The Job Market Isn't Dead, But it Seems Far Pickier These Days
The Weekend Windup #26 - Reflections, Cool Reads, Events, and More
Let me start by saying that I haven’t had a full-time job (W-2 in the US) for almost 10 years. This is largely by choice, as I realized back then I needed to do my own thing. At this point, I half-joke that I’m unemployable. That said, people are very nervous about the job market, and I get asked a lot for career advice. Here’s my attempt to answer them, with the caveat that when it comes to the current job market, I can share the stats I’m seeing and the anecdotes I’m hearing. What I can’t give you is a firsthand account of what it’s like to be in it right now. That’s why I rely on you. Let’s get into it.
The Market Hasn’t Disappeared (Yet)
The data job market is still there. It just got a lot narrower, more selective, and more production-oriented/revenue-facing. And honestly? I think the market was probably overly saturated for the last decade or so. We hired a lot of people with “data X” (scientist, engineer, analyst, etc) on their resume and a pulse, and called it a day. Those days are very much over.
Indeed Hiring Lab reported in January 2026 that hiring activity remains subdued, but nearly 45% of data and analytics postings now contain AI-related terms. That’s the clearest signal that demand shifted, not vanished (yet). If you’re still doing generic analytics, SQL plus dashboards, and moving data from point A to B, you may be insulated by your industry or your team right now. But if something happens and you go looking for work, you’re at a big disadvantage if you haven’t incorporated AI into your workflows. That’s just the reality of what the stats are showing.
Dice’s February 2026 tech jobs report showed U.S. tech postings up 12% month over month in January, but still 3% below January 2025 levels. And that same report showed that AI skill requirements appeared in 58% of U.S. tech postings, up from 51% in December. The market is telling you “AI, AI, AI” all day. I don’t know what else to say at this point. Go pick up those skills.
On Layoffs: Read Between The Lines
Layoffs are real. Snowflake announced layoffs this week. Dell too. It doesn’t feel like a week goes by without hearing about more cuts.
Reuters reported on March 19 that Challenger, Gray & Christmas linked AI to about 7% of planned U.S. layoffs in January. Goldman Sachs economists estimated that AI was responsible for 5,000 to 10,000 net job losses per month last year in the most exposed industries.
But I’m still skeptical of some of these headlines. I’m skeptical of Block’s announcement that they’re cutting 40% of their workforce due to AI. I don’t think AI is that good yet. But…it might get there before we’re ready.
What I do think is happening is that executives are anticipating how good AI will get, and they’re adjusting now. They’re skating to where the puck is going. If you’re waiting for the puck to arrive in your career or in your company, you’re probably already too late.
Something clicked for my friends and me around late November or December of last year. The agents and the models (specifically OpenAI and Claude) suddenly got really good. Like, game-changing good. Tasks that were once too difficult for these models were suddenly completed rapidly.
If you extrapolate where these models are likely going, I certainly wouldn’t bet against them (they might hit a wall, but we’ll see when we get there). I expect the tasks we do will become increasingly easy for these agents to finish. But as I’m finding and have written about many times, as AI knocks out my task list, it’s not as if I’ve discovered a new pile of free time. Quite the opposite. For every task I knock out, there seems to be ten more that magically appear.
I can’t say this is a universal thing, since the surface area of tasks one can automate and replace with new tasks highly depends on the work one does, and on whether one is employed and on one’s organization’s willingness to flexibly adapt to new levels of abstraction and possibilities. But in general, AI doesn’t mean work disappears. The type of work shifts to more abstract levels. It also means employers aren’t hiring for yesterday’s data roles anymore. The smart ones are anticipating where work moves, the tasks involved, and if that’s done by a human, a machine, or both.
In data orgs specifically, what’s getting cut is pretty consistent: cost-center work, low-context/low-value reporting work, duplicated analyst roles, disconnected research that never touches a decision, narrow ETL pipelines that a managed service/vibe-coded solution could replace tomorrow. If your work isn’t close to production, decisions, or money, you’re inevitably cooked.
A Thought On The Data Disciplines
Let’s take a quick look at my thoughts on where the various data disciplines (engineering, analytics, science) stand today. The bottom line is that the low end will get commoditized quickly.
Data engineering is resilient since data is the lifeblood for AI, but AI-assisted coding and managed services have reduced the premium on routine pipeline assembly. Moving data from point A to B is quaint, but not even the bare minimum. The premium now includes platform engineering, data and semantic modeling, governance, observability, performance, and support for ML and agent workloads. These will very soon be table stakes.
Here’s the big mental shift you need to come to terms with: you’re not just building pipelines and data for humans anymore. You’re building for bots, too. Understanding what AI-first infrastructure and data models look like, since that’s where everything is going. It’s super early right now, but at the rate things are changing, you need to be on top of this ASAP.
Analytics is under the most pressure in its generic form. Classically, you’d build reports, a dashboard, and maybe offer “insights” (whatever that means). That’s disappearing faster than the Great Salt Lake, near where I live. Employers aren’t saying they need fewer decisions. They’re saying they need fewer report jockeys and more people who can build decision systems and work with the business to produce value (i.e., money). Think metrics design, experimentation, semantic layers, executive and manager decision support, which aren’t going away, and are only getting more valuable.
Data science is a weird one. BLS projects 34% employment growth from 2024 to 2034, about 23,400 openings per year. I’m curious what that number would look like if updated today, but long-run demand looks strong. Paradoxically, short-run entry for juniors is harder. The market favors applied data scientists who can influence product, revenue, risk, and operations, especially those who can actually ship. Many data scientists have fallen into the notebook trap over the last decade. Notebook-first data science is weak. Production-facing data science is where it’s at.
Some people will call me crass for saying this, but the bottom line is that you need to operate where decisions and money are made. I might be crass, but I’m also realistic.
If You’re Working in Data, Here’s Your Action Plan
Here’s a quick-and-dirty, common-sense action plan that includes five things that are both real and not novel. I’ve ranted about these in various places, but here they are in one place.
Get closer to production. Ship things. Own pipelines. Own metrics. Own experiments. Own monitoring. Own cost and performance. Bottom line: own shit. If your name isn’t on anything running in production, that’s a problem.
Get closer to money. Revenue, margin, risk, fraud, retention, conversion, operations, and executive decisions. The farther your work is from a number someone cares about, the easier it is to cut. I don’t know how to put it any other way.
Become a hybrid on purpose. Data engineers should understand semantics, governance, AI workloads, and software engineering. Analysts should understand experimentation, metrics design, and business operations. Data scientists should understand software engineering, deployment, and measurement. I called this the Great Convergence years ago (before the AI wave), and it’s happening. The siloed specialist days are numbered.
Use AI as leverage, not branding. You don’t need to become an AI thought leader or an influencer on LinkedIn (please don’t). You need to show that AI makes you faster, broader, and more effective in your actual work. That’s what 45% of job postings are really asking for.
Build proof of work and value. Not “I know Python and SQL.” Instead, I reduced warehouse spend 28%. I rebuilt the semantic layer so AI agents operate better according to metrics X, Y, Z. I productionized model monitoring. I cut dashboard sprawl in half. I fixed data quality issues that were corrupting financial numbers. I increased revenue or profits by $X or Y%. Especially as roles and tasks are thrown up in the air, I expect titles will matter less (what is a “senior” in this new world where things change every week?), and being able to show proof of work and value will matter far more.
Shit WILL Get Real, So Have A Plan B
You can do everything I just said and still get laid off. Or still have not found a job. So let’s talk about what else you can do.
This is partly inspired by Claire Sullivan’s lunch-and-learn earlier this week, where she gave a genuinely great clinic on solo entrepreneurship and the realities of going out on your own. Highly recommend.
My take is you need a sharp value prop, a painful budgeted problem you know how to solve, and the willingness to go talk to people about it. Start with services, not products. SaaS is a bad idea for most people starting out, even before the current SaaSpocalypse environment. If you want to make quick money, services win. With a product, you’re always constrained by your product’s view of the world; the sales cycle can be long, and the competition is fierce. Services offer flexibility, especially if you’re selling to small businesses (the US alone has over 33 million). Sell a narrowly scoped outcome. Package your process after you’ve repeated it three or four times. Then sell it as a flat-rate offering. Some of you might find hourly a better option. My experience is that hourly is fine to gauge value early, but find ways to productize things. The incentives are cleaner for everyone. But do what you need to do. The main point is to make money, right?
I don’t suggest starting with a course or a community either. Unless you already have an audience, distribution is the hard part. If you doubt me, try building an audience from the ground up…if you’re starting from zero or a few hundred, get back to me when you’re even at 1K subs. Now try 10K, 100K, and beyond. It’s WAY harder than it looks. BTW, this might also be a BS metric, since simple subs and views alone can be gamed. Play the long game and prioritize engagement and relationship-building above all else.
More practically, engage with people. Go to local meetups. Talk to people. But don’t pitch. This is counterintuitive, but the moment you become a salesperson trying to extract money from someone, that’s how they’ll see you. I always made it a point not to pitch people directly. I just figured that when the time is right, and the budget is there, they’ll come back. A lot of them did.
Also, don’t only think about selling to big companies, which are often inundated with pitches from consultants and product companies. Small businesses have pain points, likely have a budget, and just need guidance. All too often, people target enterprise customers and ignore the small-business market that would actually pay them tomorrow. Cash flow is key, and it’s better to be paid today than wait 90+ days to get paid on net terms from some big company’s procurement department.
The good service lanes (for big and small clients) for data people right now: data platform audits, AI agent setup and assistance, warehouse cost optimization, analytics engineering cleanup, semantic layer design, experimentation and metrics reviews, exec dashboard redesigns, AI-readiness assessments, data modeling, and governance advisory. These are just some that come to mind that I’m seeing people do, but each of these solves painful, recurring, and budgeted problems. That’s the whole game.
Bottom line is, the world is changing under your feet. Have a plan B, reorient your work toward AI, and keep moving forward.
Have a great weekend,
Joe
Here’s this week’s Freestyle Friday podcast. Available on Spotify, Apple, and wherever else you get your podcasts. Please support the show with a review. It means a lot.
Modern data modelers need to live at the intersection of business and tech. Ellie.ai allows you to collaborate effectively with business while maintaining credibility with the Tech team. Get contextual support from AI, reverse engineer anything building a repository of sources with synthetic AI generated contextual metadata while delivering insights via an MCP Server and integrating anything with full blown API support.
The full stack data modeling future is here today!
Thanks to Ellie.ai for partnering on this newsletter.
Awesome Upcoming Events
All technical. All data. All day.
On March 26, 2026, there’s a very cool single-track, deeply technical data engineering event in San Francisco.
Capped at 100 practitioners.
No salespeople. No sponsors. Just talks + the hallway track, followed by happy hour.
When: March 26, 2026 · 10am–4pm + happy hour
Where: L4, 945 Market St, San Francisco
Thanks to Confluent for partnering on this event and newsletter.
🇸🇪 Sweden! See you at the Data Innovation Summit in Stockholm.
I’m doing a keynote and workshop on Mixed Model Arts: Data Modeling in the Age of AI.
May 7 - keynote
May 8 - workshop
Here’s 10% off: SD10OFF (good for the event. Workshop is not included)
Register here
📝 If you’re using AI in your work, we’d love your input for this March pulse survey.
It ends today (Saturday March 21) at 11:59PM PT.
It’s anonymous and only takes a one minute or less. Your response is very much appreciated.
Survey here.
Cool Videos and Reads
This week’s Lunch and Learn with Clair Sullivan.
Tech layoffs are brutal right now, and job searches are dragging on for months. Here’s the plot twist: going solo is often faster AND more stable than landing your next role. Whether you’re still employed or already laid off, you’ll learn how to create your own opportunities instead of waiting for permission.
In this other episode, I sit down with Demetrios Brinkmann (godfather of the MLOps Community) to talk about the absolute Wild West of AI right now.
We cover how fast coding agents are changing the game, the reality of vibe coding your own CRM , and how Demetrios’s community saved $20,000 just by ditching bloated enterprise tools.
But we don’t just talk tech. We get into the weeds on the content creation pipeline, from the bizarre rise of AI OnlyFans to the “Doorman Paradox” of automated content.
Finally, we spill some serious inside baseball on the tech sponsorship game, calling out the sheer audacity of heavily-funded startups expecting free labor from communities , and why protecting your reputation is worth more than any quick paycheck.
Here are some things I read this week that you might enjoy.
AI still doesn’t work very well in business, reckoning soon • The Register
We Have Learned Nothing - Colossus
The Guinness Inspector Teaching America How to Pour the Perfect Pint
The 8 Levels of Agentic Engineering — Bassim Eledath
Reverse-engineering Claude’s generative UI - then building it for the terminal
Define once, use everywhere: a metrics layer for ClickHouse with MooseStack
The Human Skill That Eludes AI
Find My Other Content Here
📺 YouTube - Interviews, tutorials, product reviews, rants, and more.
🎙️ Podcasts - Listen on Spotify or wherever you get your podcasts
📝 Practical Data Modeling - This is where I’m writing my upcoming book, Mixed Model Arts, mostly in public. Free and paid content.
The Practical Data Community
The Practical Data Community is a place for candid, vendor-free conversations about all things tech, data, and AI. We host regular events such as book clubs, lunch-and-learns, Data Therapy, and more.
Great content as always 🫶🏻 I particularly relate with the "moving Data to point A to point B Is not enough anymore", and I am trying to experiment with the things a modern data Engineer Is expected to provide on top of that. Do you have any good read / resource to recommend on semantic layers, embeddings etc. that practically translate to real world scenarios? thanks
Appreciate your takes, very grounded and helpful.