When a principal engineer at a 4,000-engineer organisation sits down on a Sunday in May and writes the most honest internal account of what AI has done to a senior role that has been published this year, it deserves close attention from anyone who employs senior knowledge workers.
Jamie Hurst's essay "Is this sustainable?", published on 24 May 2026, runs through what he describes as three years inside an AI-forward organisation as the principal engineer leading generative AI work in developer experience. It is not a productivity boast. It is not a vendor case study. It is the inside view of a senior individual contributor saying, calmly and specifically, that the role he is in now is not the role he was hired into, that it has expanded in ways that should not arithmetically fit into a working week, and that the parts of it which are giving way are the parts the organisation will miss most.
Your employees know the truth. Does your EVP? At Fathom we measure the "Credibility Gap" between your promise and their reality.
For HR, talent acquisition, and employer brand leaders, the piece is a leading indicator. The dynamics Hurst describes have already been documented in aggregate by UC Berkeley's Haas School of Business research on "workload creep" and by the Upwork Research Institute's finding that 77% of employees using AI report it has added to their workload. What Hurst supplies is the texture beneath the statistics: how a serious, motivated, well-positioned senior engineer actually experiences the trade.
Key points
- A principal engineer's first-person account, supported by independent research, suggests AI-forward organisations are quietly trading away the human-facing parts of senior roles to absorb expanding output expectations.
- Mentoring, 1-2-1 time, and unstructured thinking are the first things to be squeezed when AI compresses the cost of technical work but not the cost of organisational alignment.
- Senior roles are absorbing AI's productivity gains faster than junior ones, contradicting the common narrative that entry-level work is most exposed.
- The "AI specialist" label is currently an asset and may become a liability within eighteen months, mirroring the arc of earlier platform specialisms.
- Scope expansion of org-wide AI programmes is creating measurement gaps that HR and TA functions will be expected to help close.
The collapse of the gap between idea and demo
Hurst's first observation is the easiest to celebrate and the easiest to underestimate. Three years ago, he writes, a meaningful proposal followed a familiar choreography: write, align, build a small proof of concept, get a team assigned, ship six to twelve months later. Now, he and a colleague can build the thin proposal and a working PoC together, demo both within a couple of weeks, and use the demo to drive the conversation.
The benefit is real. The cost of building has collapsed. The cost of aligning organisationally has not.
This is the asymmetry that makes the rest of the essay land. When three different teams can each produce a working solution to the same problem in the time it used to take to write a proposal, the bottleneck moves from engineering to coordination. Hurst's specific example is merge request review, where homegrown bots have proliferated faster than the organisation can consolidate them. Building a new bot is now cheaper than adopting someone else's. Cohesion gets harder to achieve, not easier.
This pattern echoes what the Berkeley team described as task expansion: the work product accelerates, the work itself does not shrink, and the organisation finds new work for the freed capacity faster than it can consider whether it should.
Employer branding now has a measurement standard. The Talent Gravity Standard is a six-driver framework for quantifying employer attractiveness and the gap between brand promise and employee experience.
There is a second-order effect Hurst names that deserves close attention from people leaders: this shift advantages those who can build fast with AI tools and disadvantages those who cannot. Engineers who have adopted the tools effectively get heard more often, get their proposals taken seriously more often, and shape direction more than those who have not. He calls it "a skills redistribution happening inside every AI-forward org right now." Most organisations are not tracking it. Most internal mobility, succession planning, and performance management frameworks are not yet calibrated to it.
Where AI actually landed first
The dominant narrative about AI's impact on white-collar work has been that entry-level jobs are most exposed and that seniority is a hedge. Hurst's experience is closer to the opposite. AI, he argues, landed on senior roles earlier than it landed on junior ones, because senior engineers are the people positioned to recognise where AI can apply across the SDLC, write the proposals, navigate the organisation, and now also build the thing themselves.
The result, in his case, is that he codes more than he has in years. Three years ago he coded perhaps once a fortnight, mostly throwaway PoCs. Now he codes most days, with the disposability of those PoCs as the unlock. When prototypes are cheap, three approaches can be tested in the time one approach used to take.
At the same time, his strategic writing load went up. The week did not get longer.
Hurst is precise about what gave way. Two things, both invisible on a dashboard.
The first is mentoring. He has less time for 1-2-1s than he did three years ago. This was not an accident, he writes, it was a choice made under pressure. The 1-2-1 work does not benefit from AI tooling. It cannot be backloaded. It requires dedicated time and attention, and when other parts of the role expand to consume the available hours, mentoring is the first thing to go. He calls it "a real problem, both for me and for the engineers I should be developing."
The second is thinking time. The productivity gains from AI, he writes, got captured by output volume rather than output quality. Expectations rose to absorb the speed-up, and the slack that used to exist between tasks, the unstructured time where strategic thinking actually happens, got eaten first because it is invisible on a dashboard. Most of his thinking now happens on holiday.
This is the part of the essay that should land hardest in HR functions. The technical work got cheaper. The human work, the mentoring, the alignment, the problem definition, became disproportionately expensive relative to everything else, and therefore got squeezed.
Where the data agrees
Hurst's account is one engineer's experience, but it sits inside a research picture that has become unusually consistent.
The most rigorous recent study is an eight-month ethnographic project led by Associate Professor Aruna Ranganathan and PhD researcher Xinqi Maggie Ye at UC Berkeley's Haas School of Business, published in Harvard Business Review on 9 February 2026. Ranganathan spent two days a week embedded at a 200-person technology company for eight months, supplemented by tracking internal communications and conducting more than 40 interviews across engineering, product, design, research, and operations.
The research identified what the authors call "workload creep" through three specific mechanisms.
Task expansion: AI lowers the barrier to entry for complex technical tasks, so product managers begin writing code, user researchers begin taking on engineering tickets, and senior engineers, in Hurst's case, begin building what used to require a team. Job scope widens without any adjustment in formal expectations.
Blurred boundaries: AI prompting happens during lunch, on the commute, in the evenings. The work is voluntary, often framed as enjoyable experimentation, and therefore easy for leaders to overlook.
Implicit pressure: when colleagues visibly do more with AI, expectations rise for everyone.
A professional home for employer branding. The Institute of Employer Branding Professionals is being built and you can help shape it.
The pattern Hurst describes maps to all three. His scope has expanded, his hours have stretched, and the expectation set he is now working against was reset by the productivity demonstration he himself helped to create.
Independent data points in the same direction. Upwork's research found 71% of full-time employees report being burned out, 77% of employees using AI say the tools have added to their workload, and one in three say they will likely quit within six months. Subsequent reporting on the Upwork data identified a finding that should arrest any employer brand leader: workers who report the highest AI productivity gains are the most burned out, with an 88% burnout rate among the most productive AI users.
A March 2026 piece in CIO, drawing on the same Upwork research, captured the perception gap: 96% of C-suite executives expect AI tools to boost productivity, only 26% say they have proper AI training programmes in place, and only 13% say they have a well-implemented AI strategy. Hurst's account is what that gap looks like from inside a senior role: real productivity gains, real expansion of scope, no organisational apparatus to manage what happens next.
The mentoring problem in two directions
The mentoring squeeze Hurst describes is one half of a tightening loop that HR functions need to understand together.
The other half is the documented contraction of junior hiring. Indeed has reported a 60% drop in junior engineer listings over two years, and analysis from consultancy ARDURA places junior engineer hiring at a structural decline from 15% to 7% of total IT hiring in major markets. The economic logic ARDURA describes is straightforward: a junior engineer requires approximately 20-30% of a senior's time during the first months, and when companies model that against the perceived productivity of a senior with AI tools, the math often favours not hiring the junior at all.
The result is a system where the people best positioned to mentor have less time to do so, the organisation hires fewer of the people who would benefit from mentoring, and the supply of future seniors quietly thins out. Hurst describes one side of this. The hiring data describes the other. Neither side, on current trajectories, fixes the other.
For employer brand teams, this matters in two specific ways. First, the engineers most worth retaining are the ones most likely to be experiencing exactly the squeeze Hurst describes, and their externally visible signals of dissatisfaction (review sites, public posts, departures) are the metric the employer brand is judged on. Second, the talent pipelines that conventional employer brand work assumes will exist in three to five years are being throttled at the source by hiring decisions being made now.
The label that becomes a liability
The fourth observation in Hurst's essay is the one most likely to be missed by readers who stop at the productivity discussion. He raises what he calls the labelling problem.
He is now identified with one thing in a way he was not before. He notes that this is good for current visibility and career mobility, but that labels become liabilities when the landscape shifts. He points to the platform principals who got known as "the Docker person" or "the Kubernetes person" five years ago, and observes that they are now either grandfathered into their roles or carefully repositioning. He places himself, deliberately, at the asset stage of that arc, but flags that the liability stage exists.
Helping HR, talent acquisition, employer branding, and company culture professionals find careers worth smiling about.
This is a talent management problem that has not yet shown up in most workforce planning conversations. The senior specialists whom organisations are currently treating as their most valuable AI assets are the people most likely to be in the most precarious career position eighteen to thirty-six months from now, depending on how the tool landscape settles. The generalist platform knowledge they traded away to specialise is the thing that would have made the next pivot easier.
Whether the AI specialist label ages well depends on a question nobody can yet answer: how durable the current generation of AI engineering patterns proves to be. Hurst is betting on the organisational change experience, knowing how a large engineering organisation actually adopts new technology, who needs to be convinced, what governance constraints mean in practice. That is a defensible bet. It is not a certain one.
Scope expanded because the discipline expanded
The final observation in the essay deserves attention from anyone responsible for setting team-level metrics, OKRs, or performance frameworks.
Hurst's scope did not expand because he was promoted. It expanded because developer experience as a discipline expanded underneath him, and it expanded because AI made the operational issues the platform had been carrying intolerable. A weak local development setup is annoying when an engineer hits it. The same weakness becomes a serious bottleneck when an agent is trying to operate on the codebase. The fixes that serve developers now also serve agents, and agents scale in ways developers never could. The investment economics flip.
The uncomfortable bit, in Hurst's phrasing, is that this did not happen because DX got better at making the case for itself. It happened because AI exposed how much the platform was already holding back. The discipline got its moment through external forcing, not through advocacy.
The implication for HR business partners and talent leaders supporting AI-adjacent functions is sharp. The case for investment in those functions is no longer one that the function itself has to make from the bottom up. The case is being made by the technology, and the function is being asked to scale to absorb it. That changes hiring volume, hiring profile, role design, comp band placement, and reporting structure, often faster than the surrounding organisational machinery can keep up with.
Hurst flags one consequence directly: the measurement apparatus to demonstrate impact at board level for AI-led DX work does not yet exist. DORA was built for a previous era. Adoption metrics tell you what people use, not whether the organisation is better off. He and his peers are inventing the measurement framework while running the programme. That is uncomfortable but unavoidable, and it is precisely the kind of problem HR analytics functions will be drawn into helping solve, whether or not they currently see it coming.
What this means for HR, TA, and EB leaders
Hurst's essay is not a policy proposal and EBN is not in the business of converting individual reflections into prescriptive frameworks. What can reasonably be said is that the patterns he describes are appearing too consistently, across too much independent research, to ignore.
For HR leaders, the question is whether the performance management, capacity planning, and wellbeing infrastructure currently in place can detect workload creep at the senior individual contributor and senior manager levels. The early signal will not be missed deliverables. It will be the quiet erosion of mentoring time, the migration of strategic thinking into personal time, and the absorption of productivity gains into expanded scope without any conversation about whether the new scope is sustainable.
EBN
For talent acquisition leaders, the implication runs in two directions. The senior AI specialists currently in the market are extremely visible and extremely sought. The package required to retain them, three years in, may need to include things employer brand teams are not used to offering: protected thinking time, structural relief on meeting load, explicit permission to deprioritise. At the same time, the contraction of junior hiring deserves direct examination. The succession pipeline that current workforce plans assume is not being built at most organisations right now.
For employer brand teams, the most important consequence is the gap between the productivity narrative organisations are telling externally and the lived experience of the senior employees inside them. Hurst's essay is a polite, careful account of that gap from someone who is still, by his own description, finding the work rewarding. Less careful versions will appear on review sites, in alumni networks, and in the slow, attritional patterns of who leaves and who stays. The employer brand consequences of those will be visible in eighteen to thirty-six months.
Closing reflection
Hurst ends his essay with a sentence that is uncomfortable to read precisely because of the tone he writes it in. "The honest version of what it feels like, from inside it, is paddling harder to stay ahead of an ever-increasing current. It works for now. It won't last forever."
It is the calmness of the sentence that makes it useful. There is no crisis claim. There is no resignation announcement. There is no proposal for what should change. There is a senior, motivated, well-positioned principal engineer describing his own working life in a way that, if he is right, foreshadows what the senior populations of most AI-forward organisations are going to look like over the next two to three years.
There is a separate question, which Hurst does not answer and which HR leaders will need to, about whether the productivity story being told to boards and investors and shareholders matches the productivity story being lived by the people generating it. The Berkeley research suggests the gap is significant and growing. Hurst's account suggests the gap is closer to home than most executive teams currently believe.
The question that hangs over the piece, and over the wider research it sits inside, is whether organisations will reset expectations to reflect what the technology actually delivers, or whether they will let the gap close through the slower mechanism of people leaving and being replaced. The first option requires deliberate, uncomfortable conversations at the senior leadership level. The second requires nothing, which is what makes it the more likely outcome.
Takeaways and search questions
Is AI making senior engineers more productive or more burned out?
Both, according to consistent recent research. The Upwork Research Institute has documented that workers who report the highest AI productivity gains also report the highest burnout rates, at approximately 88% among the most productive AI users. The pattern Hurst describes is that productivity gains get captured by expanded scope and output volume rather than reduced hours, with the human-facing parts of the role paying the difference.
What is workload creep and where is it documented?
Workload creep is the term used by UC Berkeley researchers Aruna Ranganathan and Xinqi Maggie Ye in an eight-month ethnographic study published in Harvard Business Review in February 2026. It describes the progressive expansion of work demands beyond sustainable levels through three mechanisms: task expansion, blurred boundaries, and implicit pressure from colleagues' visible AI productivity.
Why is mentoring time falling in AI-forward organisations?
Because AI compresses the cost of technical work but not the cost of human-facing work. When senior roles expand to absorb more strategic, hands-on, and meeting-heavy work, the time that cannot be accelerated by AI tools, including 1-2-1s, mentoring, and unstructured thinking, is the first thing to be squeezed because it is invisible on dashboards.
Did AI affect senior or junior roles first?
The dominant narrative is that AI threatens entry-level roles first. Hurst's account, supported by patterns in the Berkeley research, suggests AI landed on senior roles earlier in AI-forward organisations because senior engineers are positioned to recognise where AI can apply, write the proposals, navigate the organisation, and now also build the solution themselves. Junior hiring has also fallen sharply, with Indeed reporting a 60% drop in junior engineer listings over two years, but that is a separate dynamic.
What is the AI specialist labelling problem?
The risk that being known as the AI person in an organisation is currently an asset for visibility and career mobility, but may become a liability if the tool landscape shifts and the specialist knowledge built up proves perishable. Hurst draws an explicit parallel to platform principals who became known as the Docker or Kubernetes person five years ago and are now either grandfathered in or carefully repositioning.
What should HR leaders look for as early signals of AI-driven burnout?
The early signals will not be missed deliverables. They will be the quiet erosion of mentoring time, the migration of strategic thinking into personal time, and the absorption of productivity gains into expanded scope without explicit conversations about sustainability. Standard engagement surveys may not detect this; targeted questions about thinking time, mentoring time, and after-hours AI use will.
How is AI changing the case for investment in developer experience and platform teams?
AI has changed the investment economics because the operational issues that were tolerable when humans were the only consumers of the platform stopped being tolerable when AI agents began magnifying them. The case for investment is now being made by the technology itself rather than by the function. This is shifting reporting structures, team sizes, and seniority profiles faster than the surrounding organisational machinery can adapt.
Will current AI productivity gains continue at the same pace?
The honest answer is unknown. Hurst is explicit that a significant portion of what he knows about applying AI in a software development context will be obsolete within eighteen months because the tools and patterns are still moving fast. The organisational learning, knowing how a large engineering organisation adopts new technology, is more likely to compound and remain valuable.
SOURCES
| # | Source | Publisher | Used for |
|---|---|---|---|
| 1 | Is this sustainable? | Jamie Hurst, May 2026 | Primary source for the principal engineer's account; all direct attributions and quoted observations. |
| 2 | Research: AI Doesn't Reduce Work, It Intensifies It | Harvard Business Review, Feb 2026 | Primary publication of the Ranganathan and Ye eight-month ethnographic study; workload creep framework. |
| 3 | AI Doesn't Reduce Work, It Intensifies It: What the HBR Study Means | AI First Founders, Feb 2026 | Berkeley study methodology details; Ranganathan two-day-per-week embedded research; 40+ interviews. |
| 4 | UC Berkeley Study Reveals AI Boosts Productivity But Increases Worker Burnout | Creati.ai, Feb 2026 | Three mechanisms of workload creep; task expansion examples (product managers writing code, researchers taking on engineering work). |
| 5 | AI makes the workday longer and more intense: HBR study | The Register, Feb 2026 | Workload creep voluntary-experimentation dynamic; cognitive strain framing; researcher direct quotes. |
| 6 | AI Promised to Save Time, Instead It's Created a New Kind of Burnout | Decrypt, Feb 2026 | Task expansion framing; engineer quote on doing "the same amount or even more"; Upwork 77% statistic cross-reference. |
| 7 | New Study Finds AI May Be Leading to "Workload Creep" in Tech | Interview Query, Mar 2026 | Implicit pressure mechanism; the AI productivity trap structural framing. |
| 8 | Upwork Study Finds Employee Workloads Rising Despite Increased C-Suite Investment in AI | Upwork Research Institute | 77% added-workload figure; 71% burnout rate; 96% C-suite productivity expectations; 1-in-3 likely to quit within six months. |
| 9 | Increased AI expectations without guidance leads to employee burnout | CIO, Apr 2026 | 26% AI training gap; 13% well-implemented strategy gap; 96% C-suite productivity expectations. |
| 10 | The Human Cost of 10x AI Productivity | Tech Trenches, Apr 2026 | 88% burnout rate among most productive AI users; senior engineer ownership-of-outcome framing. |
| 11 | The Developer Burnout Crisis: What 2026 Data Reveals | DevX, Jun 2026 | Tool churn and patience contraction; recovery practices framing. |
| 12 | When AI Replaces Junior Developers, Who Will Become Senior? | Medium, Aug 2025 | Indeed 60% drop in junior engineer listings; junior hiring contraction context. |
| 13 | Junior developer crisis 2026: why hiring dropped 50% | ARDURA Consulting, Dec 2025 | Junior share of IT hiring falling from 15% to 7%; 20-30% senior time cost of mentoring; the math behind hiring decisions. |
| 14 | 5 strategies for mentoring junior developers in the AI era | Engineering Enablement, 2025 | Critical thinking erosion framing; Microsoft GenAI critical thinking research reference. |
| 15 | "AI brain fry" is real and it's making workers more exhausted | Fortune, Mar 2026 | St Louis Fed 1.1% aggregate productivity figure; broader debate context. |
Unnamed CHRO
