How to use AI in your workflow without making things more complicated

AI was meant to simplify work. It was meant to help teams organise tasks, read information, suggest priorities and reduce the time lost to operational management. In theory, work management seemed one of the most suitable contexts in which to introduce it: lots of data, lots of activities, lots of constant updates.

In practice, however, many AI features are tried for a few weeks and then abandoned. In 17 interviews with designers, project managers and content strategists, 8 people stopped using AI after less than a month, while another 5 use it only for marginal tasks. No one really sees it as an integral part of their workflow.

The problem is not that AI is useless in absolute terms. The problem is that it is often placed in the wrong spot, on top of systems that were not designed to make it truly work.

AI does not solve chaos if it is built on top of chaos

Many traditional work management systems were created to offer maximum flexibility: tasks, boards, pages, custom fields, databases, comments, documents and different views for every team. At first, this freedom looks like an advantage, because it lets each organisation build its own way of working.

Over time, however, that same freedom can turn into disorder. Information multiplies, priorities become hard to read, responsibilities overlap and context is scattered across different tools, chats and documents. When AI is added on top of this fragile structure, it does not find a clear system to enhance. It finds fragments.

And this is where the paradox begins: a technology designed to reduce cognitive load ends up increasing it, because it requires people to guide it, correct it, check it and constantly explain to it what the system cannot understand on its own.

Why AI in traditional work management systems disappoints

AI works when it has access to readable context. It needs to know what is urgent, who is responsible for what, which activities are blocked, which decisions have already been made and where human judgement is still needed.

In traditional work management systems, this context often does not exist in a structured way. It is implicit, distributed, negotiated in meetings, hidden in messages or scattered in comments. That is why AI may be able to generate a summary, create a draft or suggest a task, but struggles to become truly useful in day-to-day work.

At that point, the user has a very simple feeling: “I’d be quicker doing it myself.” Not because the technology is not powerful, but because using it requires too much preparation, supervision and correction.

The real issue is delegation

Integrating AI into work does not simply mean adding a smart assistant inside a platform. It means deciding what can be delegated, what must remain under human control and which decisions require confirmation.

When this logic is not designed in, two opposite mistakes happen. In the first case, AI is used only for marginal tasks, like summaries or generic copy, and therefore does not really reduce cognitive load. In the second case, it is given decisions that require context, experience and organisational sensitivity, but the system does not have enough information to make them well.

The point is not to use more AI. The point is to use it at the right point in the workflow.

Why AI agents are not enough on their own

There is a lot of talk today about AI agents, meaning systems capable of planning actions, using external tools and completing complex tasks with a certain level of autonomy. But even AI agents do not work in a vacuum.

To be truly useful, they need clear goals, defined roles, formalised processes and precise rules for oversight. They need to know when they can act alone, when they should propose a solution and when they must stop and ask for confirmation.

The problem is that many teams do not yet have this structure even among people. Responsibilities overlap, tasks are ambiguous, priorities change informally and decisions live in calls, chats and scattered documents. In this context, AI cannot work miracles: it can only amplify what it finds.

What is really needed: AI-native systems

A traditional work management system adds AI as an extra feature on top of an existing structure. An AI-native system, by contrast, is built with AI at the heart of its architecture.

The difference is substantial. In an AI-native system, processes are modelled from the start, roles are clearer, priorities are traceable, workloads are visible and delegation rules are built into the way the system works. AI does not have to reconstruct context after the fact, because the context is already part of the structure.

This makes a much more practical use of AI possible: not a side chatbot to query every now and then, but a system that reads the work as it happens, understands where to intervene, suggests next steps and knows when to leave room for human judgement.

The problem is not AI. It is where you put it.

AI amplifies what it finds. If it finds clarity, it multiplies it. If it finds confusion, it multiplies that too.

That is why many AI integrations in work management systems do not work: they try to automate processes that were never truly designed. They add intelligence on top of fragile structures, without solving the underlying problem.

The future of work management will not simply be made of more AI. It will be made of better-designed systems, capable of making work readable before they automate it.