People Decisions Under AI

Um there's mostly other questions is how can I enable the people that I have within my company? How can I help them making better decisions? And um more importantly, how can I build build that killer team that I can trust? Next up is Louis who is going to tell us a lot about how psychometric tests and behavioral analytics can be used in order to build a killer team as well as to enable your people to have better judgment in decision- making. Welcome Louis.

Thank you. Do I get Perfect. Thanks. There we go. Thanks so much for having me and uh thanks everyone for hopefully listening for the next 20 25 minutes or so.

Uh I'm a part of Atlas People and um what we do at Atlas People is that we combine psychometric assessments with relevant company context and AI analysis and we use this to optimize team to pick out the right people to join the team. So, it's about getting the right people on the bus, making sure that you play them in the right positions, and also helping making sure that that that everyone is playing to their advantage and that we get the right people on board. And um and this is what we're working with. So, I'm going to talk a little bit about how we use AI and people decisions and also a little bit about the market and the trends that we're seeing. But a few facts here uh for you guys that might make this seem all right.

But a few quick stats. Um 75% of all M&As and direct investment failures are driven by people and cultural issues rather than a flawed financial model. 46% of all new hires are a complete failure within 18 months. And the price of a single C level mishire is six to 10 times the annual salary. So getting these people decisions right is very important because it's very very uh expensive not doing so.

Uh also another theme that's uh I think a lot of wealthy families and and family offices are focusing a lot on is the succession. So uh 90% of wealthy families lose their fortunes by the end of the third generation. 25% of those failures are due to unprepared heirs inheriting leadership roles and 45% of family offices now site family member unpreparedness as a major acute operational risk. So what am I trying to say here? Well, listen carefully for the next 20 minutes or so because this is very important to you.

Every investment, every bet on a strategy is at the end of the day a people decision. So, quick show of hands. Who in the room has used AI for an important people decision within the last 12 months? There we go, Kevin. But it's quite few.

Uh, not a lot of people. I know that you guys have as well simple because you useless but um this is a a thing that I think in a year or so a lot more hands will be raised um because this is happening very quickly. Um but if we're looking at the market as of today, we see that AI is primarily used to automate manual processes that we already have. It's not changing the way that we're making decisions. It's rather optimizing the way that we do stuff.

So that could be CV screening and ranking, interview scheduling, performance reviews, compensation analysts, simple workforce analytics, and employee support but um but we believe that we can go way further than this. So this is from a recent publication from anthropic and uh the red area that we see is uh the observed AI coverage and the blue area is u the potential the capability in the market. So what I'm trying to say here is that we will see AI taking over a lot of the processes that we have in business and this will be the same for people decisions. But of course there is a gap and the biggest gap is in areas that require human judgment and people's decisions do that. So uh we shouldn't expect for the managers making the big decisions that our jobs will be completely taken over by by AI but AI will be informing the decisions and helping us making better decisions in the years to come.

So there are of course some some common risks when we you look at using AI and people decisions. Um some of the ones that we hear a lot is the built-in bias. If we train AIs on historic data and that historic data has discrimination in it. Well then we will see that in the models as well. A good example is Amazon.

They did a a CV um CV bot that could go through all of the applications. Unfortunately, that bot was not very fond of females. So, uh they ended up not hiring any females because they all kind of left in Topfunnel without a human being uh being involved. So, uh that was yeah, that was quite embarrassing. Um we have the blackbox problem.

If we don't understand what the AIS are doing, we don't understand the reasoning behind the choices that it makes or the scorings etc., uh then we can can be in trouble on that. We have the laws of human neurons. If we use it in topfunnel for example in recruiting and we haven't even met the person then we miss stuff like culture we miss stuff like uh the chemistry in the room uh and there's a lot of context that we do not get unless we feed it to the AIS and we don't necessarily have this context yet. So we might lose out on stuff if we use it only in top funnel. So we want to move it further down in the bottom of the column where we actually really make the decisions about who to hire or not rather than only automizing stuff in the top funnel.

Then we have workforce homogenization. Of course, if we're optim optimizing for a certain pattern, then we'll end up having a team of people who looks very much like each other. And we know that that is not how we build good strong teams. we need diversity and um especially if we're looking at innovation then it's a big problem if uh if we only have the same type of people and then of course there's the legal and compliance bit with the EU AI act and uh and emerging regulations uh of course there's a big liability if we let AI make decisions without having humans in the loop so a few best practices when we use AI uh for people decisions we need to define the criteria method and data input very carefully. Uh the AI is only as good as the input that we give to it and uh it's only as good as the questions that we ask.

So we we need to be very aware of this. Then we need to constrain the LLMs. LMS are basically made to answer all our questions. They are made to solve all our problems and sometimes they do so even though they shouldn't. Um so we need to be very aware of what we want them to do but also what we do not want them to do.

Then we need to build guardrails. We need to set very clear boundaries on scoring, waiting, interpretations and so forth when we use LLMs. And we always need to keep the human in the loop. So going back a little bit to to the first point of the presentation. This is very very important.

We need humans to define the criteria and the method. We need AI analysis to inform our decisions and then we need the the human beings to make the final call. And uh I'll give you an example of how this could look in re in in real life. So whenever we help a company in doing recruiting for example, we always do what we call a match report. So what we do is that we do a psychometric assessment of the candidate.

Then we add the relevant business context. So that is what does this role require? What do we want the person to do on an everyday basis and so forth. Then we test the manager and we test the rest of the team so that we can also match this psychometric profile of the candidate towards the team and the manager because that's very important for uh having high performance and having an engaged team and having people who actually have a good time when they go to work. Um we did this uh this recruiting for a company.

they were looking for a CFO and um they they had a candidate who looked very good on a lot of parameters but when it came to the bookkeeping and process discipline of doing the same work over and over again every day he scored quite low. Um he had everything else when it came to being a strategic CFO and he was also very good with people. He also had a care had a a kind of a a natural profile for negotiation ability and strategic thinking and stuff like that. So we had a conversation with the CEO instead of just discarding the candidate because oh there there was a low score or whatever it could be. We had a conversation with the CEO about what is the most important things to get in this role and then he had a conversation with the candidate and then they came to the conclusion that he should join the company but they should take the bookkeeping part and then outsource that.

So instead of just discarding a candidate they actually optimized the roles that it so it was the best possible fit for this candidate. And what he did just a few months after starting was calling all of the suppliers and renegotiating all of the deals and saving the company a lot of money. So they got a lot more but we needed the human judgment not just the AI. All right. So if we get all of this stuff right, if we don't fall into the traps, we believe that with this combination psychometric assessments that lets us look into natural human behavior and transform that into math and text.

If we take the relevant business context, if we have the right method and we run AI analysis, then we can really, really, really help guide all of the important people decisions that we are making. No matter if it's an investment into a company and trying to assess if it's the right team for this strategy or if it's recruiting or whatever people decisions it is, we believe that AI can help us in that with this combination. Um, I'll give you another example. So we're currently working with a team or with a company. They uh onboarded a new CEO and four months after onboarding this CEO, the board started seeing the signs that something was wrong.

They didn't know exactly what it was, but they knew something was wrong. So they hired us to find out what were the root root causes and what could they do from here. So we started with a foundation of a psychometric assessment of the CEO and the leadership team and then a quick 360 survey for the leadership team as well around the CEO and pretty quickly this was done in in a few days and then running the the AI um analysis pretty quickly we could see kind of okay what is the outline of what is wrong here. Then we added the strategy and KPIs and the role descriptions as well as some cultural research. And then we ran the analysis once more and it was very clear that we had structural issues in the company.

So unclear reporting lines, unclear roles, unclear decision areas. We had a leadership style mismatch. So a very different psychometric profile on the CEO compared to the rest of the leadership team and also a CEO coming from a different culture as well. So a lot of clashes on that and a lot of misunderstanding and then there was some regional fragmentation and there were people afraid of actually speaking up and voicing critical feedback and that was also down to history in the company as well. Um so what we did was running the analysis, giving the insights and then pointing towards what can we do from here.

So we did interviews after that and giving back all of the psychometric assessments, all of the data to the people in the team as well as the CEO. So they understood not that we have problems but why does these problems actually happen and what can we do about it? How can we make our collaboration better? And then the company also chose to redefine the organizational structure, the uh reporting lines as well as um as ownership. And uh what we saw four months after that was structural clarity and improved collaboration and morale.

The board and CEO gained awareness of the management team's fit for strategy and it turned potential CEO firing to building up around the CEO but now with the data and knowledge to actually build the team stronger and none of this would have been possible without AI. We couldn't have done it nowhere near we can assess one person and look into the psychometrics of one person and stuff like that. But the amount of data points in this data set, I think it was somewhere around 500 documents and so many data points in these psychometric assessments. There's no human being that could be anywhere near this kind of analysis. So AI is the game changer for us in this.

We would never be able to do it without. And it doesn't even stop here because now there is a data foundation and an AI engine that is built with all of these informations in it and that can keep that can keep helping us in making the right decisions as we move along. So every time something changes within the company that could be a new hire, a diff a role, a role shift, new strategy, whatever it is, we could put that into the AI engine, do another analysis, then it will highlight the risk factors. It will highlight how we can mitigate them and so forth. So instead of just having oneoff reports, we can now keep working with the company over and over again on the data foundation that's already been built.

And that is what we believe the future will look like. Predictive people intelligence that turns data into clear evidence-based recommendations. And how does this apply for family offices? Well, as I said in the start, every bet that you make is at the end of the day a people bet. Uh whether you're a new family office building up a team, you know maybe what roles you need to fill, but who are the right people for these roles and who are the right people for me?

How do we create the best possible culture and a place where we we are in alignment? You can help on that. Establish family offices. Say we're hiring for an in increasingly comp complex team. Um who are we going to support in what way?

How do we train the right people to move in the right directions and so forth can help on that succession planning as I touched upon upon earlier. And then of course people due diligence uh when we do direct investments. This is also something that we do quite a lot. And by combining the psychometrics, the strategy, the goals, etc. Then we can actually start really seeing a clear picture of where do we have our strong suits, where do we have our risks and how could we potentially mitigate them or at least what are the most important conversations to be had.

And where we see the market going, we see that um people data will become as fundamental to running a company as financial data. We see we will see more and more going from the admin automation to actual in decision intelligence. We will see in recruiting uh that we will probably see more and more that we go from kind of the topfunnel standardized stuff to the actual real people decision in bottomfunnel. the ones that really matter and the ones that are expected expensive to get wrong and then we will see a shift from the one-off reports to continuous predictive causal people intelligence. That is where we will see the the general market moving.

Uh and if we're looking at people due diligence, we expect it becoming more and more standardized kind of like financial and legal and commercial. Uh we will see predictive analytics on team level modeling for investors. We'll see investors moving from gut field to datadriven founder assessments and then portfoliowwide uh people intelligence across all holdings. That is kind of where we see the future and it's very close. So the decision that matters most still need our judgment.

We can't outsource it 100% to AI but we can definitely get a lot of help in getting these decisions right by using AI analysis. That's all for me. Thanks so much.