SuccessConnect '23 - under the hood of SAP's generative AI news, and beyond

As expected, Tuesday's SuccessConnect keynote opener was loaded with HXM-related AI announcements, building on SAP's recent announcement of its Joule copilot, which will provide a digital AI assistant across a range of SAP products and workflows.

But would this show be all-about-AI, all-the-time - at the expense of other important topics to HR customers?

And when will this AI functionality be released? How are the issues of customer data privacy  being addressed, in the context of AI-related data movement?

As I write this, there is a full day of the event (and embedded analyst program) still to go. But I already have some useful answers to these questions. Start with the all-AI question: during the keynote, SAP made clear that they want to tackle a broader purview of HR challenges and possibilities. I welcome this type of messaging for two reasons.

First, I prefer to validate vendor news with customer proof points. With generative AI, I can't really do that - too few customer are ready to talk. I can press vendors on issues of accuracy, bias, explainability, and AI design principles, but not on results. Customers are knee deep in talent and workforce management issues that go beyond the reach of where generative AI can go - at least today. At a user event, you never want to miss an opportunity to deliver for those customers - with more mature functionality.

And yes, for SAP, like most forward-thinking vendors, that mature functionality does include many forms of AI already in use. It is only generative AI that lacks the proof points I am looking for. SAP's news center includes fresh news on several important HR topics. We got into all of that, including details on SAP's AI-for-HR plans, during an analyst session with David Ludlow, Group Vice President, Product Strategy and Research at SAP SuccessFactors.

On generative AI release timing

SuccessFactors customers won't have to wait long. As Ludlow confirmed, two generative AI scenarios are generally available with the upcoming SuccessFactors release in November - along with a slew of digital assistant scenarios.

This is the ability to create a job description, and the ability to create interview questions or suggestions for a manager to prepare [for an interview]...

The two gen AI use cases are embedded in SuccessFactors, powered by Microsoft OpenAI. Then there is the enhanced digital assistant:

There are about 40 use cases for the digital assistant [via Joule] that are being delivered next month, primarily focused on Employee Central and core HR type transactions, but there's a few talent ones as well. We're putting in some more capabilities throughout next year as well.

Such as?

One at the top of the list [for Joule in HR] is manager self-service, a scenario for the manager to create a new position within Employee Central - that was one of the top ones being requested. Another is the incorporation of Large Language Models, to assist in FAQ-type claims. So what is our vacation policy? What was our benefit policy - and then actually get information back from there.

More gen AI plans are in the works too:

In May next year, use cases planned for delivery are using gen AI to provide candidate screening in the recruiting area. And then the ability to suggest employee goals about development, as well as performance goals... And then we'll look at some other things as well, as we mature.

In addition to gen AI, predictive AI for HR is expanding also: 

On predictive AI, once again, the ability to take information that's being generated in the system, and then to use that information to provide a suggestion or a recommendation for somebody to pick it up. The best example here would be to say, 'Okay, it looks like as a result of your performance review, you picked up this skill. Would you like me to update your profile?'

Looking under the SuccessFactors R&D hood - investments in AI and beyond

Ludlow says a top question is where SAP SuccessFactors is investing its R&D. The answer goes well beyond AI:

A key question we get asked a lot is: 'Where do you spend your R&D? Where does the money go?' I divide this into four or five areas - a non-exhaustive list. We're making a lot of investments in user experience to connect the end users of the services you're providing to the technology that's providing. Secondly, build a great future-ready workforce - all about skills, capabilities, internal mobility, upskilling, and rescaling HR agility, compliance. The ongoing updates we're making to the solution - including other legal compliance that's coming in, but also all of the customer requests or requirements for innovation and ongoing improvements that we're making to the solution as well.

[Then there is] integration and extensions,.We continue to refine the integrations within SuccessFactors, as well as deliver new integrations to connect SuccessFactors more seamlessly to the rest of the SAP suite.

Of course, AI is a crucial part of that R&D also:

For simplicity purposes, I like to divide this into the three main types of AI that we are embedding into SuccessFactors, and the use cases that we're delivering associated with it. The first is generative AI, also known as ChatGPT.

Secondly, conversational AI, also known as digital assistant and chatbots, and the third, predictive AI - so taking information in the system, and making suggestions and recommendations to end users.

Ludlow's answer indicated how SAP SuccessFactors thinks about AI overall - with humans decidely in the loop:

When we talk about AI, we will never take a decision for the end user, the decision machine or the software simply makes a suggestion or recommendation to an end user, but it's still up to them to actually take the decision to update something. So in that case, we help reduce or eliminate bias, and any other kind of bad data that may get into the system. On the generative AI side, we're partnering very closely with Microsoft in SuccessFactors, leveraging some of the open API services running out of this Azure, that is ChatGPT 3.5 turbo.

Ludlow also noted the evolution of conversational AI:

We've had some conversational AI capabilities in the solution, some using native technology, SAP, SAP for decision to retire that technology. Instead, we've incorporated IBM Watson code into the platform. So going forward, the IBM Watson capabilities will drive our digital assistant in SuccessFactors.

SuccessFactor's UX redesign - will AI disrupt it?

I've focused on clarifying the AI news/approach, but that gives short shrift to the range of new offerings underway. Another big topic was the enhancement of the SuccessFactors UX. Ludlow explains:

So in SuccessFactors, we've been on a multi-year journey to redesign and reimagine some of the user interactions of the applications within SuccessFactors. So these include things like reimagining the goal experience, reimagining the performance experience, reimagining the homepage.

We've just delivered, in the September off cycle release, a completely new redesigned user experience for recruitment. In the November release, we will deliver a brand new user experience for the [Work Zone].

SuccessFactors is also part of SAP's Horizons project, which standardizes UI style sheets across products, from S/4HANA to SuccessFactors to Ariba. Some of those UI views have already shipped, but the Horizons project is ongoing, and outside the scope of this piece. What caught my ears? SuccessFactors' push to embed functions in users' preferred collaboration environments, including Microsoft Teams and Google Workplace:

For commonly-used employee/manager self service for transactions, we're doing a lot of integration to Microsoft Teams. So things like give feedback, receive feedback, request a day off, enter time, these kinds of things, we can actually connect the SuccessFactors transactions directly into the Teams environment. So an end user might never have to leave the Teams envrionment to do these very simple and basic HR transactions.

We delivered some of the first integrations already in the May release this year. We'll deliver some additional ones in November, and then into next year as well. We are also in the process of doing the same with Google Workspace.

Bringing interactions/transactions to the users' preferred workspace or UI makes sense. But I asked Ludlow: when you think about putting a prompt-based AI in front of software, doesn't that fundamentally disrupt UX plans? He responded:

I think the disruption will occur from multiple perspectives. The ability, like I mentioned, to integrate the self-services into Teams is one aspect, However, I can also do that through the digital assistant as well. I think you can actually get to a point where the homepage just becomes similar to a Google search bar. And so I want to do this, or I want to do that, and you do it by conversation.

If you think about an [enterprise] homepage today, fundamentally, it just has multiple links, right? Now I can go here; I can go here. And there may be some search technology to find the right link. But in this case, if you really take it to the Nth degree with these digital assistants - it's just "I want to do something,' and the system will respond to what I want to do. It makes it so much easier.

My take - airing out AI issues and more

How AI disrupts enterprise UX led to a spirited discussion. There is no one right answer, but the obvious note is that an AI conversational front end won't be able to complete complex transactions anytime soon. Nor would that type of conversational interface work for all employees. Some will insist upon their screens; some will need them.

Good UX design must expand to anticipate all of that. In general, going to where users already are, including Teams/Google etc - and automating as many processes as possible along the way - is the winning design principle. The conversational AI interface will start with simpler transactions, and then we'll see. The Joule copilot should eventually be able to jump in with relevant, role-based next steps.

One of the most interesting discussions we had during the analyst sessions was the potential for more transformative AI. How AI could change the user experience is just one small way to get into that bigger conversation. Will SuccessFactors be able to lead its HR customers through more profound AI changes - changes that could change the nature of work, and certainly how we manage talent and skills? Time will tell.

From my digging so far, an important piece will be SuccessFactors' progression on their own skills ontology, as well as using AI to enable real-time updates of that skills ontology. That's hardly the only AI use case, but if you want to transform work, that comes down to talent, which comes down to skills, which comes down to skills data that users buy into, trust - and help to maintain.

I ran into a few customers that are eager to do more with SuccessFactors on skills, but in the short term, they have gone for other third party skills solutions. This was a random sample, not necessarily a trend. Nor does that preclude SuccessFactors AI using that externally-integrated skills ontology - that should not pose any problems. But: I think the continued emphasis on skills ontologies will strengthen SuccessFactors' overall AI push. 

During Ludlow's session, analysts asked the expected question about customer data privacy and LLMs. What I'm learning is that even when vendors use external LLMs, if they do this right, the LLM vendor in question does not really have any access to that data - not for model training or anything else. LLM data movement is an important question, but it's a problem a responsible AI vendor can avoid. Ludlow's answer underscored this:

The way we're working with Large Language Models is: we are not sharing data out. So we're using data, but the relationship we have with the supplier is they may not use it to train their LLMs. We would never send personal data out to begin with, but [this applies to] company level data as well. A lot of companies don't want that shared out. So once again, even though it's not PII type data, we are not allowing the LLM suppliers to use that data to train their models.

That brings me to pricing. I've said before that I like many aspects of SAP's AI strategy, with the important exception of pricing. I stand by that criticism for now, but AI pricing is something of a moving target, for most if not all vendors. Pricing discussions were not part of the agenda at SuccessConnect, but I'd expect more details on AI pricing for SuccessFactors to be revealed soon, obviously by the November release. I would expect these pricing plans to be 'umbrella' plans across SAP's products.

During our analyst sessions, I pressed the issue on a couple topics I believe are important to customers. However, those sessions were mostly under NDA, except for Ludlow's. The good news is that what we discussed under NDA is largely shorter term, so that content should eventually be appropriate to surface. One thing I can say: talking to a number of analysts about SuccessFactor's positioning, they believe that SuccessConnect can go much further in tying SAP's HCM approach to its formidable industry competencies (and software reach).

Yes, you can get that immersion at SAP Sapphire, but getting more of that industry vibe at SuccessConnect is high on the list of improvements from analysts. I think that fits well with SuccessFactors' goal to support business cases that HR leaders need for boardroom relevance - including the collaboration needs of a more fluid workforce. Ergo, the progress made on integrating more deeply with Fieldglass.

In the meantime, I am also recording podcasts with long-time SuccessFactors community leaders. Those podcasts will allow me to go directly into what matters to customers now - including AI impact and a lot more. Customer sessions were abundant - as were project lessons. Stay tuned.

Updated October 5, 6am UK time with a number of small tweaks fro reading clarity.