How to Build Your First Automated Workflow in Under 10 Minutes

If you're running a business by yourself or with a small team, you already know the struggle: there are never enough hours in the day. You spend too much time just keeping up with the business (supporting customers, paying bills, maintaining inventory, etc) when you could be focusing on growth instead.

The good news? Automating these tasks is easier than you think, and it doesn't require a computer science degree or expensive software.

Why Automating Repetitive Tasks Pays Off

The time savings aren't even the biggest benefit of automation.

Yes, automating a task that takes 20 minutes per day saves you 120 hours per year. That's valuable. But the real transformation happens when you stop thinking about those tasks. When you're not mentally tracking whether you remembered to follow up with that lead, or worrying about whether the invoice got sent, or wondering if you missed an important email. Research on cognitive load and task-switching consistently shows that the mental overhead of tracking open tasks often costs more productivity than the tasks themselves. Cal Newport's work on deep work frames it similarly: the value of focused, uninterrupted work is destroyed long before you sit down to do it, by the anticipatory anxiety of unfinished tasks in the background.

That mental overhead, the constant context-switching and task anxiety, is what actually kills productivity. Automation eliminates it completely so you can focus on what matters.

What You'll Need

Before we dive in, here's what you'll need to get started:

1. A Rills account - Sign up for an account at rills.ai. You don't need a credit card until you select a plan and start your free trial.
2. A use case in mind - Think of one repetitive task that frustrates you regularly. Good first candidates:
   • Triaging customer support emails
   • Qualifying new leads from your contact form
   • Following up on pending invoices
   • Summarizing daily Slack conversations
   • Updating project status in your CRM
3. 10 minutes - That's genuinely all the time you need for your first workflow.

Don't overthink the use case. Start simple. You can always build more complex workflows later.

Step 1: Define Your Trigger

Every workflow starts with a trigger, the event that kicks off the automation.

In Rills, triggers can be:

• Time-based: "Every Monday at 9am" or "Daily at 6pm"
• Event-based: "When a new email arrives" or "When a form is submitted"
• Webhook-based: "When my CRM creates a new lead"
• Manual: "When I click the Run button"

For every new workflow, we recommend starting with a manual trigger. This lets you test the workflow on-demand without waiting for a specific event. You can always add additional triggers later when you've validated that the workflow is doing what you want.

Example: Let's say you want to automate the process of qualifying new leads from your website's contact form. Your trigger would be "Manual" for now, and you'll run it once you have a lead to process.

In the Rills dashboard:

1. Click "Create Workflow"
2. Give it a name: "Qualify New Leads"

That's it. We automatically add a manual trigger to every new workflow.

Step 2: Add Your AI Agent

This is where things get interesting. Instead of writing complex if/then rules, you describe what you want the workflow to accomplish in plain English.

In the workflow builder:

1. Drag an "AI" node from the node palette on the left onto the canvas
2. Connect the right output handle of the "Manual trigger" node to the left input handle of the "AI" node
3. Click the "AI" node on the canvas to configure it
4. In the "System Prompt" field, define the agent's role and what a good lead looks like for your business. See this example:

You are a lead qualification specialist for a freelance brand strategy consultancy. We help founders and marketing directors of small consumer product brands (food, beverage, beauty, lifestyle) define their positioning, messaging, and visual identity.

Our ideal client:
- Company stage: Pre-launch to Series A (revenue $0–$5M)
- Decision maker: Founder, CEO, or Head of Marketing (someone with authority to greenlight a project)
- Pain: Struggling to stand out in a crowded market, inconsistent brand across channels, or preparing for a retail pitch/fundraise and need a polished brand story
- Project budget signal: Mentions an upcoming launch, investor deck, trade show, or retailer meeting; these signal urgency and real budget
- Bad fit: Enterprise brands with in-house creative teams, agencies looking to white-label our work, or anyone asking for logo-only work with no strategic component

A Hot Lead has a specific deadline or event driving urgency (e.g. "we pitch to Whole Foods in 6 weeks"). A Warm Lead has a genuine brand problem but no clear timeline. A Cold Lead is vague, out of scope, or clearly price-shopping.

5. In the "Prompt" field (the per-execution instructions), reference the incoming lead data using variables from earlier steps in the workflow. For a manual trigger you would supply these manually, but when this is eventually hooked up to a form submission, an email, or a CRM's webhook, they would come from those steps:

Qualify the following inbound inquiry and determine how well this prospect fits our ideal client profile.

Inquiry details:
- Name: {{ lead.name }}
- Email: {{ lead.email }}
- Company / Brand: {{ lead.company }}
- Their role: {{ lead.role }}
- How they found us: {{ lead.referral_source }}
- Message: {{ lead.message }}

Evaluate this prospect against the ideal client profile in your instructions.

Respond in the following format:

CATEGORY: [Hot Lead | Warm Lead | Cold Lead]

REASONING:
[2–3 sentences explaining why this prospect fits or doesn't fit. Be specific; reference details from their message.]

RECOMMENDED ACTION:
[One sentence describing the next step. e.g. "Book a discovery call this week (mention the Whole Foods timeline)", "Send our brand audit questionnaire to assess readiness", "Politely decline; out of scope"]

The AI will now analyze each lead according to your criteria. Notice you didn't write any code. You just described the task in plain English, exactly like briefing a human assistant.

A note on prompt structure: The System Prompt sets the stage once, it's the agent's "job description" and never changes. The Prompt runs on every execution and pulls in live data via {{ variables }} passed from your trigger. Keeping these separate makes your prompts easier to tune over time. Anthropic's prompt engineering guidance echoes this separation: clear role context in the system prompt, task-specific instructions per execution.

Pro tip: The more specific your System Prompt is about what a good lead looks like, the more consistent your results will be. Vague criteria ("good fit") produce vague outputs. Concrete criteria ("mentions replacing a specific tool") produce actionable ones.

Step 3: Set Your Review Preferences

Here's what makes Rills different from traditional automation tools: you decide what runs automatically and what needs your sign-off.

You can now add a "Human Review" node onto your canvas after any step. This node can determine a confidence level of the workflow's execution up to that point and based on its configuration route to your mobile phone for review. You set the threshold for what requires your oversight.

For the lead qualification step:

1. Click on the "Human Review" node to open the configuration panel
2. Find the "Review Threshold" field
3. Set it to 80%

This means:

• If the AI is 80% confident or higher, it proceeds automatically
• If the AI is below 80% confident, it pauses and asks for your review

When you're starting out, we recommend setting thresholds high (80-90%). As you see the AI making good decisions, you can lower them to reduce manual oversight. Each review request that goes to your phone includes its confidence value so you can get a sense for what an appropriate threshold looks like. Rills will also suggest changes over time to improve confidence and adjust the workflow design to increase quality over time.

What review looks like: When a step needs review, you'll get a mobile notification. Tap it, review the AI's proposed action and reasoning, then swipe right to approve or left to reject. Each review takes about 5 seconds. You could also click into the card to suggest edits.

Step 4: Test and Iterate

Now it's time to see your workflow in action.

1. Click "Save" and then "Publish"
2. Manually trigger the workflow with the required data (a test lead's information)
3. Watch the workflow run

You'll see:

• The AI analyzing the lead
• Its categorization and reasoning
• The confidence score
• Whether it would have required your review

Did it make the right call? Great! If not, that's valuable feedback. Click "Edit" and refine your instructions to be more specific about what you're looking for.

Because instructions are plain English, you can iterate without debugging code.

Common First Workflows

Here are popular first workflows by business type:

For service businesses:

• Qualify inbound leads from contact forms
• Triage customer support requests by urgency
• Follow up with clients who haven't responded in 3 days
• Generate weekly client status reports

For e-commerce:

• Flag suspicious orders for manual review
• Send personalized follow-ups based on purchase history
• Update inventory across multiple platforms
• Process refund requests

For content creators:

• Summarize comments and feedback across platforms
• Identify collaboration opportunities in your inbox
• Schedule content based on engagement patterns
• Track mentions and respond to high-priority ones

For SaaS products:

• Onboard new trial users with personalized guidance
• Identify churn risk based on usage patterns
• Qualify demo requests
• Update CRM with product usage data

Pick one that resonates with your biggest pain point. The workflow you're excited to eliminate is the one you'll actually use.

What Happens Next

Once your workflow is running:

1. It learns from your reviews - When you approve or reject AI decisions, the system learns your preferences and suggest improvements to your workflows. For the full mechanics of how this works, see how confidence scoring improves over time
2. You reduce manual oversight - As confidence scores climb, you can choose to review less often
3. You add more complexity - Chain multiple steps together, add conditional logic, connect more tools

The goal isn't to automate everything on day one. It's to eliminate one annoying task, see the value, then expand from there with additional workflows or more steps. For more ideas on which workflows to tackle next, see the five high-impact automations every solopreneur should set up.

Common Questions

"What if the AI makes a mistake?"

That's exactly what the approval system prevents. High-risk actions get reviewed by you. Low-risk actions run automatically. You control the system.

"Do I need to connect my tools first?"

Not for your first test. Rills can work with manual input while you're learning. Once you're ready, connecting tools takes about 30 seconds per integration.

"What if I want to modify a workflow later?"

Workflows aren't set in stone. Click "Edit" anytime to update instructions, adjust review thresholds, add steps, or change triggers. Your past executions remain in the history.

"How much does this cost?"

Workflow Credits and AI Credits are what you pay for. The logic, approvals, and infrastructure are included. The base subscription includes usage credits and you can pay for additional usage with a price limit to prevent overspending if you want to. See our action credit pricing explainer and pricing page for the full breakdown.

Your Turn

You've just learned everything you need to build your first automated workflow. Here's your action plan:

1. Right now: Sign up for Rills and create your first workflow (10 minutes)
2. This week: Run it on real data and adjust the workflow based on results
3. This month: Identify your second automation opportunity

The hardest part is starting. Pick one task that annoys you every single day and automate it in the next 10 minutes.