When AI Redefines Work with Shlomit Gruman

Thomas Kohler:
Today’s guest is Shlomit and I’m really looking forward to this episode already. We met at the Human AI Dinner, which you mainly hosted, but I could be a co-host as well. And you’re now a Transformation and People Executive and also four times Chief People Officer.

And I’m really looking forward to this episode when we talk a bit about all the changes regarding AI and what does this mean also for organisation and change and uncertainty. Before we do that, let’s start with a small introduction about yourself.

Shlomit Gruman:
Yeah, thank you. Thanks, Thomas. And it was a great pleasure to do the dinner together.

I thought it was an amazing evening. So you already introduced who I am professionally. My name is Shlomit.

I live currently in the Netherlands, where actually for the past almost seven years, been in organisations for 25 years now. Also as a Chief People Officer for 13 years, worked and lived in five different countries, worked in tech, in consumer tech, banking, SaaS, seen all the different, I would say, use cases, hyper growth, transformation, growth, investments, divestments, acquisitions. And at the same time, my motto in life is, I don’t know, or more I don’t know than know.

So of course, all these experiences and navigating a lot of uncertainty and ambiguity, it built a lot of muscles for sure. And I have a lot of bruises and a lot of learnings. At the same time, and I know we’re going to talk more about it today, especially in the times that we are living now, we actually mostly don’t know than no.

And I’m really happy to be here today.

Thomas Kohler:
Likewise. And what, so what made you organise a dinner or an event around human x AI, I think was the overall tagline. Why did you come up organising this?

I often see this from providers that are doing a lot of dinners. But for somebody like you, who is now in, let’s say, executive, also organising it, I think that’s very proactive and great. But what’s maybe behind that?

Shlomit Gruman:
Oh, okay. Now that’s going to be a story. So first, I extremely passionate about bringing these two worlds together, the human and AI.

I myself started the journey into AI or deeply into AI around eight months ago. So it happened as part of experimenting in an advisory capacity, which is currently what I’m doing. So I decided very intentionally, as mostly I don’t know, to educate myself and immerse myself in AI.

Today I only work with agents. So I used to lead teams. Now I lead agents.

Always prefer people, I have to say, you know, but I fully understand and get how blended the world is becoming. And more and more we will see leaders managing both humans and agents. So I got very curious about a topic.

I started learning it, going very deep into the different use cases, obviously the different tools that are out there, solving different problems, being part of a community also that is all dedicated for that. And then I felt, OK, you know, I have all this experience, muscles, bruises from leading change, transformation, obviously a lot of judgement that came from this world and now also understanding of AI. So I got really, really fascinated with how I’m bringing these two things together.

So managing change in the AI era with the learnings of transformation, scaling changes, et cetera, and with what I know now about AI. So that’s kind of a big passion of mine. You can call it leading AI revolution, leading AI transformations, helping CEOs and companies to navigate all these changes and make sense of this moment.

And then also to bring leaders together to really reflect on this topic of what actually that means when you need to lead both and redesign work in companies. As I truly believe that many companies don’t get that this is not a tool adoption, but work redesign. So that’s been a long story into the point of the dinners.

The dinners is just an opportunity to connect with other leaders on this topic. And I’m very happy that I was able to find great sponsors that are ready to bring people together to such an event. We had two of them so far.

So we had one in Amsterdam, second in Berlin, both of them also with a company called Juno Journey and obviously different with yourself in Berlin and different partner in Amsterdam. And it’s just been a great success. We learn from dinner to dinner in terms of what people want to hear, the conversations, the panel.

But it’s just like, you know, it’s people want to connect and that’s a great opportunity to do so.

Thomas Kohler:
Nice. And now you made me curious on examples. You manage agent.

What agents and for what workflows? What specifically do you do with it?

Shlomit Gruman:
Yes. So I have different use cases starting from managing all my expenses. So that’s a clear workflow that I had to build as, you know, there are things that I’m very passionate about and I intuitively I enjoy doing.

And there are things that I less enjoy doing. So I build an agent that deals with all my expenses and things that I just need to keep track on on a regular basis. Obviously I am doing a lot of work with AI when I need to create documents.

Of course, I apply my judgement, but AI helps me a lot to structure, to get the correct research that sometimes is needed inside context into it. And then we do some iterations around that. I obviously work with the note takers for all my meetings.

I analyse them. I also have a special project in Claude of analysing all the note takers and all the learnings out of them. I’ve been doing some things on Lavaball.

So it depends. So I, notebook LM, one of my favourite tools, I’ve been building infographs there and presentations. So it’s either agents that I have either on Chachapiti or Claude, or some workflows that I created, or I’m just using the tools and I’m mostly on the pro versions of the tools.

So I’m using the more enhanced versions of them, solving for different problems that I need to solve.

Thomas Kohler:
Okay. I also think that it’s really great. We just got ISO 27001 certified.

It’s like an information security norm internationally. And it’s really, I would say, intense to get that also for an organisation already with a lot of employees. And I think we used a consultant plus their internal documentation as recommendations and then also build our custom agents just for this ISO certification with the PeopleWise context plus also the context from the consultant plus also with, I would say, prompting in order to use also relevant context for different situations that we could build also a lot of customised documentation for workflows that were maybe already there or we already used and then also adapted them in a documentation, but also created internal documentation that is then, let’s say, the employee facing documentation out of the official ISO certification documentation.

And I think we were one, so the consultant told us and also the TÜV-SÜD that we were one of the fastest certifications because we did already everything AI native and we also really did not have any major issues, right? It was just really agentic and AI generative from day one in setting this up from scratch. And this was really fast and accurate.

And I was really surprised that we saved a lot of time and a lot of annoying work. A second example, what I also use it for is financial planning. Often I can just also in the car, I have my, let’s say, voice conversation with somebody.

I go to a different, I would say, project type where they have certain instructions on how to behave. And then I use a lot of conversations like we hired now these new people. What’s the impact on, for instance, our financials?

What do we need to take care of in terms of infrastructure? Because it’s like pre-set what factors we have in terms of cost versus revenue. And then I always know when going into the car where we’re standing financially.

It’s of course a draft and so on. And it’s also just rough estimations. But that’s often enough until the real financial executive summary is coming from the internal controlling, which is often a bit delayed because it’s detailed and it’s 100% correct, right?

So there are so many great use cases. Did you also see in organisations that you maybe consult how AI is used in a good way, but also in a bad way, maybe?

Shlomit Gruman:
Oh, yeah. I mean, so first, just reflecting on what you said, I think, and that’s what I saw about myself. It helps you to become so much more efficient in things.

So like the speed of it, like to your point, right, I give this data and then I get this output immediately, like in a few minutes. If I’m thinking like I’m now like when I need to do some work on Excel, I’m just going into Gemini, I give it some, you know, data or context and then I get an Excel, you know, built for me. This is just incredible.

So I’m just thinking like the time that it saves us to do the thinking work and the value add work, this is just incredible. So and another point of reflection from what you said is the point, and that links me also to organisations, you said something extremely important. We can just build it from scratch, right?

Like we build it. It’s there. It exists.

We go. For many organisations, the problem is that they need to break their workflows. They need to break their work in order for something new to emerge.

Thomas Kohler:
And maybe that the workflow was set up in a non-AI native way and then you have a problem.

Shlomit Gruman:
Yeah, but it was most likely set up in a non-AI native way, because when you think about change and transformations, you are changing workflows that were built some years ago or processes that were built some years ago. And with AI, it doesn’t make sense anymore or the process can completely look different because you are now bringing agents and humans or you bring agentic AI, which is about agents working with agents, which is a different story. So you really have to redesign work.

And what I’ve seen, what I’ve been observing is that, of course, you know, in organisation, mostly people know how to use chat GPT or some sort of chat, you know, either Co-Pilot or Gemini. But the problem is that it’s not actually changing the work itself, or rather, the companies are not capturing the value of that usage. It’s what I call shadow AI.

There is shadow AI happening, but it’s not necessarily structured. So one mistake that I’m seeing companies are doing, they are treating it as a tool adoption and not as a work redesign, meaning they are not really sitting and looking at every task and getting very, very clear about what has to stay with humans, what has to be augmented with humans and what needs to go entirely to AI. And AI will do a better job or can already solve for it at the fullest.

So, of course, you still want a lot of judgement in the process, but not everything requires humans in the loop.

Thomas Kohler:
Do you have an example process where this could be already done? Would you describe it?

Shlomit Gruman:
Yeah, I mean, let’s take even recruitment, right? I mean, I think it’s an easy example in our context. AI with the right context, if you give it the right context, it can certainly source for candidates, it can certainly point to the more suitable candidates, resumes.

Again, context is extremely important because you want to make sure that AI doesn’t have bias in the process. So the humans have to tell the agent, whatever you’re using, what to pay attention to, absolutely. But then AI can help you to do a lot of the work that before you needed sourcers for it.

I mean, like I remember the previous model of recruitment, you have recruiters and you have researchers, you have sourcers, you have somebody that schedules interviews. A lot of it today can be done easily by AI.

Thomas Kohler:
What I see and I can also mirror that, what I see in recruitment at the moment, I think for the sourcing piece, I did not see a product that is really working well. But if you use it well for market mapping, which would be, let’s say, the pre-step of sourcing, it works very, very well. Then also for building funnel reports, that’s not what all the companies do, but I think should do, is to really analyse existing data or past data.

And then also just build a forecast when we see, let’s say, an interview, let’s say an account executive in Munich for selling to enterprises. You use this data and you want to show your hiring managers or your company how long it takes for somebody to get into the company and then get productive. Because to sell an enterprise deal is usually nine to 18 months, right?

So every enterprise seller that is in the first year maybe did not even do a sale. So when this is the context that you can take also in terms of hiring and setting expectations, AI is really great for that because you can showcase diagrams, what you said, showcase funnel metrics based on benchmark data or internal data and also showcase this to the hiring manager, to the business to say, hey, in order to solve your problem, we have this market we can hire from, then we have a number of available talent or not available talent, just a number of talent. And then the question is how many of them are available to us? Then when you gather data in terms of recruiting, you can ask for the salary expectations, their target attainment and build a market map to say, okay, in these companies, these people who achieve a hundred percent earn that amount of money and these are roughly the activities what they do in sales cycles in order to achieve the target, if we cannot provide a better job offering than that, we most probably want to hire from there, but cannot hire from there because why should they leave if we don’t have a better offer, right?

And if somebody then would change, maybe they did not perform well there, will they perform here, right? So I think this is something where you can then really have a lot of honest conversations, but it’s just, again, important that you can do that only if you have the data foundation, the data model, and a lot of companies, they don’t gather this data, it’s just somewhere and it’s not consolidated, right?

Shlomit Gruman:
Yeah, so exactly. So I think there are a couple of things here also from what you described. One, data is key because if you’re connecting your tools, like it was always the case, right?

Even before AI, but it’s especially important with AI because the wrong data, wrong outputs, that’s like, or if you are not very clear about what data you are connecting the system to, you’re not going to get the result that you’re expecting. So data, internal data, external data, extremely important and apply judgement also on this data, on this raw data that you’re connecting it to. And I think the second thing, it’s like, let’s take recruitment and use this iceberg model, like above the line, at the line, below the line, everything that has to do with interviewing people and making decisions, this is humans, right?

You still want to have recruiters doing the process, hiring managers and so on. Then you have the augmentation, you know, you’re using AI to help you do the market research.

Thomas Kohler:
Or the scheduling, as you said, that’s also really good for that.

Shlomit Gruman:
The scheduling, I think actually scheduling is even below the line. Scheduling can be like, just give it to AI, you know, I will solve the process for you, but then there are things you want to augment that humans are still in the loop in like understanding the data, applying judgement to the data, interview, the interview itself, it’s also another good example where you can use a very sophisticated note taker that will even give you notes based on your values and mindsets that are important to you in the company, and still you’re assuming you’re in the driving seat to make the final decision, so I think that’s a very good example where, where you take a process and you redesign it with AI, thinking very intentionally data, which systems are connected, and then being very, very intentional in designing what stays with humans, what is being augmented and what, just give it to AI, you know, and free up all this time for more meaningful work, and this is just one example, right? I mean, I can talk about, I, I built an onboarding plan using an AI, and that was just incredible, like in half an hour, I got like top-notch onboarding plan, like when you need to do some massive hirings and you want to build a plan on what’s going to happen in week one and week two and week three, including like a bootcamp, just by giving some context about the company, I got like an, like an incredible onboarding plan, which of course, you know, needed some iterations, that’s, that’s normal, but again, the speed of, of which, you know, I got it, something that maybe two years ago would have taken several days to build.

Thomas Kohler:
Yeah. And also several stakeholders from several perspectives, right? That all need to sign off something and something, and now you can just propose a proposal that is already maybe considering their perspective when you prompt it in the right way and we have, and then you have access to the right information, and then they can also review stuff faster or accept stuff faster.

And then, yeah, you’re just faster and more accurate. I also used it for internal onboarding purposes that we have a very specific plan and so on, and then I think it’s really structured, but there are still the, let’s say human challenges to find the right time for that and right, and then making sure that people are really doing it. So that’s maybe the human part.

Shlomit Gruman:
Yeah, there is still the human part, precisely. And I just like on something that you said, I just wanted to add, prompting is, is, is extremely important, no doubt about it, and there is also the concept of super prompter. I would say that context is like the key.

Context is the king, is the queen, whatever. So it’s, yeah, prompting is good, but what is really important is that whatever you are using, and that’s why data is important and connecting it to everything that is happening in the company, AI needs to know your context and the, and the more it knows it’s your context, whatever it is, it can offer you better solutions, whether it’s like getting to know you as a person or the company or what you’re trying to achieve. So context management is becoming key in really getting the most out of AI. Yeah.

Thomas Kohler:
And I think therefore you just need to find a fine line, right? What is, what is the right data you can also feed into and also what you maybe want to augment or synthesise as data when things getting maybe too personal or too sensitive. And I know the US and the EU have different perspectives and opinions on it.

I think maybe the USA way is a bit extreme on unregulated, unregulation, and maybe the EU is a bit extreme on regulation, right? So maybe something in the middle is also fine to start with.

Shlomit Gruman:
Yeah, no, I think safety and security and governance around it is very important, otherwise people are not going to trust it.

Thomas Kohler:
Yeah. Yeah. So what I also think is necessary that testing there is really important, but from testing to getting it into production, that’s a different piece.

And I think this is where the human aspect is so important because ultimately you need to change habits and behaviour on different workflows. It’s actually nothing than basic change management, actually. How do you do that properly to adopt, let’s say, AI testing process or test process in a testing environment into a company-wide process?

We did not do that really properly and roll it out. We are somewhere in the middle with some processes, but still definitely not there that I could say, oh, this is rolled out, it’s working, the company is using it, it’s part of the DNA and core process. So that would be interesting for me also on how you do that.

Shlomit Gruman:
Yeah, I love this question because, you know, for me, this is the product mindset that is just now being amplified. And what I mean by the product mindset is you always need to approach everything with a huge degree of iteration and it’s a mindset, the mindset of launch and learn or test, fail, learn, you know, whatever you want to call it, is that, of course, you know, you have context, you have data, which is a big part of the discovery, the discovery process in everything that you do, and then you have to define what is the smallest move that I can next take, like experiment with, and then you go out, you get feedback and you build on that feedback. And I think that’s extremely important in every aspect of the company to adapt this mindset.

With AI, it’s becoming more critical because it’s a lot of, you can experiment endlessly today. You can just, you have an idea, you go and build something, you try it out, you get feedback, great, you amplify. No, you kill it, you move on to the next thing.

So it’s sometimes also hypothesis that you need to test and get input on. So I think it’s really mindset. It’s the mindset of constant iteration.

It’s the mindset of curiosity, of constantly learning, of it’s okay to fail, you know, as long as you are not destroying the company, because for certain things you have to be very careful about, like you have to create the guardrails and create boundaries around it, but let’s say most of the things we can, we can test and we can learn and sometimes fail and that’s okay. And we, we also going to learn from it. So it’s just this mindset of constant iteration and treating everything that you do, not as a process, but as a product.

Now I’m not against the word process. I think it’s a good word. And eventually you need a workflow, needs to be clear, like step one, two, three, et cetera, but the mindset should always be like, you know, we don’t know, things are changing so fast that we constantly need to iterate and most of the things that we are doing are up for iteration, whether it’s with our customers that their needs are changing constantly and there is feedback that is coming from them or something that like you need to build internally. Also the needs of your employees are changing different stages in the company and you need to iterate.

So from my perspective, it’s a lot about the mindset, then creating the environment that encourages that. I mean, if it’s, I can talk a lot about psychological safety and that it’s okay to test and it’s okay to learn, that it’s okay to talk about your learnings. This is where leadership comes into play.

Because at the end of the day, if people don’t feel safe, they’re not going to experiment, they’re not going to try, they’re going to play it safe and then that’s going to kill your constant iteration mindset.

Thomas Kohler:
And the incentivation is also important, right? Because if you, let’s say, create the right workflow that is then eliminating your job from a narrative, that’s also maybe not what you then naturally will do.

Shlomit Gruman:
Yeah, none of it. I mean, and I can understand it’s very human, right? It’s very normal that people will be fearful of their jobs and about their future, and I think this is also where leadership comes.

So beside, you know, the mindset and creating the right environment is to allow people to see the opportunities in all of it, because I can say, you know, and I can talk openly about the fact that, and I have it from my own experience, careers are changing in front of us, right? So it’s, if people are looking for stability, it’s very difficult to find these days, it’s very difficult to guarantee that, oh, things are just going to be stable the way it is for the next foreseeable future, because we really don’t know, and the experience that we have for the past, especially five years or so, five, six years, is showing us that we can’t predict anything, you know, it’s like, it’s so many unknowns, and the only thing that you can absolutely control is yourself, you can absolutely control your mindset, you can absolutely control your skills, your knowledge, your learnings, and that’s something that leaders, I think, have a huge responsibility to show people the path to it, you know, like hope is willpower and waypower, you need to have both to have hope, and that’s the job of the leaders, you know, you are showing them the way, of course it comes from your own will, but you can show the way and really encourage people, even with this AI, like to explain to them what is AI fluency, help them to develop AI fluency, encourage them, encourage them to experiment and learn and constantly be in this iteration, their personal even iterations process, so yeah, I mean, I think it’s a lot about just looking at yourself and what you can do.

Thomas Kohler:
That’s a great final word, we’re already at time, thank you so much for the episode, it was really great having you and hope to see you soon.

Shlomit Gruman:
Yeah, thank you, thanks for the conversation.