This probably comes as no surprise, but I’ve been thinking a lot about AI lately.

I think about and use AI on a daily basis, so that part isn’t new, but I’ve been thinking particularly about the cross-section of AI and those who are entering or yet to enter the workforce.

A couple of key pieces of data came out recently that genuinely make me wonder how anyone under 18 expects to pursue a traditional career path.

The first was a great new report from Ravio on the tech job market in 2025. The headlines are:

1. entry level jobs down 73%

2. AI job titles up 578%

The second was something I’m sure everybody saw blasted across LinkedIn, a study out from MIT that AI might be eroding critical thinking skills.

In sum, those that used ChatGPT for writing essays had the lowest brain engagement (compared to control group), got lazier over time — resorting to copy and pasting direct from AI — and goes on to suggest that usage of LLMs could actually harm learning, especially for younger users.

I’ll pause there. The paper is not yet peer reviewed and the test group was only 54 people, but yeesh. That’s a pretty confronting message even if only partially true. I know it actually made me take stock of my AI usage, and I’ve actually started forcing myself to do more things without AI support because it’s put the fear of god into me about not becoming over-reliant on it.

But what it flagged for me is that, for young people, they face a double edged sword.

On the one hand, entry level roles are diminishing (companies are hiring more mid- and senior-level people) and AI capability is more important than ever to having a career. On the other hand, they might be more susceptible to AI brain rot (my words) than more experienced people, and probably should be the ones least using it.

Something I’ll add, that many of you reading would no doubt know, is that AI is good for giving you about 60-80% of what you need. The rest of it relies on your skills and experience to discern its response into something that is actually useful.

So in a world where entry level jobs are reducing, companies require and assess AI competency, but where that same AI usage might be making you dumber (again, my words), what are young people supposed to do?

Maybe the cloud does have a little silver lining.

I actually think the same technology that is reducing opportunity for young people in a traditional sense is now creating it in another.

I’ve frequently been involved with emerging professionals in range of different capacities throughout my career. Whether at university hackathons or bringing interns into an organisation. One thing they don’t lack are ideas and enthusiasm. It’s infectious and impressive.

In many ways they benefit from not having had that creativity and lustre crushed out of them by a corporate role. Something their AI enabled peers probably don’t have in as much abundance.

It’s never been easier to convert that enthusiasm and creativity into a concept and then a product, using the very tools that took away their job opportunities.

Have you tried building something in Replit or Lovable? It’s never been easier, and we’re only limited by our imagination — something our younger generations have in spades.

So I think there is hope for our young people, but I don’t think it resides in a traditional career path.

It will be in the entrepreneur generation. The one’s that come up with new and unique ways to solve problems we could only dream of solving or maybe didn’t even know we had.

What do you think? Is my read right, or do you have a different take on where future generations will go? I’d love to hear it.

Matt

LOOKING TO GO FROM GOOD TO GREAT?

Startup People Summit

If you’re a Head of People, you have the hardest job in the company.

You’re across the broadest range of expertise — everything from hiring great talent to building high performance systems — All on a shoestring.

Often we do it in isolation. Unsure where to turn or who to ask so that you can avoid reinventing the wheel or even just speak to someone who is on the journey too.

We put others development before our own because we exist to enable others success.

This is the why of the Startup People Summit.

To showcase the incredible things we do in our roles.

To create a space to learn from others.

To meet others where we are, and avoid the isolation.

To feel seen, recognised and supported.

This is a space for you.

Choosing the right salary data is pivotal for any Head of People. This article guides you through defining needs, shortlisting vendors, and scoring options, ensuring you build defensible compensation practices that strengthen your credibility as a leader.

What Salary Data Should I Buy?

The scene is set. You’ve finally been given the go-ahead to upgrade your approach to compensation and buy a benchmark dataset.

Hoo-frickin-ray!

No more cobbling together salary bands from free benchmarks and god knows what other unreliable sources you’ve come across.

You’ve only been asking for this forever — but now is the time to actually buy the thing and you’re starting to wonder ‘which one is going to be the best’?

This might be the most pivotal decision you make when it comes to how you build your comp practices. 

Defending your comp bands is right up there with one of the most credibility making or breaking things as a Head of People.

I can’t recall just how many times I’ve been challenged by an exec with their one data point, compared to my benchmark with hundreds or thousands.

I even fondly recall a VP of Sales once telling me that the salary band I provided for the BDR role he was hiring in Sweden was ‘too low’ because — and I quote — “I’ve been there before and do you know how expensive a beer is? They’d never be able to buy a beer with that low a salary”.

😑 (some days you either laugh or you cry).

So it pays to get this part right, because believe me, you will be grilled on it.

Where to Start?

What we fundamentally want from an exercise like this is to buy a data source that will show us benchmarks for the skills/roles we hire, in the locations we hire them, across the kinds of companies we hire from.

To this end, you want to define the things you require in two buckets:

1. Those that are non-negotiables, and 

2. Those that are nice to have.

An example of your ‘non-negotiables’ might be:

• Data for United States and Canada

• 75%+ job family coverage

  • Including minimally: software developers, product management (...)

• Companies with a profile of:

  • X industries

  • X to Y headcount, and/or

  • X to Y capital raised/stage, and/or

  • X to Y revenue.

Some of your ‘nice to haves’ might be:

• It’s within your price range

• It has broader compensation elements (not just base salary, but also variable compensation, equity etc.)

Once we’ve defined this list of ‘needs’, our job is to find the dataset with the biggest overlap between those needs and their data. Expertly illustrated in the venn diagram below.

Now we know what we’re looking for, it’s time to compare it to the field of benchmark data suppliers out there.

Shortlisting Vendors

It’s always helpful to consider 2-3 benchmarks before making your final decision. If you’re in the technology industry and wondering which one’s to look at, I’ve already compiled a list of the major providers to consider (but many of them cover broader industries, too).

Check it out here.

From here, it’s time to start demo’ing the products. Be clear about your needs, and be insistent that they share with you the details of the countries, job families and companies they have data coverage for. Even if it's the nicest looking platform, it’s useless if the data isn’t effective.

Make sure to create a consistent way to score each vendor on each of your criteria. You’re welcome to use boring performance management language for this — you won’t offend anybody:

1. Exceeds — has more than we require

2. Meets — has what we require

3. Does not meet — has less than what we require

Once you’re satisfied by the dataset, make sure you consider other important aspects too. Functionality varies greatly between vendors and you’ll often find some that have features like:

• Automatic integration with your HRIS 

  • Which prevents you from having to submit manual spreadsheets of your data and reduced admin.

• Easier user experience

  • Some of the big players have systems that feel like they require a rocket science degree to use, but they often have deeper datasets. Consider carefully compared to younger players with glossy systems that are shallow on data.

• Many also provide features such as the ability to create bands from the market data, or to set target percentiles to compare your workforce (saves you bumbling around in a spreadsheet).

 Some of these might be a simple ‘yes/no’ vs a score, but you’ll get the gist of it.

Picking a Winner

Done well, this should be a fairly clear cut process for identifying the winner, which should be the one with the highest score.

But consider this. While the vendor with the highest score might be the obvious choice, if you’re budget conscious, don’t discount the one that meets (vs exceeds) if more suitably priced.

By running a process like this, not only can you be sure you’re getting what you need (and pay for), but it gives you a great audit trail for your decision making process.

No longer should startups be buying a dataset because their VC recommended something in their portfolio, or because it's the one somebody has heard of before. We’re buying something we can rely on to give us the data we need and make effective decisions.

Because at the end of the day — when you’re dealing with what is often the largest line item for your company — a process like this will ensure you’re finding a benchmark that builds your credibility as a commercial leader.