Making AI work in HR (EN)

00:00:00: How can organizations introduce AI in HR successfully?

00:00:05: Today, we explore the journey behind responsible AI adoption.

00:00:17: Welcome to HR.

00:00:19: What's Next!

00:00:20: Quick insights new perspectives and practical ideas on the future of HR.

00:00:26: I'm Hinata your host.

00:00:28: let's get started.

00:00:30: what can organizations learn from real-world AI experiences?

00:00:35: Today, we'll explore Greynke's approach to AI in HR from evaluating use cases and creating the right framework for putting solutions into practice.

00:00:46: As a global financing partner, Greynker helps small- and medium-sized businesses finance investments through flexible leasing solutions.

00:00:55: Joining me today is Jürgen Mechler, Programme Lead HR Transformation Technology at Greynken.

00:01:03: Great to have you with us today.

00:01:06: Yeah, thank you very much!

00:01:07: Thanks for having me.

00:01:08: When you started exploring AI at Grenker how did you approach the topic?

00:01:14: How did make sure that AISolutions were introduced in a structured and responsible way?

00:01:21: The way we faced this challenge was We came up and defined as three-stage concept For understanding evaluation And activation of an AI capability.

00:01:32: I think all of you are finding yourself in a situation where many vendors come up with AI capabilities.

00:01:38: And to get this on the road and your business life, it's quite a challenge!

00:01:44: We're running through our three-stage principle.

00:01:46: In first stage we want fully understand how the AI capability works.

00:01:53: This does not only mean above surface.

00:01:55: so what do you see?

00:01:58: What is response like for example if were talking about the large language model?

00:02:03: Not only like that, it's also relevant what is happening below the surface.

00:02:07: So to see where are data flows going?

00:02:09: Is data still in your control or has it been moved maybe to an open large language model by you?

00:02:16: don't have full control about the data... Maybe your sensitive data which was being processed into AI case.

00:02:22: so You should really be sure to fully understand instead number one What is going on here and not only what other marketing slides like of the vendor.

00:02:32: In stage two, we are collaborating with our internal control board for AI.

00:02:37: And this is a board as the set of people of X-Verts or AI and there you have business owner in touch with your vendor providing their capability e.g.

00:02:47: Akoso or SAP.

00:02:50: After fully understanding the use case You can now present and pitch your use case to this board This Board considering also an evaluation based on the EU AI Act to make sure, have a proper and external valid classification of your AI use case.

00:03:08: To see if it's going be non-critical or highly critical in AI use cases.

00:03:14: depending upon that result you need to leverage project plan later when it comes to stage three activation.

00:03:22: For the activation, you take the inputs from stage number one and two.

00:03:27: So what is your AI case like?

00:03:29: And maybe what's a risk or the regulatory perspective that we need to consider as a result of evaluation in step two?

00:03:38: when stage three brings this into life technically but also on an emotional side it has strong impact usually for processes and people so change management is quite important here.

00:03:53: That's a really thoughtful way to build structure around something that could easily be chaotic.

00:03:58: I really like that three-stage Way of working, you know first understanding then evaluating also looking at the importance of regulations right and Then a proper rollout.

00:04:11: Can you share one or two examples?

00:04:14: Of how AI is already helping employees at Granker day today?

00:04:19: Yes sir.

00:04:20: So the very first two use cases we brought live are on one hand side conversational AI and, in other hands a text generation support because you can imagine as an employee maybe don't know who should ask for some specific questions.

00:04:41: For example What is your holiday entitlement like?

00:04:46: Who's your HR contact or what my job about.

00:04:51: These questions, they can be covered by an embedded AI capability in our HR portal and it's a chatbot available.

00:05:02: people can just reach out to at any time on the browser also their mobile yeah depends of solution you go for but capabilities that its available.

00:05:12: And it can also be extended for other use cases, if they are more relevant to you.

00:05:18: So we have a high accessibility of that.

00:05:21: and just breaking down the hurdle thinking about who should I reach out to ask their question?

00:05:26: You could engage with the conversational AI at second-use case of AI power text generation.

00:05:35: this is about many moments where people are confronted text, some text content in the HR portal.

00:05:44: For example In a moment of having a performance review providing feedback or setting goals for an employee and role-of-the manager for example.

00:05:53: And there this text generation feature is quite handy because you can not only adjust tone extend length You also make sure to have no bias.

00:06:06: It also restricts wording you use.

00:06:09: So even if your mind uses harsh wording, it's going to be corrected and presented in a straight novel way.

00:06:15: so that is very convenient for people to save time but also have better experience when using translation capability to give feedback just in the mother tongue of an employee which maybe not capable yourself as manager.

00:06:35: how something is every day as a chatbot or smarter writing tool can genuinely make people's work easier.

00:06:41: I love that, you know?

00:06:43: You're addressing the topic of AI bias because that is really important and even there is help with things like tone and translation.

00:06:54: looking back though What are the biggest lessons you've learned from introducing AI at Granker?

00:07:00: And what advice would you give to other organizations just starting out.

00:07:06: Yeah, I think considering our three stage principle and also involving control instances in your company it's highly valuable to start early because all this procedures of approvals and pre-engagement with stakeholders you need to have before you can actually kick off your project.

00:07:28: You should gain full understanding, make sure that you have enough time to run through these processes as well to maintain constant engagement throughout the whole evaluation delivery and also life phase of your AI capability.

00:07:43: And when we started off picking distinct cases, We looked at what are the most relevant angles to our employees?

00:07:50: For example.

00:07:51: We've been talking to our HR colleagues through a help desk and then to determine What are the more frequent questions you're gonna be asked by managers and employees and we try To make sure that our conversation LAI from The very first moment picks up these Questions and providing an appropriate response That people can work with.

00:08:09: Because just imagine if you present a use case and bring in it live with low impact, You would sort of crash the wave before you can even surf.

00:08:20: So that's relevant to pick high-impact use cases And last but not least making sure people stay on track.

00:08:30: Just outline what is this story gonna continue like?

00:08:32: because you can't imagine If we bring life for first use case It won't be your last.

00:08:37: Yeah, it's just a starting off of the journey.

00:08:39: And this is also how you can engage with people and also adapt acceptance to say okay maybe It's not very best experience because its continuous process of evolvement and development That what peoples should understand.

00:08:54: They are more like mentally on track To be curious, test things Also find edges.

00:09:02: Maybe they gonna come up With some new use cases Which you pick for your next phase of rollout.

00:09:09: Having a starting point is really important and I like how you said that when you start early, picking the right use case does seem to make a difference in terms of how it takes off.

00:09:23: Later on during this journey.

00:09:24: obviously people need to be engaged from there.

00:09:30: When it comes to using AI responsibly though What principles matter most, especially around governance compliance bias and decision-making?

00:09:43: Yeah.

00:09:44: I think you can imagine governance compliance IT security another control instances.

00:09:49: they are very important to have a full evaluation of your AI use case And that's why we brought this all in together instead.

00:09:58: number two off our implementation approach.

00:10:01: So at first We are not only engaging with the control instances, we also collecting a lot of information and documenting the results.

00:10:10: Documenting on informations from the vendors how is data being processed?

00:10:15: I know that what we've been talking about earlier but also documenting which decisions have considering what state of knowledge, because maybe there an AI capability that you're using in production will develop with change.

00:10:30: Will be extended by the vendor?

00:10:32: You might need to rerun evaluation Because may be data flow changes Maybe and outcome after I use case changers And out-of-the sudden a formerly maybe non critical use case which could be executed autonomously will change to a critical or highly-critical use case.

00:10:53: And this would trigger some clear steps you need to involve, for example put human in the loop which has not been part of before in process and so still have full control on understanding what is going on in AI when using such tool.

00:11:13: It's clear that keeping that human in the loop and staying on top of documentation is essential.

00:11:18: One thing gets more critical, right?

00:11:20: I mean we always talk about how you know AI is there to support but You always need Human oversight To actually make decisions.

00:11:30: Looking ahead where do you see The biggest opportunities for AI at Grand Care And what areas are you most interested In exploring next?

00:11:40: We're currently looking at AI use cases in recruiting.

00:11:43: I can imagine recruiting being one of the areas or an HR where AI can be a strong benefit, and all the procedures.

00:11:52: for example candidates would just be able to upload their CV And then AI capability might pick up that information matching this To the current vacancies and automatically present the best fitting opportunities to the candidate and they can decide to apply for it or not.

00:12:10: Secondly, the candidate's information could also be matched with internal requirements and presenting a matching score to the recruiter.

00:12:18: To determine which candidate would be properly good choice or not.

00:12:24: And here like I just said that the recruiters come along being human in loop because it is considered highly critical process if, for example a candidate would automatically be refused because of low matching rate.

00:12:38: So that's highly critical!

00:12:40: And we see two other high impact use cases in our organization and this is when you think about extending your conversational AI.

00:12:47: Extending in two areas where the highest impact on employees.

00:12:52: Area number one are common questions.

00:12:55: with regards to time tracking We want to give answers throughout chatbot providing answers depending on policies we have, but also considering latest time tracking information from the employee in a system and maybe providing some answers.

00:13:11: Depending on that or even some help where the conversational agent might pick up and carry on some further steps in the system to help the employees for example That an additional working-time is being considered the right way.

00:13:25: And The second extension Is the area of benefits?

00:13:28: I think every one of us loves benefits And this agent should be able to not only provide information what benefits our employee is being entitled too, but also considering which ones for example are already part of current payroll.

00:13:43: So maybe there's a gap and maybe an employee can just opt in for additional benefit?

00:13:49: The agent can pick up that information.

00:13:51: may we provide these two an approval have human interloop again at the end providing the output and a new benefit information to payroll, to have this available for your employee at next moment of payroll.

00:14:05: Everybody definitely loves benefits!

00:14:09: Those sound like some genuinely exciting directions to keep an eye on especially with recruiting and expansion in what chatpots can do.

00:14:18: Thank you so much Jurgen for sharing our experience and for giving a clear practical view of what AI governance looks like in HR, especially at GranQ.

00:14:31: Yeah thank you very much.

00:14:31: I enjoyed it!

00:14:33: Thanks for listening to HR.

00:14:34: What's Next?

00:14:36: If you enjoyed this episode make sure to subscribe.

00:14:39: See You next time.