What skills do knowledge workers need in the Agentic Era?
Jared Chung
Founder at CareerVillage.org. #WorkforceDevelopment #AIinEducation #volunteerism
The bar is about to rise for entry-level knowledge workers in a big way. But there are several ways to talk about the new bar – which is the most effective framing to use?
Over the last several months in The Agentic Era, we've explored how autonomous, general-purpose AI agents are developing, and we’ve started to explore the implications they could have for reshaping the workforce. One of the most-requested subjects Agentic Era readers ask about is skills. To succeed in this new era, workers will need new skills—but there are a variety of perspectives on what those skills are. And as I’ve explored that topic over the past year, through talking to educators, employers, technologists, and our own non-technical staff exploring Agentic AI in their day-to-day work, I believe I’ve come to a bit of an outlier position on this skills question – one that personally gives me a bit of hope.
The emerging consensus
First, before I broach my slightly contrarian position, it’s worth acknowledging that a loose consensus seems to be forming that the ideal skill sets for thriving in the Agentic Era aren't about technical wizardry or deep AI expertise at all. Most of the people I’ve compared notes with have come to recommend a combination of two families of skills: (1) light technical skills (call it “AI savvy, not AI expert”) and (2) a lot of the core skills that have been in focus for the past decade: communication, problem solving, critical thinking, relationship building, emotion and empathy, project management, etc. Over the years we’ve called them soft skills, core skills, durable skills, SEL skills, and a few others. (there are some distinctions between these different terms I’m skipping over – my goal is not to conflate them, but rather to affirm that they are widely considered essential to the workforce of today and of the future). I share this perspective, although I’ve found that it doesn’t land well when we talk to employers who are already working on their company’s Agentic AI transformation.
But I believe I’ve found a reframing that does resonate with those employers.
A reframing both familiar and novel?
The most critical non-technical skills for managing AI agents are already well-known and widely taught: they're management skills.
Specifically, the essential skills for the Agentic Era involve vision-setting, goal definition, delegation, quality control, and strategic communication—skills traditionally associated with effective management. Crucially, these management skills don't replace but instead build upon essential foundational abilities like communication, critical thinking, strategic thinking, and entrepreneurship. These durable skills, recognized as essential and marketable for more than a decade, form a critical foundation upon which strong management capabilities can be developed. And they fit neatly into the way managers are already starting to talk about AI Agent oversight (see: FastCompany’s article “Who Manages AI Agents, IT or HR?”)
What Does Managing Agents Look Like in Practice?
Consider the workflow of a typical knowledge worker operating in an agent-driven workplace:
If these steps sound familiar, that's because they mirror precisely what good managers do every day. And it seems in my small-scale testing with employers that this branding around core skills is resonating well.
The Great News: We Already Teach These Skills
About one-fifth of American workers currently occupy managerial roles. That means vast educational infrastructure, corporate training programs, and management-development courses already exist to teach precisely these skills. Institutions from universities to workforce nonprofits already know how to equip people with management competencies. Additionally, apprenticeships and on-the-job training programs have proven effective in cultivating management capabilities, making these pathways especially valuable.
Unlike the more esoteric skills associated with the generative AI era (like "prompt engineering" or debugging hallucinations), Agentic AI skill sets are familiar territory. We already have models, methods, and even curricula that can be swiftly adapted for teaching management.
There’s still a technical skill set needed
Yes, there's a technical component—though it's relatively light. Workers will need basic literacy about AI agent capabilities, risks, guardrails, and compliance. And hands-on experience using AI and staying on top of the new developments as they emerge. But the technical core skills required aren’t cutting-edge computer science; they’re familiar, accessible, and teachable.
And subject matter expertise
Unless and until we no longer value human subject matter expertise (refer to: Machines of Loving Grace), any manager of AI agents will also need a good understanding of their functional subject area. Programmers managing coding agents. Marketers managing marketing agents. And so on. I still see credentialing and training playing a huge role in developing subject matter expertise for quite a long time, although the pressure for those skills to be acquired earlier in education and faster will likely increase.
The Hard Part: Scaling Management Education—Rapidly
If we buy that this framing is a good approach, it leads us to the real challenge: although management skills are known and teachable, they're traditionally reserved for a select few. The Agentic Era might require us to democratize and rapidly scale these management skills to nearly all knowledge workers (entry-level and experienced).
To achieve this:
What We Need Now: Re-Skilling
We recognize that "manager" isn't always the most popular job title. But I’m starting to believe that we must accept—and even embrace—that in the Agentic Era, management is the new entry-level. Imagine what it could look like if:
This skills gap may be more addressable than previous framings around Agentic Era skills, but the scale of the task is immense. And being a great manager is not easy, as all of us surely know.
What do you think?
Some of my own colleagues at CareerVillage don’t necessarily love this framing, and it might not resonate for anyone who has ever had a poorly-performing manager (who hasn’t!). The truth is that being a good manager is hard, and if we’re expecting every entry-level knowledge worker to be a good manager, that will take a lot of work. On the other hand, a skills framing that seems to resonate with employers is a valuable asset, and I wonder if we have an opportunity to reframe the discussion around how AI Agents are wielded to firmly center people through a management-skills-oriented framing.
Dear readers, I’d love to hear your thoughts!
🙏
Post-script 1: Entrepreneurship
It’s also worth mentioning that there will likely be a premium on entrepreneurship skills for quite some time. Entrepreneurship skills are a close corollary to management skills, and I would expect the returns to successful entrepreneurs building business in the Agentic AI era to be higher than in previous eras. And I would argue that acquiring entrepreneurial knowledge, competencies, and experience is likely to be well regarded by employers in the Agentic Era looking to hire managers for their AI agents.
Post-script 2: What skills do we “lose” in the process of this transformation?
What skills are discarded, or rendered useless?
A colleague asked me this question and I have to admit that I found it much harder to name knowledge work skills that will be rendered useless than it was to name skills that will be in high demand.
My uncle used to be a typesetter at a newspaper printing press in the East Bay. He lost his job when newspaper printing presses went fully digitized in the 80s/90s. But so did hundreds of other occupations back then. From a skills standpoint, I’m not exactly sure how to name the skills that were less important — surely the skill of physically putting type into a press, but that’s hardly the bulk of what he did. He lost his job because the industry was replaced, not because the bulk of his skills (teamwork, communication, ability to labor physically for hours a day, etc.) were obsolete.
How do I map that to the current AI automation phase? We’re losing entry-level jobs quite a lot these days. What are the specific skills that are not in demand? I think it’s hard to name those. My guess is that we lose a LOT of skills. A lot of small knowledge and competency areas. Huge amounts of legacy software will probably go away, so any skills based on operating those software platforms will be moot. You know how to use Photoshop v23? Moot after Adobe releases some new version of Photoshop with 100% AI-driven changes. AI will likely continue what Grammarly started, and further erode the utility of using calculators. But these are micro-level skills. I don’t think that the macro-level subject matter expertise (marketing best practices, financial planning, client service, etc.) will be devalued, even if the industries may get disrupted. Are those less-sought-skills worth mentioning in the article though?
As you can see, it’s harder for me at this point in time to reason confidently about the skills that will be less important than it is to reason confidently about the skills that will still be in high demand.
Post-script 3: A letter I shared recently with a school district about skills
For those of you who have made it down this far, and want to read further, check out this email I sent just this past week to Rochester Public Schools in my capacity as a member of their IT and Computer Science Steering Committee. I think it puts some specific real-world examples to the ideas I mentioned above…
Matthew and Steering Committee members,
I'm afraid that I will not be able to join you all on June 4th. … I wanted to take a moment to share that our perspective at CareerVillage.org is that the past few months have been very dynamic in AI, with some important albeit early emerging implications for K12 education. I must apologize for the length; If I had more time I would have written a shorter letter.
We're continuing to see a significant proliferation of AI tooling (we're referring to it internally as "the Cambrian Explosion phase"), but with a much clearer focus in recent months on AI applications for non-technical users. The half-life on noteworthy tools is extremely short (often just weeks from obscurity to notoriety and then back to obscurity), but some recurring themes seem to be sticking: steady energy pushing coding assistants forward (with honorable mention to the highly controversial "vibe coding" fad which had several of our non-technical staff members building websites, apps, and internal tools for the first time in their lives), steady energy pushing browser agents and OS agents forward (most notably tools like Browser-use.com, Manus.ai, ChatGPT Operator, Project Mariner), and a lot of energy pushing the maturation of monitoring and evaluation tooling. On the LLM front, we're mostly seeing a proliferation of model offers at different points on the cost-speed-size-quality spectra. To most end users, the difference between reasoning models, diffusion models, small models, open models, models doing deep research, etc. comes across simply as different tradeoffs between helpfulness and speed, with little interest in how they differ under the hood.
What does it mean for our K12 students at a time when technology is this dynamic (chaotic)? I believe there are few implications worth considering:
1. For all students (not just IT and Computer Science): Much has been said about the skills needed for knowledge workers to effectively wield AI. Suggesting that students invest in EQ is a common claim, but we have a different perspective: we conclude that employers will have a high demand for, and difficult time hiring, employees who can effectively manage the breadth of work autonomous systems will be doing. To put a fine point on it: managing AI is still managing, and being an effective manager is a valuable skill. Although 1 in 5 Americans is a manager of other people, I'm sure we all know our fair share of terrible managers. Great managers need great communication skills, problem solving, critical thinking, to evaluate the quality of work products, to understand the implications of work being done, and to understand what it means for the direction for the business. These are many of the core skills K12 has been working to build for years now, but framed in terms of what it takes to be a great manager. We sometimes talk internally about "MBAs for Kids" as a way of provoking what might need to change to prepare young people to become ready for the knowledge work of the near future.
2. Should students still learn to program? Although there's a lot of volatility in the software engineering job market, we're still not convinced that learning to program is any less valuable than it was a year or two ago. We just expect that the way software engineering or systems engineering works will continue to change with AI tooling. I think that means it's still good to learn Python or other programming languages. Where possible, embracing AI tools is probably wise.
3. Teaching students how to navigate this chaotic tech moment: With hundreds of new tools being launched every day, it seems futile to try to stay at the forefront of every tool release. That's true for everyone, myself included. But I think there's a teachable moment there for our young people: keeping track of how the tech is evolving is a valuable skill in itself. Allocating time in class, routinely, to practicing learning about the latest tools, trying them out quickly to understand what they're about, and then zooming out to reason about the themes and trends that we are seeing in how the technology is developing. I think building that muscle of how to navigate the tech change itself can be hugely valuable to them if the tech continues to change routinely.
4. For students interested in Agentic AI: Where students are ready to, building agentic AI demos can have short learning curves (once you have Python under your belt) and have really impressive results that can inspire learners to go deeper. We've found that using https://github.com/browser-use/browser-use/ specifically is a real jaw-dropper.
5. For students interested in IT Security: As the private sector adopts more digital automation using Agentic AI systems over the coming years, the security and oversight surface area they have to manage will continue to rapidly expand. The emerging consensus from CIOs I've spoken with is that the demand for IT security professionals is likely to remain strong and rising for quite some time.
6. And lastly: teaching students how to navigate the job market: Career navigation ("the skills to GET a job") is its own set of skills, completely independent from the skills needed to perform job duties ("the skills to DO a job"). With a more rapidly changing labor market, the need for better career navigation support increases. This is our bread-and-butter at CareerVillage, and we strongly recommend K12 leaders level up their career development supports district-wide (not just for IT or Computer Science learners).
Once again, my apologies for missing the upcoming meeting. I hope some of the ideas in this letter are helpful, alongside the hugely valuable perspectives from other Steering Committee members.
- Jared
Coach | Social Impact and Philanthropy Leader | Former School System Leader | Love & Justice at the Center
2moRemarkable newsletter. Just read the whole series. So many insights!