The End of Visible Work

We were talking recently about agent visibility, and I found myself returning to a question that feels increasingly important for every organisation building with AI: what happens when the work is still happening, but fewer people can see it?

Not because people are hiding anything or the work has disappeared, but because the work has moved somewhere else and this is the shift that matters.

For most modern organisational life, leaders have used visibility as a proxy for understanding work. If they could see the factory floor, the office, the paperwork, the customer queue, or the team gathered around a desk, they had some sense that work was happening.

It was never a perfect measure. But it was visible. And then work started moving.

It went from factories to offices then to computers, then to remote work, then to systems and finally now from systems to agents.

Each shift made work more abstract; an older form of visibility disappeared and forced organisations to invent new ways of understanding what was actually happening.

As a founder, I think we are at another one of those moments.

The Old Story of Automation Was Efficiency

For a long time, we talked about automation mainly as a story of efficiency.

A process had too many manual steps so a system was built. Work became faster. Errors went down. More volume could be handled by fewer people. The organisation became more productive.

That story is partly true, but it is incomplete.

In IT and operations, many of us saw this years ago. Entire teams that once performed repetitive operational tasks were replaced or more often compressed by systems that allowed one person to do the work of many.

One example, from my early IT days, I keep thinking about is insurance policy writing.

There was a time when whole floors of people were involved in preparing, writing, checking, and issuing policies for clients. The work had a physical presence. You could walk into the building and see the operation. The workforce was the workflow.

Then technology arrived that changed the shape of the work. Suddenly, one trained user could produce hundreds of policies through software. The work did not disappear. The policies still had to be created. The rules still mattered. The decisions still had consequences.

But the visible work changed, the floor got quieter, the process moved into a system and the organisation became more efficient, but also more dependent on something fewer people understood.

That is the part we do not talk about enough.

When work moves into systems, visibility does not automatically move with it.

The Same Visibility Gap Appeared in Remote Work

Remote work gave us another version of the same shift.

Before remote work became normal, a lot of management culture still depended on physical presence. People came into the office. Meetings happened in rooms. Documents moved across desks. Teams formed around proximity.

Then work moved into email, Teams, shared documents, project boards, chats, calls, and cloud systems.

Again, the work did not disappear. In many cases, there was more work than ever.

But the visibility changed.

Leaders could no longer rely on walking around the office to understand what was happening. They needed calendars, status updates, shared dashboards, weekly reviews, asynchronous documentation, and new rituals of coordination.

Some organisations adapted well.

Others simply recreated office behaviour through digital surveillance. They confused activity with output. They measured logins, green dots, meeting attendance, and message volume, because those were the easiest things to see.

And that was the warning sign.

When work becomes less visible, organisations often reach for the easiest visible signal, even if it is the wrong one.

Now agentic work introduces a deeper version of the same problem.

An agent does not sit at a desk.

An agent does not look busy.

An agent may not send a message until the work is complete.

An agent may gather information, transform it, compare sources, draft outputs, trigger workflow steps, and prepare a recommendation without creating the familiar signs of activity that humans are used to seeing.

The risk is not that no work is happening, but that it is happening without enough organisational visibility and that is why I think the next big challenge is not simply building agents.

It is making agentic work accountable.

The Pattern Keeps Repeating

The pattern seems to repeat.

First, work is physical.

In a factory, work is visible because it is embodied: machines, materials, workers, supervisors, outputs. You can see bottlenecks, hear production, walk the line.

Then work becomes administrative.

In offices, work becomes paper, phones, meetings, filing cabinets, inboxes, and approvals. Still visible, but less direct. The output is not always a physical object. It may be a decision, a contract, a policy, a report, or a grant application.

Then work becomes digital.

Systems start absorbing tasks. The policy writer example fits here. The work moves from teams of people into software. The output increases, but the process becomes harder to inspect unless the system was designed with logs, permissions, reporting, and controls.

Then work becomes distributed.

Remote work breaks the relationship between location and productivity. The organisation needs new ways to understand progress, not just presence.

Now work becomes agentic.

Agents can operate across systems, gather information, draft responses, update records, escalate exceptions, coordinate with other agents, and involve humans at approval points.

This is not just another productivity tool.

It is another shift in where work lives.

And every time work changes location, visibility has to be rebuilt.

At Cura, We See AI as Work, Not Chat

At Cura, we anchor on one belief: AI is work, not chat.

That distinction changes everything.

If AI is chat, the interface is the product.

If AI is work, the workflow is the product.

• Who requested the work?
• What sources were used?
• What decision was made?
• What assumptions shaped it?
• Who approved it?
• Can we reconstruct it later?

These are not UI questions. They are operating questions.

And they are the questions every serious agentic organisation will have to answer.

Agentic work is not just another software adoption curve. It is an operating model shift.

Not because agents are mysterious.

Because work is moving again, and visibility has to move with it.

Three Lessons for Agentic Organizations

The first lesson: visibility is not control.

A manager near a team may feel informed. A dashboard may show activity. An agent log may record steps. None of that guarantees understanding.

Useful visibility answers the questions that matter:

• What is happening?
• Why is it happening?
• What needs human attention?
• What can be trusted?

The second lesson: every major shift in work creates a visibility gap.

Factories made production visible. Offices made administration visible. IT systems made work scalable, but often hid the process. Remote work broke the link between location and productivity.

Agentic work will expose something even bigger: how much trust depends on knowing who did what.

With agents, that question gets more complicated.

Was the decision shaped by a person, an agent, a workflow rule, or a source that should have been challenged?

The third lesson: preparing for agents is not a technology project. It is an operating model project.

If agents are going to work inside real organisations, leaders need clear answers:

• Who owns the process?
• Where do approvals happen?
• What should never be delegated?
• What should always be reviewed?
• What evidence should be retained?

That is why observability will become a defining layer of agentic organisations.

Not dashboards for dashboards' sake.

Operational visibility.

A way to see work that no longer looks like work.

Why This Time Is Different

This time, we can see the shift coming.

Past waves of automation often arrived as local changes: a system here, a workflow there, a team structure slowly reshaped.

The pattern was obvious in hindsight. With agents, it is visible now.

Leaders, auditors, boards, regulators, funders, and customers will all ask the same question: what are the agents doing?

So the opportunity is not just to build agents faster.

It is to prepare better.

That means designing visibility into the work from day one.

Before something goes wrong.

Before adoption stalls.

Before people lose confidence.

Before someone asks, "Why did the agent do that?"

What We Need to Build Next

The question now is simple: what should the visibility layer for agentic work look like?

To make that visibility real, organisations need to answer a more practical question: how is the agent getting the work done?

The visibility layer should not only show that an agent produced an output. It should show the path it followed to get there: the goal it was given, the systems it accessed, the sources it used, the intermediate steps it completed, the assumptions it made, the checks it performed, and the points where a human was asked to approve, correct, or decide.

That process record matters because trust in agentic work will not come only from the final answer. It will come from being able to inspect the work enough to understand whether the answer was produced responsibly.

It should also make outputs visible. What did the agent create? A recommendation, a draft, a completed task, an updated record, a routed exception, a summary, a decision package, or a workflow handoff? Just as importantly, what was the agent not able to create, and why? Was information missing? Was permission unavailable? Did the task require human judgment? Did the agent hit a boundary that should not be automated?

This is where visibility becomes more than monitoring. It becomes organisational understanding. Leaders do not need to watch every agent action in real time. But they do need a reliable way to see the work pattern, review the evidence, understand exceptions, and know where accountability sits.

A list of agent actions is useful. But it is not enough.

Organisations need an operational record of work.

A way to see the process, handoffs, decisions, sources, approvals, exceptions, and outcomes.

A way for leaders to understand the work without micromanaging every step.

A way for humans to supervise agents without becoming the bottleneck.

A way to benefit from invisible work without losing accountability.

The future of work may be less visible in the old sense. But less visible does not have to mean less accountable.

Every time work moves, management has to move with it.

Governance has to move with it.

Trust has to move with it.

And visibility has to be rebuilt.

That is the next phase of agentic work.

Not replacing the office with a digital workforce, but learning how to see work again.