Spring event highlight - SAP customers react to generative AI at SAP Sapphire, and a data imperative emerges

(Tony Caesar, ASUG Board Member, in Orlando)

Time for highlights from the spring generative AI enterprise software conference season! Yes, vendors earned that strikethrough - we heard way too much generative AI happy talk from the keynote stage.

But this is categorically different than the blockchains-are-magical keynotes from five years back.

Generative AI is surging on the backs of consumer adoption. Not to mention the sharp rise in "shadow generative AI" within enterprise walls, a risky endeavor that companies must get a handle on. Vendors simply had to respond to this fervor, by showing customers they have a generative AI roadmap.

Generative AI - how vendors got into keynote trouble

But we ran into trouble also:

• The generative AI risk factors were mostly downplayed (but that is the top question any CIO or CFO needs answered).
• The gap between customer priorities and vendors' AI talking points was often quite large.
• Other areas of tangible progress, such as the impact of robotics on the shop floor, recommendation engines on e-commerce, or automation in finance, missed out on being showcased - even though they are much more mature (and therefore applicable) at this moment.
• Enterprise vendors have a terrific opportunity to address some of generative AI's shortcomings around data privacy, hallucinations, governance, reinforcement learning and explainability. But more time is needed - and that wasn't always clear from the enthusiastic keynote stage.

However, on rare occasions, vendors put candid talk about generative AI in front of attendees as well. One standout example? The ASUG Annual Conference keynote, which took place during SAP Sapphire Orlando (the keynote is streamable on YouTube). Here's a few highlights from that keynote, along with other customer views from Orlando.

SAP customers react to generative AI - ASUG perspectives

Carolyn Dolezal, Chief Operating Officer, framed the AI discussion - asking panelists what they heard on the ground in Orlando. ASUG CEO Geoff Scott said he saw striking differences versus ASUG's past survey data:

When we did the ASUG Pulse of the Customer survey last year, AI and ML was way down the list of things that were a priority. Integration is always at the top of the list. There's other things that go on the list. But AI/ML was way below. That was before ChatGPT came out in November of last year, and started to really take the market by storm.

AI was already a factor. But as Scott says, the mainstreaming of AI was sparked by ChatGPT:

What I would term as a more back office technology, something that we would talk about internally, but was very hard to actually touch and feel, now became something that you could actually type sentences into - and questions into a web browser and get answers back. That changed the game. Something that was out there all of a sudden became very important - became top of mind for CEOs, boards. And by default all of us.

As Scott told conference attendees: it falls on IT leaders to explain how generative AI can be harnessed. The business wants answers - and fast.

The challenge, I think, is that we're not exactly sure what it all means yet... It's not just a fun thing to play with. It has significant greatness and significant consequences. So we now as technology professionals, are responsible for figuring out how to make it work in the enterprise, and come up with a coherent answer before others do it for us.

Risk management is a big piece of this:

What we find is a mishmash of tools that are out there doing their own thing, that are outside the boundaries of reasonable control. I happen to think it's going to be a game changer. But I also think we have a lot of distance to travel.

Tony Caesar, Senior Vice President, Information Technology, Cradlepoint, and departing 13 year ASUG Board Member, added his views:

This technology is not new, but it is new. AI has been around for a while. And now generative AI and ChatGPT is becoming really used in the consumer area. What I'm nervous about as a CIO, is: are there dangers to my organization?

What do I mean by that? I'm on a council with some other CIOs. One person gave me an example where someone in the organization took customer data, and put it all into ChatGPT to create a proposal. The concept sounds really cool, right? 'Hey, I can get a professionally created, presentation proposal.' But now all of that customer's data and pricing information is in ChatGPT. So I'm looking at things from a different perspective.

Caesar doesn't want to hold back innovation. But data leaks are a risk too far:

It's not that I want to be a control freak. But I want to make sure that from my perspective as a CIO, I don't start having data leaks or IP leaks going on through ChatGPT.

For management teams, time is of the essence:

We're going to have to figure out: how do we control it? Because if not, it's going to control us.

SAP's generative AI plans versus customer priorities: it all starts with data

In fairness, SAP is as rigorous with customer data privacy as any vendor out there. But the dilemma customers face right now is: most enterprise software vendors don't have their own large language models (LLMs) yet. Even if they do, customers may want to use a different one - or one that is more advanced in some aspects. It will take time for vendors to establish how the data segregation will work, and what opt-in/opt-out warnings customers will see, if they need to access third party LLMs and apply customer data.

I believe Caesar's concerns will be addressed in time. But the spring season was too soon for complete solutions. SAP's generative AI partnership with Microsoft, announced at Sapphire Orlando, makes a lot of sense - that's a good enterpirse partner for large language model collaboration (I wrote about the HR use cases here: Are SAP's AI ambitions in line with customer priorities? Inside the SAP SuccessFactors generative AI news). As SAP's Meg Bear told me:

We're looking at: how do we solve big problems? How do we do this with a lot of transparency? How do we have a lot of understanding for the fact that this is going to evolve over time, so that the burden is not held with our customers, but that we can absorb both the upkeep and the compliance sides of this, to provide good guidance for them to make decisions on where and how to use these tools.

This does not mean customers are in a holding pattern. Talking to customers across events, a clear imperative has emerged: get your data platform together. Yes, this is not a new priority - but there is a new urgency. SAP, of course, thinks it has fresh answers here, via its SAP Datasphere offering (announced March 2023, with some promising views from initial Datasphere partners).

Whether customers choose SAP's data platform or not, there's no question about a newfound data urgency. This holds up across industries and departments. I find it encouraging how many customers seem to understand: if your AI lacks quality data, your AI will lack quality. What about getting a return on that arduous data platform work? Well, generative AI is still on the horizon for most, but other forms of AI are rewarding that data integration effort.

Getting from data to AI result - an SAP Business Networks customer scenario

Example: at Sapphire Orlando, I conducted interviews with notable SAP Business Network (and S/4HANA) customers, e.g.Sapphire Orlando '23 - Hitachi Energy's logistics lessons from its global SAP Business Network project. A second major SAP Business Networks customer told me about their emphasis on data accuracy:

We have much greater data accuracy with the information that we're getting out, as a result of the new process - and as a result of the systems that we're using with SAP and Ariba. Some of the things that we're able to benefit now more fully is through leveraging of Ariba sourcing. We've recently turned on the AI for that. It was a parameter that we turned on, to give us supplier recommendations, question recommendations - if we add certain materials.

Getting to that initial AI result took work: it required global process standardization.

Now that we've standardized our processes, we're getting the data to fall into the right columns. Now, what can we do if we unleash AI on top of that? It's something we're excited to start looking into.

As generative AI for the enterprise matures, these efforts will create "data possibilities."

We want to explore more AI possibilities with our data. I know it's a hot topic right now with ChatGPT and Bard and all that... Now that we have the data set to unleash AI on, I think we can really see the true value.

It's not just doing new SAP projects. It's doing those projects with a certain mindset about data.

We have that data set because of the tools we're using, in part because of S/4HANA and Ariba - and in part to us taking that opportunity to globalize and standardize our processes, and cut back on customizations.

My take - validate AI tech with peer discussions

Whatever customers might say, vendors are pressing ahead with generative AI. Enterprises will have much more mature generative AI offerings to kick tires on by this time next year. In the meantime, customers have their data prep work cut out for them. In my view, the biggest customer mistake would be to undertake that work in isolation. During the ASUG keynote, James Johnson, Vice President, Chief Information Officer at James Hardie Industries, and ASUG Board Chair, hit that issue head-on:

Data is the lifeblood of any business enterprise... But the paradigm that technology is separate from business must shift, starting with leadership. We need to understand that the function of IT is to be your business partner first, to bring value to the company.

I've had the privilege of working in technology for 30 years... I've also had the honor of being a CIO at four companies during that time. The main thing that I've learned in that time is that coming together, to share experiences both good and bad, is a valuable force multiplier. I found that value specifically in the ASUG community.

Scott laid out one more AI warning:

We had a really good conversation this morning with a number of board members about this. The fundamental question that came up around the table is: does this unlock human potential and creativity? Or does it make us lazier? I'm hopeful that unlocks creativity and potential; I'm worried it just makes us lazier.

I share Scott's concerns. In the end, we'll get the AI we design - and deserve. If we design these systems properly, they should enhance human creativity and "higher level" productivity - at least in the ten year view. But there is a very real risk we'll fall short of this ambition.

One big mistake is what I call the "AI overreach" - misunderstanding or idealizing what the tools can do: using them as justifications for "efficiency layoffs" or adding to the substandard body of content that nobody cares to consume. Generative AI is full of AI overreach scenarios. Better outcomes are up to us - and diligence is required. Bring your enterprise BS detector also; you might need it.