Asana's Workload view will show you, in colour, that Priya is at 140 percent of capacity next week. It is a good feature. It is also only as truthful as the effort values sitting on the underlying tasks, and on most teams those values are filled in for about the first three weeks after somebody sets Workload up.

A capacity bar chart showing one person over the red line, beside a list of unestimated tasks that are invisible to the chart
The tasks with no effort value are invisible to the chart. They are usually the reason the chart is wrong.

What Asana Workload already does

Worth stating plainly, because the native feature is more capable than a lot of teams use. From the Workload tab you click Add effort and choose a numerical custom field to weight tasks by. You can measure in hours, in points, or simply by task count if you have not set up a field. Effort updates automatically when the underlying custom field changes, whether that happens in a project or in someone's My Tasks.

You then set a maximum capacity per team member. When assigned work exceeds that capacity in a given week, Workload shows a red line indicating that person has too much on. Portfolio workload extends the same view across multiple projects, which matters because almost nobody's week is contained inside one project.

That is a genuine resource-management feature and it is not the thing teams struggle with. The struggle is that it runs on data somebody has to maintain.

Why the red line stops being trustworthy

Workload is a calculation over a field. The calculation is always right. The field is frequently wrong, and it goes wrong quietly, which is worse than going wrong loudly.

What Workload showsWhat it cannot tell you
Total effort per person per week against capacityThat eleven tasks this week have no effort value and are counted as nothing
A red line when someone is overWhether they are over because of real work or because one task was estimated in the wrong unit
Effort as recorded in the custom fieldThat the team's five-point tasks have quietly been taking nine points of time since spring
Who is over capacityWhich specific task should move, and who has the context to pick it up
The current week's allocationThat the same person has been over capacity for six consecutive weeks
Assigned workThe unassigned work that is coming and has nowhere to go yet

Every row in the right-hand column is answerable from data Asana already holds. None of them are answerable by a chart, because they need something compared, remembered, or judged. That is the shape of work worth delegating.

What to hand an agent

"Balance the team's workload" is the wrong instruction, and an agent given it will start reassigning tasks, which is exactly what you do not want. Scope it to the data and the proposal.

ElementWhat to specify
InputAll tasks in the portfolio for the coming two weeks with assignee, due date, effort field and project, plus the last quarter of completed tasks for calibration
Missing-data passList every task with no effort value, and propose one based on comparable completed tasks, clearly marked as a proposal
Calibration passCompare estimated effort against actual completion time on recent work and report drift, per person and overall
Rebalance optionsFor anyone over capacity, propose two or three specific tasks that could move, with who could take each and why
PermissionsRead and propose only. It does not reassign tasks, change due dates, or write effort values
EscalationWhere it has no comparable work to estimate from, say so rather than guessing a number

The permissions row again. Reassigning someone's work is a management act, not a data operation. An agent that silently moves a task has made a decision about two people's weeks without either of them being asked. Propose-only keeps the useful part and drops the part that damages trust.

A working setup, end to end

Every Friday the agent reads the coming fortnight across the portfolio, runs the three passes, and posts a short report: unestimated tasks with proposed values, estimate drift, and rebalance options for anyone projected over capacity. The team lead spends ten minutes accepting proposals, which writes the effort values, and takes the rebalancing conversation to the standup rather than to the tool.

The reason the estimate proposals matter more than the rebalancing suggestions is a bit counter-intuitive. Rebalancing is a judgment call a lead can make in seconds once they can see the picture. Filling in forty missing effort values is a chore nobody does, and it is the thing that makes the picture visible in the first place. Fix the data and the chart starts telling the truth on its own.

For the mechanics of connecting to Asana without handing over more than the job needs, connecting an agent to a private API covers scoping, and AI agent security best practices covers approval gates for exactly this propose-only pattern. To make it a standing Friday job rather than something someone triggers, see writing a prompt for a recurring agent.

It also pairs with the rest of the project-tracking upkeep an agent can absorb: Asana inbox zero on the notification side, ClickUp task automation and Jira backlog grooming for teams running the same pattern in other trackers, and sharing an agent with your team once the report is worth more than one person seeing.

What goes wrong

We run this pattern on our own planning, and three things reliably bite.

The first is estimating from titles. "Fix onboarding bug" could be twenty minutes or three days, and an agent that estimates from the title alone produces numbers that look authoritative and are noise. It needs the description, the subtasks and ideally the comparable completed work. Where those are thin, the honest output is an escalation, not an estimate.

The second is treating capacity as a target. Once a chart exists, there is a pull toward filling everyone to 100 percent, and an agent asked to balance workload will optimise for exactly that. It is the wrong goal. Teams need slack for the work that arrives unplanned, and a schedule with no slack converts every surprise into a missed date. We ask for rebalance options only for people over capacity, and explicitly not for people under it.

The third is the quiet one, and it is the same trap as velocity in sprint reporting: if estimates drift upward over time, workload looks healthier while the same amount of work is happening. That is why the calibration pass is in the spec rather than optional. Comparing estimate against actual is the only thing that keeps the other two passes meaningful.

The honest limit: this does not do resource planning for you, and it does not know that Priya is about to go on leave or that Sam is the only person who can review the payments work. It keeps the data clean enough that the human doing the planning is looking at something real.

Frequently asked questions

How does Asana Workload calculate capacity?

You choose a numerical custom field to represent effort from the Workload tab using Add effort, measured in hours or points, or you can leave it as task count. Asana totals that effort per person per week, and you set a maximum capacity for each team member. When assigned work exceeds their capacity, Workload displays a red line. Effort updates automatically whenever the underlying custom field changes in a project or in My Tasks.

Why is my Asana Workload view inaccurate?

Almost always because tasks are missing effort values, which count as zero and are therefore invisible to the chart, or because the estimates stopped matching reality and nobody recalibrated. Workload's arithmetic is not the problem. The field it reads is maintained by hand, and maintenance is the first thing to lapse once the novelty of the setup wears off.

Can an AI agent reassign tasks in Asana automatically?

It can, and it should not. Moving work between people is a management decision that affects two people's weeks, and doing it silently is a fast way to lose the team's trust in both the tool and the agent. Give the agent read access and have it propose rebalancing options with reasoning, then let a human make the call in a conversation.

What can an agent estimate effort from?

Comparable completed work is the only reliable basis. A task's description, its subtasks and the actual completion time of similar past tasks give an agent something to reason from. A title alone does not, and an agent estimating from titles produces confident numbers with nothing behind them. Where no comparable work exists, the correct output is to flag the task for a human estimate.

Which Asana plan includes Workload?

Workload is a resource-management feature on Asana's higher tiers rather than the free plan, and portfolio-level workload sits alongside it. Because Asana adjusts plan names and inclusions periodically, check the current pricing page for your region before budgeting, rather than relying on any third-party comparison, including this one.

How much does it cost to run this?

Less than the planning meeting it shortens. On Gravity the free tier covers one agent at $0 a month, which is enough to run a weekly report against a single portfolio and see whether the proposals are any good, and paid plans start at $20 a month with $20 of usage included, with extra usage available beyond the plan. For how that compares across the category, see our roundup of the cheapest AI agent platforms.

Sources