Most people who discover AI tools make the same mistake I did at first. They pick one, try to make it do everything, and wonder why the results feel inconsistent.

The better approach is simpler: stop trying to find one tool that does it all, and start thinking about which tool is right for which job.

For me, that split comes down to two tools. Manus and Claude. They do very different things, and once I understood the difference clearly, the way I work changed completely.

The Simple Version

Manus is my research agent and virtual assistant. It goes out into the world, gathers things, builds things, and runs things — with minimal input from me.

Claude is my thinking partner. It helps me make sense of what Manus finds, work through decisions, develop strategy, and produce work that requires context, nuance, and judgment.

Manus acts. Claude thinks. I direct.

What Manus Does

Manus is at its best when the job is clear and the output is concrete. Give it a task with a defined endpoint and it gets to work — often without you needing to be there at all.

In practice, this is what I send to Manus:

Research tasks

Go to [competitor site] and do a full teardown of their onboarding flow. Screenshot each step, note what's working and what's confusing, and compile it into a report.

Finding and compiling information

Search LinkedIn and Lu.ma for events in Singapore and Australia in March and April focused on female entrepreneurship. Compile into a spreadsheet sorted by date.

Building tools

Build me a web app CRM to track my LinkedIn outreach. Include a pipeline tab, analytics dashboard, and a weekly habit tracker. Deploy to a live URL.

Scheduled recurring tasks

Every weekday morning at 6am, pull the three most relevant industry news items for a founder in my space and send me a summary under 200 words.

Automating repetitive workflows

Every Friday, generate five reflection questions based on my week. Keep them specific, not generic.

The pattern is consistent: a clear brief, a concrete deliverable, and Manus handles the execution. I don't need to be present. I come back to a finished result.

What Claude Does

Claude doesn't go out and gather things. It works with what it's given — and that's exactly the point. The quality of what Claude produces is directly tied to the quality of context you bring to it.

This is where the thinking happens.

Strategy and positioning When I was working through brand positioning for OhhWells, I brought Claude the market research, the customer insights, and the competitive landscape — and used it to pressure-test ideas, challenge assumptions, and sharpen the message.

Writing and communication Blog posts, outreach emails, partnership proposals, ad copy. Claude doesn't just write — it writes in context. The more it knows about your brand voice, your audience, and your goal, the better the output.

Decision-making and frameworks When I was restructuring how our product team works, I used Claude to think through the tradeoffs — what the triad model was optimising for, what we were losing by keeping it, and what a better structure might look like.

Synthesising research into action Manus delivers the raw material. Claude turns it into something usable. A competitive teardown becomes a strategic brief. A pile of user quotes becomes a set of design principles. Raw data becomes a decision.

The Workflow in Practice

Here's how a typical research-to-decision cycle looks when both tools are working together:

Step 1 — Brief Manus Define the research task clearly. What do you need to know, where should it look, what format do you want back?

Step 2 — Manus goes to work You step away. Manus browses, compiles, screenshots, and builds. You come back to a finished document or deployed tool.

Step 3 — Bring the output to Claude Paste in the research, add context about your situation, and start the thinking conversation. What does this mean? What should we do about it? What are we missing?

Step 4 — Claude helps you decide and act You leave with a clear position, a strategy, a piece of writing, or a framework — shaped by real information, not guesswork.

Step 5 — Back to Manus if needed If the decision requires more building or more research, Manus gets the next brief.

The loop is tight. Each tool feeds the other.

The One Thing That Makes This Work

Context.

Manus doesn't need much of it — it just needs a clear task. But Claude needs to know who you are, what you're building, what decisions you're facing, and what you've already tried. The more context Claude has, the more useful it becomes.

This is why Claude Projects matters. I use it to store the background that Claude needs to be genuinely useful — brand positioning, audience insights, strategic decisions already made, team principles. Every new conversation inside a project starts with that context already loaded.

Think of it this way: Manus is a brilliant executor who works best with a clear brief. Claude is a sharp thinking partner who works best when they know the full picture. Your job is to give each one what they need.

A Simple Rule to Remember

When you're about to open an AI tool, ask yourself one question first:

Do I need something done, or do I need to think something through?

Done → Manus. Think → Claude.

Start there. The workflow builds itself.

I share the exact prompts, project setups, and workflows I use across both tools. Subscribe to get them.