What Manus Does, and Where It Fits a Studio

Manus calls itself an autonomous general AI agent. That means software that plans a job, works through it step by step, and hands back finished work. A chat assistant mostly just answers.

Each task runs in a sandbox: a cloud virtual machine, which is a computer Manus runs in the cloud for that one task. It has a network connection, a file system, a web browser and software tools. Inside it, Manus can open websites, save notes to files, install software and run code.

Here is an example from your world. You write: "Research how 20 outdoor-clothing brands talk about repair, one row per brand, with links." Manus writes a plan, opens each brand's site in its browser, records what it finds in a spreadsheet, and returns the spreadsheet with a short report. While it works, you can watch it browse and click, and you can replay the session afterwards.

The difference that matters for a studio: a chat assistant gives you text to act on. Manus does the steps itself and gives you files. That makes it more useful, and it also means you need to check what it actually did.

Manus returns work as files you can open and edit:

• Research reports and data analysis with charts
• Slide decks as editable .pptx, PDF or web slides, plus a speaker-notes file. You can apply your own agency template.
• Websites and web apps, including prototypes built from Figma frames
• Images and short video, made with third-party models listed on Manus's models page
• Documents and spreadsheets

For a creative director, this covers most of the preparation around an idea: the category scan before the brief meeting, the register behind the moodboard, the first build of the concept deck.

Keep one warning from the track in mind from day one. A correctly named file proves nothing. Concept_Deck_v3_FINAL.pptx tells you nothing about its quality, whether its images are cleared, or whether anyone approved it. Manus can produce a polished deck quickly. Whether that deck holds the idea is still a separate question, and the answer comes from you.

Manus helps most with multi-step jobs that have a clear format and can be checked against sources:

• Questioning a brief. It lists what's missing, what to push the client on, and which lines are evidence and which are hope.
• Category research across many brands. It covers positioning, tone, colours and recent campaigns, one row per brand.
• Reference registers. It records each reference's job, where it came from and its usage status.
• Decks in a template. It builds a first version of the concept deck in your .pptx template.
• Public touchpoint audits. It checks public pages against the brand guidelines and records each issue with a screenshot.
• Prototypes from Figma frames. It builds a working web prototype for testing.

What these jobs have in common is that they are wide or fiddly, and a person can check them against something: a source, a guideline, a template or a frame.

Take the track's teaching case. A heritage jacket brand is launching a repair service and wants owners to bring back the jacket they already have. Manus could scan 25 brands' repair and resale messaging by Monday morning. Your team would spend that time judging the results instead of collecting them.

Three jobs stay with you, however good the tool gets:

1. Deciding which route is worth defending. Manus can make ten routes before lunch. It can't tell you which one you'd stand behind in front of the client.
2. Judging what a stranger should feel. In the jacket case, that might be: this jacket is a record of where I've been, and it's worth keeping. That is a point of view, and a point of view needs an author.
3. Releasing work under the brand's name. Someone has to answer for it to the client, to rights holders and to the public, and that can't be an agent.

One way to remember this: AI makes trying an idea almost free, so you'll have plenty of routes. A route you would defend is as rare as it ever was.

The jacket case shows the risk. In the brief, the wear (the frayed cuff, the elbow patch, the scorch mark) is the one true thing. A generated frame that cleans up the sleeve may look better, but it breaks the idea. Manus won't notice that. You have to.

The voices in this track describe two opposite mistakes.

Staying out. Nick Knight, the fashion photographer who founded SHOWstudio, says in a May 2026 conversation that anxiety about a big cultural change is normal and healthy. What worries him more is creatives who refuse to engage, who "turn their back on it" (Knight). If you stay out, others decide how the medium is used.

Settling. David Droga, founder of Droga5 and now Accenture's vice chair, told Semafor in June 2026 that AI will first replace the formulaic, average middle of marketing, which he calls "the mediocre middle" (Droga). His point is that AI is good at replicating best practice, so people using the same tools reach the same place. If you settle, the model's defaults become your brand's look.

The course takes the path between them. Manus widens the search and carries the making, and a person keeps custody of the idea: what the work is for, what it should make a stranger feel, and when it is ready. Both quotes come from machine captions, so treat the exact wording as approximate.

Nils Leonard, co-founder of Uncommon Creative Studio, argues that the real threat to creativity "isn't AI, it's apathy" (Leonard). He's talking about creatives who discuss the change instead of making things. So this lesson ends with a piece of work. By the end of this module you'll have run a real task in Manus.

Sort your studio's jobs into three groups:

• Hand: Manus does the job and you check it. For example, category research or a public touchpoint audit.
• Share: Manus prepares the material and you make the call. For example, it lists the gaps in a brief and you decide which questions go to the client. Or it fills in a reference register, and a person confirms every usage status, because a discovered asset counts as uncleared until someone proves otherwise.
• Keep: you decide and Manus stays out. For example, choosing the route, judging the feeling and releasing the work.

Try it on the matching exercise. Then write down three jobs from your own week and sort them the same way. Keep that list, because you'll use it in this module's lessons on access and plans.