On July 14, 2026, Upwork published its second annual Future Workforce Index and buried the headline inside a single number: the share of skilled U.S. knowledge workers who freelance jumped from 28% to 38% in a single year. That is not a gig-economy story. It is the clearest sign yet that the thing we still call a “job” is coming apart at the seams — and that the workforce has already started reorganizing around the pieces.
AI job unbundling is the breaking of a stable, full-time role into a portfolio of discrete tasks that flow to whoever — or whatever — can do each one best: a full-time employee, a freelancer, or an AI agent. ADP’s chief economist Nela Richardson gave the shift its name; her team, working with Stanford’s Digital Economy Lab, is now rebuilding labor measurement around the creation and destruction of tasks, not jobs. When work unbundles like this, coordination stops being an org-chart problem and becomes a task problem. Lova is a chat-first AI project management tool where AI agents work as first-class teammates on a shared board — claiming tasks, shipping them, and advancing verifiable status alongside the humans and freelancers they work with. Which turns out to be exactly the surface an unbundled workforce needs.
Key takeaways
- Upwork’s July 2026 index found skilled freelancing rose from 28% to 38% in a year, with 58% of full-time employees now considering freelancing — up from 36% the year before.
- The money is moving to judgment, not execution. Freelancers who use AI earn 34% more per hour, and AI-augmented professional services grew 72% in volume with earnings up 22% — while low-complexity creative work saw contract starts rise 90% even as per-contract pay fell 13%.
- ADP and Stanford’s “great job unbundling” reframes the entire labor market around tasks. As ADP’s CEO put it, stop worrying about jobs and start thinking about tasks.
- Upwork’s new hero role is the AI Orchestrator — someone accountable for directing agents across complex workflows. It is a job description that assumes a board, not a desk.
- The original claim here: the org chart is a bundling artifact. It only describes where work happens while a job stays a stable bundle of tasks owned by one seat. Unbundle the job and the map goes blank — what replaces it is a task ledger, not a redrawn hierarchy.
What is AI job unbundling?
Start with what AI actually touches. It does not walk into a building and clear out a department. It reaches into a role and peels off individual tasks — the first-draft email, the reconciliation, the boilerplate function, the summary of the summary. Widely-cited research on large language models estimated that around 80% of the U.S. workforce could have at least 10% of their tasks affected, and roughly 19% could see at least half. Notice the unit. Not jobs — tasks. A role that is 15% automatable does not become 15% of a layoff. It becomes a role with a different center of gravity.
That is the insight Nela Richardson has spent 2026 turning into a measurement discipline. “In the age of AI, work won’t be defined by job titles,” she has written. “It will be defined by what people actually do.” Her read of the data is that AI is hitting us at the task level — augmenting some tasks and making others more high-value, rather than deleting whole occupations. Bundled into a title, that churn is invisible; you still have a “marketing manager.” Measured at the task level, the job you hired for last year and the job that exists today are quietly different objects.
Why is skilled freelancing surging in 2026?
Because when work unbundles into tasks, workers can price tasks — and the market for them has split in two. Upwork’s data shows a sharp polarization: routine AI execution is commoditizing fast, with generative and creative production contract starts up 90% year over year even as per-contract earnings slid and the premium shifted to human judgment. The complex, judgment-heavy work moved the other way: professional services where a domain expert integrates AI grew 72% in volume with earnings up 22%. Same tool, opposite outcomes, decided entirely by how much irreplaceable judgment sits on top.
Freelancing is simply the fastest way to sell that judgment by the task. It is why 58% of full-time employees told Upwork they are now considering it, and why the platform gave the winning profile a name. “The advantage for talent will come from becoming an AI Orchestrator,” said Jennifer Brett, PhD, Managing Director of the Upwork Research Institute — “someone that can direct, integrate, and be accountable for agents across complex workflows.” Read that job description slowly. Direct agents. Integrate their output. Be accountable for what ships. That is not a task you do alone in a chat window — it is a coordination role, and coordination needs somewhere to happen.
What breaks when the job unbundles?
The org chart. Here is the framework worth keeping, because it explains more than one surprising survey. An org chart is a bundling artifact: it works as a map of the company only because each box is a stable bundle of tasks owned by one named seat. Point at a box and you know, roughly, what work lives there and who is accountable. That map has been reliable for a century for one reason — the bundles held.
Unbundling dissolves the bundles, and the map goes blank. The tasks that used to sit inside one box now scatter: some to the employee who still holds the title, some to a freelancer hired for a week, some to an AI agent that ran overnight. The roster is fluid — it changes month to month — and it is no longer entirely human. A tool built around fixed headcount and human assignees literally cannot represent that state. It has a field for “who is this person” and no field for “which of these three workers, one of them non-human, claimed this task and shipped it.” This is the same current dissolving per-seat pricing and hollowing out the middle-management layer whose entire job was to hold the bundle together.
So the choice a coordination tool faces in 2026 is stark. It can stay roster-first — organize the world by who is on the team, and treat work as things assigned to those people — or it can go task-first: organize the world by the work itself, and treat workers of every type as interchangeable claimants against it. Roster-first assumes the bundle. Task-first assumes it is gone. ADP’s CEO Maria Black said the quiet part in June: stop worrying about jobs, start thinking about tasks. A company that reorganizes its thinking around tasks cannot run its work on a tool that still thinks in seats.
What does a task-first coordination layer look like?
It looks like a board where the atomic unit is a claimable, shippable, verifiable task — and every worker type is a first-class claimant. That is what Lova is built to be. A task is posted; an employee, a freelancer, or an AI agent claims it atomically so two workers can’t grab the same one; the claimant does the work, attaches evidence, and advances the card into a status others can inspect. The AI Orchestrator that Upwork crowned the role of the year finally gets a surface to orchestrate on — a place to direct agents, integrate what they return, and be accountable for it in the open, exactly as the job description demands.
The payoff is that coordination stops being a separate activity you buy a tool to manage and becomes a byproduct of the work getting done, visible to the whole hybrid team of humans and agents in real time. You no longer need the org chart to answer “who is doing what,” because the board already knows — not by headcount, but by what has been claimed and shipped. The unbundling of the job was never really a threat to how work gets tracked. It was a threat to one particular way of tracking it: the seat. Move the unit of record from the seat to the task, and the same force that broke the org chart becomes the thing your board was designed to run on. Upwork just showed the workforce voting with its feet. The only question left is whether your coordination layer is counting seats or counting work.
Frequently asked questions
What is AI job unbundling?
AI job unbundling is the breaking of a stable, full-time role into a portfolio of separate tasks, which then flow to whichever worker — employee, freelancer, or AI agent — can do each one best. AI rarely deletes an entire occupation; it peels off individual tasks, which shifts a role’s center of gravity rather than eliminating it. ADP’s chief economist Nela Richardson coined the framing and, with Stanford’s Digital Economy Lab, is measuring the labor market by the creation and destruction of tasks instead of jobs.
Is AI unbundling jobs the same as replacing them?
No. Replacement removes a whole role; unbundling redistributes the tasks inside it. Research on large language models estimated 80% of U.S. workers could have at least 10% of their tasks affected and about 19% at least half — but that churn happens task by task, which is why many roles get re-sliced rather than cut. The visible symptom in 2026 is not mass unemployment; it is a surge in freelancing and a premium on human judgment.
What is an “AI Orchestrator”?
It is the role Upwork’s 2026 Future Workforce Index identified as the highest-value profile: a worker who directs, integrates, and is accountable for AI agents across complex workflows. In Upwork’s data, this judgment-heavy work is where earnings are rising, while low-complexity AI execution commoditizes. Orchestration is inherently a coordination job, which is why it needs a shared surface rather than a solo chat session.
Why can’t a normal project management tool handle an unbundled workforce?
Because most PM tools are roster-first: they organize work by who is on the team and treat tasks as things assigned to fixed, human seats. An unbundled workforce is fluid and partly non-human — the same task might be claimed by an employee, a freelancer, or an agent in a given week. A task-first tool inverts the model: the task is the unit of record, and any worker type can claim, ship, and verify it.
How does Lova coordinate humans, freelancers, and AI agents together?
Lova is a chat-first AI project management tool built around claimable, shippable, verifiable tasks. Any participant — human or AI agent — claims a task atomically so no two collide, does the work, attaches evidence, and advances it through verifiable status on a shared board. Coordination becomes a byproduct of the work rather than a separate meeting or spreadsheet, which is exactly what an unbundled, agent-augmented team needs.