The growing spread between the AI median and 75th percentile
AI comp pressure is concentrated in highly specialized roles like applied research and computer vision
Should AI engineering get a premium over software engineering pay?
It really depends on what you mean by AI engineering.
I’ve been surprised over the past few months to see a declining trend in AI compensation, both in absolute dollars and as a premium over SWE.
Seems different than a lot of what we’re hearing from comp leaders.
So last week I led Compa’s research team in a deep dive into AI offer data to understand what’s really happening in the market.
A couple big takeaways:
Comp varies widely depending on the type of AI work
The top of the market is spiking while the median continues its downward trend
Let’s dig into it.
1. Comp varies widely based on the type of AI work
Breathless headlines suggest that all AI roles are getting blank check mega offers.
Most are not.
Check out these premiums for base salary, using United States P4 as an equalizer across all jobs and sub-domains:
Top AI roles in applied research, autonomy, computer vision, etc. carry significant premiums over both AI engineering and software engineering. A similar analysis of TDC shows even bigger premiums, upwards of 30-50%.
Highly specialized skills in emerging technology, held by a small number of people in the world, command big market premiums. But more generic AI engineering work is becoming mainstream, and reflects a more modest (and declining) premium over software engineering pay.
2. The top of the market is spiking
The second big takeaway is that while the median AI engineering offer is coming down, the top of the market is spiking in Q1 this year. It’s particularly acute at more senior P5/P6 levels in stock comp:
This widening spread between 50th and 75th percentile illustrates some of the dislocation I’m hearing from comp leaders. “Crazy” exceptional offers are hitting their desks for some roles, while the majority of AI engineering offers carry a modest premium, increasingly institutionalized in separate or premium pay ranges.
AI is incredibly expensive, so consider this widening spread between the middle and top of market. Who are you competing against? Will the role work on an extraordinarily unusual technology challenge?
Define your pay strategy accordingly and focus investment where it counts the most.
Why offer data is a helpful lens into emerging jobs
Why would a comp team look at offer data?
The problem with emerging jobs is that, well, they’re emerging.
They don’t clearly translate into traditional employee-based market data by definition. Comp teams are debating the use of exceptions, premiums, and new ranges, and everything is happening in real-time — as a result, you just can’t see meaningful patterns in employee-based data for months if not years.
Offers, on the other hand, reveal emerging jobs that aren’t classified yet. In the case of AI, nearly 50% of offers are tucked into other job codes like software engineering and data science. But Compa’s matching technology looks at job titles and descriptions to peer through the noise.
Offers also represent pure market float — a point price in time, like stock market data. This makes the shape of the data uniquely conducive to studying trends, like the emergence of a new job.
We’ll keep an eye on this trend as the year progresses. Ping me with any questions!
Peer Group is a newsletter for comp leaders navigating market volatility in the new era of pay transparency! If you want to share my newsletter, you can forward this email to your colleagues and fellow comp leaders.
Want more polls and insights about today’s volatile talent markets?
Subscribe by hitting the button below.