Day 13 of 28 · OpenClaw Challenge
Multi-Step Tasks
⏱ 7 min
📊 Intermediate
So far, most of our examples have been one tool at a time — check email, read a file, browse the web. But the real magic happens when the agent chains multiple actions from a single instruction.
Today you'll see why one sentence to your agent can replace 20 minutes of app-switching.
What is a multi-step task?
A multi-step task is any request that requires the agent to use multiple tools in sequence to get you a complete result.
You don't tell it the steps. You tell it the outcome you want. The agent figures out the steps itself.
Knowledge Check
In a multi-step task, who decides which tools to use and in what order?
A
The tools decide automatically
B
You — you list the steps for the agent
C
Each tool calls the next tool in a fixed order
D
The agent — you describe the outcome, it plans the steps
You describe what you want. The agent plans and executes the steps. That's the Think step of the agent loop — the agent looks at your goal and figures out the best path to get there.
One instruction, five actions
Watch what happens when you say: "Prepare me for my 2pm meeting with the client."
Prepare me for my 2pm meeting with Acme Corp.
Pulling everything together...
Here's your prep for the Acme Corp meeting at 2pm:
📅 **Meeting details**: You + Sarah + Acme's VP of Engineering (Tom)
📧 **Recent emails**: Tom sent 3 emails this week — he's concerned about the timeline for Phase 2. Sarah replied with a revised estimate but Tom hasn't responded.
📄 **Shared doc**: The project plan was last updated Monday. Phase 1 is 90% complete. Phase 2 start date moved from March to April.
🌐 **Acme news**: They announced a new CTO yesterday — Lisa Park, previously VP Engineering at Stripe. Tom may report to her now.
💡 **Suggested talking points**:
1. Acknowledge the new CTO appointment
2. Address Tom's timeline concerns head-on
3. Present Sarah's revised Phase 2 estimate
4. Propose a weekly sync to keep alignment
Want me to send this prep to Sarah too?
↻ Replay conversation
Let's break down what happened. One sentence from you triggered five distinct actions:
Step 1 — Calendar tool: found the meeting details and attendees
Step 2 — Email tool: searched for recent correspondence with Tom
Step 3 — File tool: opened the shared project plan and checked status
Step 4 — Web tool: searched for recent Acme Corp news
Step 5 — AI reasoning: synthesized everything into actionable talking points
If you did this manually, you'd open your calendar, search your email, open the doc, Google the company, and write up notes. That's 15-20 minutes of context-switching. Your agent did it in 10 seconds.
Knowledge Check
How many tools did the agent use to prepare for the Acme Corp meeting?
A
Two — calendar and email
B
Four tools plus AI reasoning — calendar, email, files, web, and synthesis
C
Three — calendar, email, and files
D
One — it just checked the calendar
The agent chained calendar, email, file reading, web browsing, and its own reasoning into one seamless preparation. Each tool contributed a piece of the puzzle, and the AI put it all together.
The agent reports as it works
For longer tasks, the agent doesn't go silent. It reports progress so you know what's happening:
Research our top 3 competitors and build a comparison document.
Starting research — this will take a minute.
✅ Step 1/4: Identified your top 3 competitors from recent emails and market data
✅ Step 2/4: Visiting each competitor's website — scanning pricing, features, and recent announcements
✅ Step 3/4: Cross-referencing with review sites (G2, Capterra) for customer sentiment
✅ Step 4/4: Building the comparison document...
Done ✓
Saved "Competitor_Analysis_Feb2026.md" to your Documents folder.
Includes: feature comparison table, pricing breakdown, recent moves, customer sentiment scores, and my recommendations.
12 pages. Want a one-page executive summary instead?
↻ Replay conversation
Knowledge Check
Why does the agent report progress during long tasks?
A
To slow down and seem more human
B
Because each step requires your approval
C
Because it's required by its programming
D
So you can see what's happening, verify the approach, and redirect if needed
Progress reporting builds trust and keeps you in control. If the agent is heading in the wrong direction ("wait, those aren't our real competitors"), you can redirect immediately instead of waiting for a wrong result.
Manual vs. agent
Here's the bottom line. A task that requires checking email, reading a document, browsing the web, and scheduling a follow-up takes you 15-20 minutes of switching between apps.
Your agent does the same thing in one sentence and 10 seconds.
Multiply that by the 10-15 multi-step tasks you do daily, and you're saving 2-3 hours every day. That's not a small optimization — that's a different way of working.
Final Check
What's the user's role in a multi-step task?
A
The user has no role — it's fully automated
B
Describe the desired outcome and review the result — the agent handles the steps
C
Write out every step for the agent to follow
D
Monitor each step and approve individually
Your job is to say what you want. The agent's job is to figure out how. You review the result and redirect if needed. That's the ideal division of labor between human and agent.
⚡
Day 13 Complete
"One instruction, many actions. Your agent chains tools together so you don't have to."
Tomorrow — Day 14
Build Your Morning Briefing
Time for your first mini-project — build a daily briefing that combines everything you've learned.