The biggest beneficiaries of AI coding improvements

The biggest beneficiaries of AI coding improvements might not be developers, but software consultancies serving corporates.

In the past three months, I would argue that the average productivity of engineers has tripled.

What used to be 80% manual coding and 20% autocomplete is now 80% prompting and 20% reviewing. If you are not a software engineer, it might be hard to understand how big this change is.

Almost every engineer I know is subscribed to Cursor, Claude Code, or Copilot, often paid by the company. Features that used to take me three hours now take fifteen minutes.

However, corporate expectations have not changed much. To most companies, software development is still seen as an expensive task that needs to be billed by man-hour and managed by a large project team.

In reality, software consultancies are already changing quietly. Project rates stay the same, but the time needed to deliver has dropped to half of what it used to be.

There is a big information gap (with limited time) between corporate expectation and real production speed. For software consultancies, and the profitability of each project just doubled.

When the same engineering team can now handle three times more work, the only limiting factor is the number of projects you can take.

For engineers, only the fittest will survive. Some will go deep into specialized fields like AI research or trading systems, while others will move closer to business roles such as sales.

Forward deployed engineer job postings grew 1,165% year-over-year in 2025, and GTM engineering postings have more than tripled since 2024 (TBH they are similar positions, but just different names).

This feels like 2021, when software companies were hoarding software engineers in anticipation of the post-COVID boom in IT services.

This time, the boom is in tech sales and deployment roles as consultancies race to capture the demand for expanded capacity.

I do not run a large software agency with hundreds of people, so I might not see the full picture. But from a startup standpoint, it really feels like we are riding the AI wave.