By Julianna Vitolo (subscribe to her Substack here for more)
The role of human capital in society is changing, and the implications are particularly significant for knowledge workers. Today, AI primarily functions as a productivity multiplier: copilots supplement human work and render traditional processes more efficient, allowing individual knowledge workers to accomplish more in less time. Due to rapid advances in underlying LLMs, these copilots will soon become fully autonomous agents, tackling entire workflows without intervention. This will make the human-in-the-loop model we see today much less compelling from an economics perspective, and will likely lead to total replacement in many cases. Today’s knowledge worker jobs will be fundamentally disrupted, and the traditional playbook for building and scaling businesses will be rewritten.
The Great Workforce Disruption
The economics of a partially agentic workforce aren't just compelling – they're irresistible. Anyone with a basic understanding of economics could tell you that the production equation consists of two sides: inputs and outputs. Traditionally, the more inputs you add together, the greater the resulting output. AI is now generating an unprecedented level of output per worker, rendering the standard headcount-based growth model (where more heads = more productivity) obsolete. Tomorrow's startups won't need armies of engineers to scale – they'll go further, faster, with small, elite teams supercharged by AI.
Large enterprises are also adapting to this new reality. Shopify's CEO, Tobi Lütke, articulated this best when he directed employees to leverage AI to expand their individual capabilities before requesting additional headcount. Fiverr's CEO, Micha Kaufman, has similarly endorsed AI-enabled productivity enhancement over workforce expansion. Klarna’s CEO, Sebastian Siemiatkowski, formally paused hiring for new roles in 2023, encouraging employees to adopt AI solutions across the organization, and saving about $10 million annually by using AI for marketing needs, reducing in-house lawyer time, and optimizing communications roles. These anecdotes represent a larger trend – from small ventures to large enterprises, companies are undergoing a fundamental shift in how they aim to achieve growth.
The result is somewhat paradoxical and raises profound questions about the future job market:
In a world with surplus talent but limited openings, who gets hired?
How will we define career progression when traditional entry-level positions disappear?
Will this accelerate the rise of specialized boutique firms over venture-backed behemoths?
Hiring Calculus in the AI Era
Due to these productivity gains, the formula for hiring decisions is already changing dramatically. Consider software development: Why hire many junior developers when fewer senior developers with AI assistance can deliver exponentially more value?
As "vibe coding" (programming with the assistance of AI) increasingly becomes standard practice, the demand for junior developers has plummeted. Companies prefer experienced architects who can provide strategic vision and product design expertise – areas where AI still lacks sophistication. One senior developer with AI can drive a more effective output than several junior developers at roughly the same cost. Firms can also save on the reduced management overhead of smaller, more experienced teams in the long run.
Engineering is an area where this shift in mentality will be most immediate and acute, given that software developers are among the highest early adopters of copilot platforms like Cursor, Codeium, Claude, and GitHub Copilot. In Anthropic’s Economic Index (published February 2025), “computer and mathematical” roles, which are largely software engineering positions, demonstrated the greatest early adoption of Claude. The majority of the queries in this category related to “tasks like software modification, code debugging, and network troubleshooting.”
Source: Anthropic Economic Index
Ultimately, the impact will be economy-wide, extending far beyond developers. Some examples of tasks within each field that can be automated with AI include those that formerly required substantial time to filter through large corpora of data, or are considered mundane and routine (we’ll explore each of these in more depth in future blog posts):
Law
Document review and analysis, case law and precedent research, contract drafting, and due diligence automation
Finance
Investment research, risk assessment, generation of reports and presentations, regulatory compliance, and financial modeling
Consulting
Market research, preparation of client proposals, competitive analysis, and identification of inefficiencies in business processes
Product Management
Customer feedback analysis, PRDs, competitive product analyses, product roadmap prioritization, and analysis of A/B tests
Healthcare
Clinical documentation, medical research and literature reviews, automation of administrative tasks like insurance correspondence and documentation, and clinical decision support to assist in diagnoses
Education
Personalization of learning content, assessment and curriculum generation, automation of student feedback and grading, and educational research
Domain Knowledge as the Ultimate Differentiator
When everyone has access to the same AI tools and the barrier to entry for technical roles has been lowered or dismantled, how can you differentiate between candidates?
In some areas, like vertical software, domain expertise will become the critical competitive advantage. Specialized knowledge of and experience in industries like healthcare, finance, logistics, or manufacturing can't be easily replicated by generalist AI systems, and there is real value placed on understanding the nuances of a given vertical – especially in this moment, when much of the value vertical-focused startups are seeking from employees is training data to develop future efficiencies that can help scale revenue without also scaling headcount and costs. This shift favors candidates with deep industry experience who can apply AI within industry-specific contexts to solve problems.
It’s important to note, that such domain experience is not the type gleaned from a book or advanced study, but instead is found in real-world experience. The power and marketability of advanced degrees have waned over time. It used to take years to garner specific knowledge of a discipline, and the acquisition of that knowledge occurred through formal, structured, and intensive study. With deep research LLM capabilities, those advanced degrees are no longer as potent. An LLM can synthesize information akin to a PhD researcher, so the need to hire someone with a specific academic background has faded.
Employees immersed in a specific industry deeply understand its features and issues, and possess a perspective that has yet to be replicated by foundational models. In the software engineering example, experience is what allows employees to develop taste and systems-level thinking, moving beyond the pure code mechanics. This is increasingly valuable as the barrier to ship viable code has disappeared.
From Volume to Value: The New Recruiting Paradigm
Recruiting will undergo its own transformation, pivoting from high-volume talent acquisition to identifying those rare individuals who combine technical proficiency with creative problem-solving and strategic thinking. AI companies capable of identifying, contacting, and procuring these individuals for companies with openings for these traditionally “hard-to-hire” specialized roles will accumulate significant value.
Traditional skills-based assessments for technical roles like coding, investing, product management, etc. will move away from the model of real-time problem solving, to take-home, open-book style assessments where candidates will have access to any tool of their choice. The mindset shift here is that by removing all constraints and allowing interviewees to leverage any tools of their choosing, the structure of the interview will also change. Rather than testing memorization and preparedness, the resulting interview questions will gauge higher-level, strategic thinking. Questions will shift away from answering “what” you did, to “why” you did it. LLMs can do the “what,” so human employees will be responsible for determining the “why.”
We’ve already seen that the previous standards for evaluating engineering proficiency are now outdated, outsmarted by AI solutions. Interview Coder recently achieved notoriety for allowing students to cheat on live, web-based Leetcode coding interviews using their undetectable AI assistant platform. In this case, the tool not only codes for the candidate but also provides the rationale and step-by-step thinking for each line written. Applications like this are forcing hiring managers to rethink the format of interviews within the coding world and should be viewed as a sign of what’s to come for other roles as well.
The recruiting function itself won't be immune to the AI transformation, either. Truly effective AI recruiting systems will need to be multimodal – combining voice, chat, and video capabilities to create candidate experiences that rival or exceed human interactions. These systems must master the nuances of candidate assessment while maintaining the human connection that remains essential to employer branding.
Internal Structures and What it Means to “Manage”
Another possible effect of this dramatic change in work will be a new concept of what it means to manage a team. Management will no longer be reserved for seasoned employees with tenure; rather, all employees will become managers. Instead of managing people, most employees will manage fleets of agents. Human oversight will be required to review work, but not to complete it.
Strategic Implications for Founders
An agentic workforce will alter the way talent is recruited and managed, and will drastically change the HR function. For founders building AI-powered HR, recruiting, and workforce management tools, this shift presents an enormous opportunity to build solutions tailored to this new labor dynamic. Prospective ideas include:
Assessment platforms that evaluate candidates specifically on creative problem-solving capabilities and strategic thinking, areas where humans still outperform AI
For engineers, give take-homes and allow for unlimited AI access rather than live coding interviews. Interview case studies should test strategic, high-level thinking skills: the logic behind certain decisions and the overarching plan for development going forward
AI recruiters that focus on quality rather than quantity, helping companies identify candidates who will deliver outsized value. Tools that validate domain expertise and unique combinations of skills rather than just technical prowess
Ensure that these recruiting tools are multi-modal, leveraging the combined power of voice, text, and video to not only source and select the most qualified candidates for the job, but also ensure that the candidate experience is seamless and painless. Candidates should go through the process without realizing they’re missing out on the human touch and relationship element that makes current recruiters so effective
Workforce optimization systems that help companies maximize employee potential with AI assistance
Multimodal recruiting experiences that combine the efficiency of AI with the emotional intelligence of human interaction
The most successful founders won't just automate existing HR processes – they'll reimagine the entire talent ecosystem for an era where AI and human capabilities are deeply intertwined. The winners will be those who create tools that amplify uniquely human strengths while leveraging AI to eliminate mundane tasks.
This new world doesn’t just necessitate a change in how we hire, but a complete redefinition of what it means to work.
