole process. You're also right, even with what I said about completeness of suite, even with extending partnerships to supply chain, there's a bunch of other systems and other data that gets incorporated... We're an open system. Our customers either bring data in, or they use our reporting tools to build out and aggregate that data from those other areas.
Oracle has been integrating AI into its Fusion Cloud workflows for some time - but the latest functionality on "improving the accuracy of lead time measurements" is new. That means customer use cases aren't documented yet. So what gives Miranda the confidence that lead times can be improved?
It's been pretty well proven that humans aren't very good at pattern recognition and statistics. We tend to be recency-biased; we tend to be negative-biased. It's just a natural human trait. Algorithms and database algorithms do better, and the latest press will show you that if you add on top of that machine learning algorithms, it's even better than that.
Miranda's view of open enterprise doesn't translate to pure best-of-breed though. He cautions: if you use best-of-breed excessively, your planning abilities will be hampered.
What I tell our customers all the time, is: if you are a best-of-breed, or if you feel you have a need for a best-of-breed in one particular area, or a couple particular areas, that makes total sense to me. Different industries, different products are out there; we're not naive about that. However, if you said you've got forty different supply chain applications, I would contend that you're very likely be paying more for integration through data problems that [prevent] quick decision making, then you are giving up any one individual feature.
My take - where does generative AI fit in?
Miranda contends that across many supply chain and logistics areas, Oracle Fusion Cloud is a best-of-breed option. Fair enough, but I think his prior quote nailed it: customers do want choice on what apps to plug in - and you're not going to cover every industry requirement. I happen to think customers also need expert guidance here, to ensure that they don't wind up with the type of cluttered data landscape Miranda warns about.
But you can't talk AI these days without talking about generative AI. As readers know, I'm of two minds about generative AI. On the one hand, I find that the techno-evangelists exaggerate generative AI beyond what it's capable of - or will be capable of. But the enterprise has an important role to play in defining viable use cases, and setting up the guardrails to minimize the impact of human misuse or technical "hallucinations." Obviously, Oracle is working away at generative AI product infusions. What is Miranda seeing so far?
You hit my first point, which is: we've been doing this for a long time. I know ChatGPT is hot in general, but we launched our AI apps two and a half years ago, and have continued to. You see it with the AI news here, which is not generative AI at all. So that's part one.
Part two is we do have strong [generative AI] use cases in certain areas. So for example, we're working across HCM to use generative AI to produce job posts, really the description part of job posts. It does that, in my opinion, incredibly well. That's something that you know, for humans, it's kind of tedious - humans don't do it all that great. Frankly, even starting today, you can use ChatGPT, and say, 'Write me a job post for an Oracle Application Engineer.' And it's - in my opinion - quite good. So there are use cases there.
Then Miranda brought up a point I hadn't factored in: the popularization of AI via ChatGPT makes customers more engaged in AI discussions.
I think what ChatGPT has done in particular, it's made it very accessible to the end user... I was having lunch at JPMorgan Chase with their CFO. What really tipped it there: I was able to literally go to ChatGPT on my phone, type in 'Give me a job post for JPMorgan Chase,' and show it to her so that she saw it was material and relevant.
Now she and her team are coming up with, 'Okay, now I have an idea of how to use this' - not to write my high school English paper, but how to use it in the enterprise, with some control, with a human review, so you don't have it saying anything untoward. And that's really where it kicked in.
Because generative AI needs guardrails, governance, and adult supervision - not to mention quality data feeds - I see it as an evolutionary technology in the enterprise, not a revolutionary technology. Before you think I'm being insulting to ChatGPT and its ilk, I've never heaped the "evolutionary" compliment on the Metaverse or Web3. From me, that's pretty high praise.
A cultural revolution does not necessarily translate to enterprise ROI. However, the enterprise pressure cooker can show us what the tech is truly effective at. Miranda adds summarization to the list, something factors into many scenarios, from a "doctor's assistant" health care scenario I demoed last week, to financial reporting:
I would zoom out on your doctor's example and say, the other thing we found [generative AI] does an excellent job of is: summarization. So if you give it a white paper, and you say, 'Make a PowerPoint presentation of this,' I would say it does an excellent job of summarizing a paper into bullet points. Or, to your point, just take one of your TLDR emails, put it in ChatGPT, say 'Summarize; give me the four main points' - again, with some misses, it does a pretty good job of that.
There's a number of places - if you think in financials, you know, 'Analyze this financial report, and give me the highlights of this balance sheet and income statement,' it's quite good and quite close with those things.
Bring on the use cases, then. Something tells me this topic will come up at Oracle CloudWorld London as well.