Articles/Skills & Practice

What Is Cowork and What Kind of Tasks Is It Actually Good For?

A practical explainer with real professional services examples: due diligence, research synthesis, client reporting.

February 2026·7 min read

Most people who use AI regularly have a mental model that looks something like: I write a prompt, I get output, I review it and use what is good. This works for a lot of tasks - drafting an email, summarizing a document, generating a first pass at a memo.

Cowork is a different mode. Rather than a single exchange, it is an extended session where AI handles a longer, multi-step task with you providing direction, feedback, and judgment at key points. The AI does more of the sustained work; you do more of the steering.

The difference in practice

Prompt-and-response mode: you write a prompt, you get one output, you decide whether to use it or try again. Each exchange is independent.

Cowork mode: you brief the task at the start, the AI works through it in stages, you check in at defined points and give feedback, it adjusts and continues. The session has a beginning, a middle, and an end - and the output at the end reflects everything that happened in between.

The practical implication: Cowork is appropriate when the task is too complex or too long to handle in a single prompt-and-response cycle. It is how you get AI to help with the work that actually takes hours, not minutes.

Tasks Cowork is actually good for

The right candidates have two characteristics: they have multiple distinct steps, and the quality of later steps depends on doing earlier steps well. Due diligence is a clear example. Research synthesis is another. Client reporting for a complex multi-month engagement is a third.

Due diligence document review

A Cowork session for due diligence might start with a brief: here is the deal, here are the documents, here are the risk categories I need you to flag. AI reads through the material, produces an initial risk summary, you review and direct attention to areas that need more depth, it produces a revised summary, you add your judgment to the sections that require it. The final document is a real product of the session, not a first draft you rewrite from scratch.

Research synthesis

Synthesizing research across a dozen sources is repetitive, slow work when done manually. In Cowork, you load the sources, brief the synthesis task - what question you are answering, what level of detail is appropriate, what format the output should take - and AI works through the material systematically. You review the synthesis at natural stages and redirect as needed.

Complex client reports

A quarterly client report typically requires pulling together multiple data sources, summarizing progress across several workstreams, drafting executive commentary, and formatting everything consistently. Each of these steps is a candidate for AI assistance. In Cowork, you can move through them in sequence within a single session, with the AI carrying context from earlier steps into later ones.

Where it breaks down

Cowork does not work well for tasks that require real-time judgment throughout, where you cannot define upfront what good looks like, or where the task is short enough that the overhead of an extended session is not worth it.

Creative strategy development is a good example of the first. The quality of each step in a strategic discussion depends on judgment that cannot be specified in advance. You need to be in the loop throughout, not checking in at defined checkpoints.

Answering a single specific question is a good example of the third. If the task takes ten minutes and produces one clean output, prompt-and-response is the right mode.

The briefing is most of the work

Cowork sessions succeed or fail on the quality of the briefing. A good brief specifies the task, the context, the desired output format, the checkpoints where you want to review, and what the AI should do when it hits ambiguity - ask for clarification rather than guess.

The first time you run a Cowork session on a particular task type, the brief takes some thought to write. The second time, you adapt the brief from last time. By the third time, it is a Skill - a structured briefing template that anyone on the team can use on the same kind of task.

This is how Cowork connects to the Skills library: the briefing templates that work become Skills. The result is that the hard thinking you did once - figuring out how to brief a due diligence review, how to structure a research synthesis session - becomes something the firm owns.

Cowork is covered in depth in Module 3 of Apparatus 202, including how to brief longer tasks, where to put checkpoints, and how to turn your best Cowork sessions into Skills. The article on turning prompts into Skills is useful context for the connection between the two.

Next step

Ready to turn what your team knows into something that lasts?

Apparatus 202 gives your team a Skills library, connected data sources, and workflows the firm owns. One session — no ongoing subscription.