Your career ladder was written for a different jobYour career ladder was written for a different job

A career ladder is a promise about what a company will reward, and a promise only works when it describes the job people are doing today. Designer Fund's AI in Design 2026 report found that 73% of designers say expectations for output, quality, and speed are climbing while only 28% of leaders say their company has formally updated evaluation, compensation, or hiring, and that 50% of design leaders now weigh AI fluency more heavily when they hire while only 5% care less about execution quality. Put those together and the ladder you wrote two years ago asks a designer to carry a heavier job under paperwork written for a lighter one, and the leader holds the pen on that paperwork.

Keep the structure of the ladder and have someone who builds with an agent every day read every row of it, because the rows that need rewriting are the ones that no longer separate one level from the next.

Where the rows stopped working

Most ladders measure craft through the deliverable. The mid-level row talks about polish and completeness, the senior row talks about owning larger surfaces, and under the language sits an assumption that a lot of high-fidelity work means a lot of skilled hours, which was a fair stand-in for seniority when the designer who turned out a clean, complete flow quickly usually had better judgment too. A good agent setup breaks that link, because a working prototype takes an afternoon for a designer, a PM, or an engineer, and volume and fidelity can no longer tell you who is ready for the next level. What can tell you is what sat under the deliverable all along: deciding what is worth building, seeing what is wrong with something that looks right, and making the people around you faster. The stand-in has become cheap to produce, so a row built on it now rewards whatever is easiest to generate. I covered the older version of this problem in what a design career ladder actually measures, and the AI version arrives faster.

Two rows, reworded

The craft expectations stay in the ladder, and the change is a new layer on top of them that describes directing the tools: specifying what the agent should make and verifying what it returns. Each row should describe something a person does, since tools churn every quarter and a row that names a product is stale before the review cycle ends. Here is how I would reword two rows. A mid-level row that used to read "Delivers complete, polished designs for features of moderate scope with limited guidance" can read "Takes a feature from problem statement to working prototype, names the options they threw away and why, and runs a repeatable workflow with AI tools that a teammate can pick up." A senior row that used to read "Owns design for a large product area and mentors junior designers" can read "Decides what the product area builds next and what it leaves alone, reviews generated work at volume against the design system, and packages their own workflow so at least two teammates use it."

Every clause in the new wording is something a calibration committee can look for in a packet and something an interviewer can ask a candidate about, which is the test any row has to pass.

Subtract in the same pass

AI fluency stacked on top of execution quality, with nothing coming off the pile, makes the ladder heavier every review cycle. Every line you add needs a line you weigh less, and the decision should be made level by level and said out loud to the team. At mid-level, weigh the polish of any single deliverable less, since an agent produces a passable first pass, and weigh the count of flows owned less. At senior, weigh the fidelity of individual screens less and weigh judgment at volume and the development of other designers more. Execution craft stays in the ladder, because someone has to recognize when generated work is wrong and because only 5% of leaders in that survey say they care about it less, and what changes is how much weight it carries at each level.

Make judgment visible in calibration

Deliverables are easy to point at in a promotion discussion and judgment is not, so judgment has to be written down as it happens. I advocate for my reports in calibration, and the hardest part of that advocacy is evidence, so I ask people to keep a running note: the thing they decided not to build, the AI output they rejected and why, the review where they caught a problem that looked fine. Three or four of those in a packet move the conversation from how much someone shipped to how well they decided.

Move the interview loop with it

A ladder that says one thing while the interview loop screens for another gives you the worst of both. I screen for self-direction and comfort with ambiguity (AI-heavy work throws far more ambiguity at a designer than a fixed deliverable ever did), and I hire to fill a gap the team has. Whatever goes in the senior row should be something you can probe for in a conversation with a candidate, and if you can't write the question, the row belongs somewhere else.

Write it with the people it affects

The designers furthest along with these tools usually sit in the middle of your org, and they know better than you which behaviors separate a strong senior from an average one this year. I set KPIs together with my reports, and a ladder deserves the same treatment: give three of them the current rows and ask what is missing and what is dead weight. They write sharper language than a leader does alone, and they trust a document they helped write.