AI is changing how work gets done. Most companies are still measuring and rewarding people for the work they used to do. Rewriting the career ladder starts with understanding what you can actually measure, and what you value.
Most companies need to rewrite the engineering career ladder for the AI era. Very few have started.
In late October, Amazon published a revised engineer role guideline that emphasizes outcomes rather than which tools an engineer uses. The principle makes sense: judge people on the value they deliver, not whether they wrote every line of code themselves.
But there’s a catch. That works when a company has decades of operating history, established reliability and support functions, and a lot of data on how its systems perform. It can focus on outcomes because it has built much of the infrastructure needed to observe them. Some companies we work with haven’t.
For a company with 40, 100 or 400 people, the question is more immediate: what should we actually be measuring, and are we ready to measure it?
Rewrite Your Own Engineering Career Ladder
A reference point is not a template. A smaller company can’t simply announce that it now rewards outcomes rather than output and expect the organization to follow. It first needs the ability to see those outcomes.
What follows comes from our assessment and coaching work, not from a published survey. These are patterns we see repeatedly across companies at different stages of growth, and they point to two very different starting positions.
Most Ladders Still Describe the Old Job
In company after company, the engineering career ladder, role definitions and listed responsibilities still describe the pre-agentic job: writing code, owning a component, managing a team of people, and delivering against an assigned backlog.
Nobody necessarily decided to keep them that way. They simply haven’t gotten around to rewriting them. Meanwhile, the work itself is changing. Engineers are specifying work for agents, reviewing generated output, orchestrating multiple tools and making decisions about what should be built in the first place.
The result is a growing disconnect. People are being evaluated against a description of work that is changing underneath them. And managers are increasingly unsure what good performance, particularly at senior levels, actually looks like.
Two Camps, Two Starting Points
In our experience, companies tend to fall into two broad camps. Some have moved toward an outcomes-based model, where the focus is on what customers and the business actually experience. Many others still rely primarily on measures of activity and velocity: tickets closed, pull requests merged, story points completed, and now AI-assisted output.
Neither camp is necessarily careless. They simply have different starting points, and each needs a different first move to rewrite the engineering career ladder.
Velocity-Driven
What gets measured: Tickets, pull requests, story points, features delivered and AI-assisted output.
Where we usually see it: Earlier-stage companies with limited operational, reliability and support infrastructure.
The risk: Fast delivery gets rewarded even when the work creates downstream problems.
First move: Build the visibility and operational discipline needed to understand what happens after release.
Outcomes-Driven
What gets measured: Customer experience, business results, reliability, quality and operational performance.
Where we usually see it: Companies with established operations, reliability, support and quality functions.
The risk: Results that are difficult to attribute can reward the best storyteller rather than the strongest contributor.
First move: Tighten how outcomes are defined, measured and credited.
The Run Organization Determines What’s Measurable
The difference between these camps often comes down to the maturity of the run organization. By that, we mean the people, processes and systems responsible for operating what the company builds: reliability, support, quality, incident management and ongoing service delivery.
Where those functions are thin, velocity tends to become the default measure. Tickets get closed, features ship and releases happen. But what happens afterward? Did customers actually benefit? Was reliability better or worse? Did the new feature create more support work? And did the team spend more time maintaining what it just built?
If nobody is measuring those things, the organization has little basis for judging outcomes. Speed becomes the easiest thing to see, and therefore the easiest thing to reward.
This is where copying that approach can go wrong. A company without the operational visibility to measure outcomes that announces it now rewards results may simply end up rewarding whoever tells the most convincing story.
For these companies, the first step isn’t necessarily to rewrite the engineering career ladder. It’s to build the operational visibility that makes an outcomes-based model possible. That’s an executive team decision, not just a tooling or engineering initiative.
What Changes When Agents Do More of the Work?
The shift to agentic work changes more than how quickly a team can produce software. It changes where human contribution and judgment create value.
When agents can generate code, write tests, create documentation and execute increasingly complex tasks, producing output becomes less of a differentiator. The value shifts toward deciding what needs to be done, specifying it clearly, evaluating the results and understanding the consequences.
This doesn’t mean technical expertise or writing code no longer matter. They absolutely do. But writing code is no longer sufficient evidence of seniority or contribution. The ability to exercise judgment about the code, the problem and the outcome becomes increasingly important. Any effort to rewrite the engineering career ladder has to reflect that.
What a Rewritten Engineering Career Ladder Should Reward
Our working view, which we’re continuing to test with clients, is that anyone who sets out to rewrite the engineering career ladder should recognize a different set of contributions. Five areas stand out:
- Specifying work. Defining problems, breaking them down and providing enough context and direction for a person or an agent to act effectively. The quality of the specification increasingly determines the quality of the result.
- Verifying and reviewing output. Knowing what good looks like, challenging assumptions, identifying errors and understanding when an agent’s output is wrong. More output makes the ability to evaluate it more important, not less.
- Exercising judgment. Deciding what to build, what not to build, when to stop and where human intervention is necessary. Knowing which problems are worth solving is often more valuable than solving more of them.
- Orchestrating and mentoring. Helping people and agents work together effectively, sharing context, developing others’ judgment and improving the capability of the wider team rather than simply managing individual execution.
- Delivering measurable outcomes. Connecting work to customer and business results, and taking accountability for what happens after something ships.
These aren’t replacements for traditional engineering skills. They’re a broader definition of contribution. Technical depth, architecture, problem-solving and execution still matter. But the way those capabilities translate into organizational value is changing.
Consider two people. A senior engineer who uses AI sparingly but consistently makes excellent architectural decisions, catches problems before they reach production and delivers strong customer outcomes. And a junior engineer who uses AI constantly, generates large amounts of code and closes tickets at an impressive rate.
Which one does your engineering career ladder reward? If it doesn’t have a clear answer, it may still be measuring activity rather than contribution.
It’s Not Only an Engineering Problem
Engineering is where the strain often shows first, so it gets most of the attention. But the same shift is happening across the organization.
Product managers are using agents to research markets, analyze feedback and develop specifications. Designers are generating and evaluating more alternatives. Marketing teams are producing and testing content at a different scale. Customer success, finance and operations are using agents to automate work that previously required significant human effort.
In each case, people are increasingly responsible for specifying work, supervising agents, evaluating output and exercising judgment about what gets released or acted upon. Almost none of their existing career ladders explicitly recognize those responsibilities.
There’s also a management implication. If agents take on more execution, what exactly are managers and senior individual contributors accountable for? Managing more people? Producing more output? Or improving the quality of decisions and the results the organization delivers?
If engineering changes its career ladder but product, design and the rest of the organization don’t, the company creates a new mismatch. Leaders will be working toward different definitions of performance, and managers will continue to be evaluated against expectations that no longer reflect how work gets done.
That’s why we believe the effort to rewrite the engineering career ladder should be coordinated with every other function, not run as an isolated HR or engineering project. Career ladders, performance management, management structures and promotion criteria all need to evolve together.
Who Owns the Rewrite?
HR should run the process to rewrite the career ladder, but the executive team needs to own the decisions. Someone has to be willing to define what good performance and seniority actually mean in an organization where agents are doing an increasing share of the execution.
For engineering and product, that means a joint CTO and CPO conversation. Otherwise, the two functions risk developing different expectations about ownership, accountability and what constitutes meaningful contribution. The same alignment needs to happen across other departments.
Timing matters. We expect 2027 performance and promotion cycles to expose the problem in many organizations. People will be evaluated against criteria that don’t reflect the work they’re actually doing. Senior contributors may feel their judgment and leverage are invisible, while managers struggle to explain why one person’s contribution is more valuable than another’s.
We also expect boards and investors to start treating the velocity-versus-outcomes question as a signal of operating maturity during diligence. That’s our prediction, not a finding, but it’s consistent with what we’re seeing in assessments today.
Questions to Sit With
Before rewriting your engineering career ladder, it’s worth asking a few questions about how your organization currently operates:
- Which camp are you in: outcomes-driven or primarily velocity-driven? And how do you know?
- If work shipped quickly but performed badly in production, who in your organization would notice?
- Can you distinguish between someone who generates a lot of output and someone who creates a lot of value?
- When did you last revisit the career ladder for product, design, support or data, rather than just engineering?
- Who can explain what “senior” or “staff” means today without describing a job that existed a few years ago?
- When did you last promote someone for catching a problem, challenging a decision, mentoring others or stopping work that shouldn’t have been done?
A Final Thought
For years, we’ve promoted engineers for the amount of work they could deliver, the complexity of the systems they built and the teams they managed. Those things still matter. But in an agentic organization, the person who knows what shouldn’t be built, catches what shouldn’t ship or helps ten people get better results through agents may create more value than the person producing the most code.
A career ladder is one of the clearest statements a company makes about what it values. If it still rewards volume of authored work above judgment, leverage and outcomes, it’s quietly paying people for the wrong things. It’s also why we wrote about tokenmaxxing as a vanity metric.
The goal isn’t to copy anyone or rush to define a new set of AI competencies. It’s to understand where your organization is today, build the ability to measure what matters and then change how you recognize and reward contribution.
As former operators, we’ve built career ladders and run the organizations they shape. Our Product and Engineering Assessment is one of the places where we most often see the disconnect between how organizations work and how they evaluate people. If you’re trying to work out whether to rewrite the engineering career ladder now or build the groundwork first, we’re glad to talk it through.
Is your career ladder describing the job people do now?
Book a 30-minute conversation to talk through where you stand and what a ladder review would involve.