Most AI website builders will give you a good-looking homepage in a few minutes. That is not the hard part. The hard part is getting a site that knows who it is for, explains one clear offer, captures leads, takes payment, and still reads like one business after a dozen rounds of edits.

This guide walks through a practical workflow for building a service business website with Runable, an AI agent that can do research, build websites and apps, and produce documents from one workspace. The example is a consulting-style offer, an “AI Growth Audit” that helps businesses understand what is limiting their search and AI visibility. The same process works for most solo consulting, freelance, or productized service offers.

Prefer to watch? See the full walkthrough on YouTube. I build the whole thing on screen.

Why the Workflow Matters More Than the Tool

If you ask an AI tool to “build me a website,” it will guess at your audience, your offer, and your tone. Then each revision pulls it in a new direction. One prompt asks for a premium look, the next asks for something playful, a third quietly adds a service you never offered. After ten revisions, the site no longer feels like one business.

The fix is to work in a fixed order:

1. Research the market and the customer.
2. Plan the offer, positioning, and page structure.
3. Build a complete first draft.
4. Add business systems such as sign-in, payments, forms, and email.
5. Optimize, publish, measure, and improve.

Runable is useful here because research, planning, and the build can happen in the same workspace, so the research files can feed straight into the site. But the discipline is what makes the output usable. The tool just makes each step faster.

Step 1: Write a Master Project Brief

Before you touch a website builder, write one detailed brief that acts as the source of truth for the whole project. You will reuse it when you build the site, when you create marketing assets, and when you decide whether a new idea actually fits the business.

You can draft it in ChatGPT, Claude, or any assistant you already use. In the video, the brief covers six things:

• Audience: who the site is for, as specifically as you can describe them.
• Problem: what is going wrong for them right now.
• Offer: what you sell and what is included.
• Key promise: the outcome they can expect, stated without exaggeration.
• Conversion goal: the one action you want visitors to take, such as “Get your audit.”
• Design direction: how the site should feel. For a specialist advisory service, that means clear typography, restrained color, generous spacing, simple calls to action, and sections that show proof of process.

Add constraints at the end of the brief

This is the part most people skip. Tell the AI what it must not do:

• No exaggerated claims.
• No fake testimonials.
• No invented statistics.
• If information is missing, use a clearly marked placeholder instead of making something up.

That last rule is the one that saves you. A placeholder that says [Add client result here] is easy to spot and fix. A plausible-sounding invented number can end up live on your site and in front of a prospect.

Step 2: Run Market Research With a Purpose

With the brief in hand, the next step is research. The goal is not a pile of competitor links. You want research that helps you decide on the offer, pricing, messaging, page structure, SEO strategy, trust signals, and customer journey.

What to ask the research agent to cover

• Direct competitors offering the same service.
• Adjacent services competing for the same budget. For an audit offer, that includes SEO agencies, AI visibility platforms, website audit tools, and conversion consultants.
• What customers complain about and what they value.
• Objections that come up repeatedly.
• The terminology customers already use.
• Which customer segments look like the strongest fit.

In the video, I pasted a detailed research prompt into Runable and let it run. It came back after about 25 minutes and 19 steps with seven files, including a long market research plan, a competitor database with positioning and headlines, a pricing comparison, and an SEO opportunity map. Runable reported using about 2,300 credits for that research run.

Treat AI research as a first draft, not a fact sheet

The research output included competitor prices and a suggested price range for testing the core audit. That kind of output is a useful starting point for your own pricing thinking. It is not something to quote on your site or in a proposal until you have checked it. Before you rely on any number:

• Open the source the agent cites and confirm the figure is still on the page.
• Check the date. Agency pricing pages change, and many do not publish prices at all.
• Separate what a competitor actually publishes from what the agent estimated or inferred.

Step 3: Generate the First Version of the Site

Once the research is done, Runable offers to turn the report into a website or a deck. It may also suggest a format. In the video, it recommended an interactive research site. I chose a customer-facing marketing website instead, because that is what the business needs.

Runable can also show you design templates to start from. Templates are fine for speed, but if you already have a strong brief, you will usually get a more coherent result by skipping the template. Paste in your master brief and tell the agent to combine it with the research it just produced.

What a first draft can include

In the video, the build took about 20 minutes and produced:

• A marketing site and a customer portal.
• A built-in chat assistant for visitor questions.
• Calls to action to get the audit or book a consultation.
• A “what is included” section and an FAQ, as the brief requested.
• Google sign-in that worked out of the box.
• A markdown configuration and launch checklist listing environment variables and the setup still needed.

I show this step on screen in the video, from the first prompt to the finished draft.

Just as useful is what the first draft did not do. Payments were not implemented yet, so Stripe still had to be connected. There was no pricing page. The contact form was probably not wired up to send email. That is normal for a first pass. Read the launch checklist the agent generates, because it tells you exactly what is still placeholder.

Step 4: Review the Draft Like a Buyer Would

A working first version is a minimum viable product, not a finished site. Before you ask for changes, review it against a short checklist so your revision requests are specific.

Hero section

• Is the headline easy to scan?
• Does the supporting copy explain the scope of the offer?
• Does the primary call to action stand out?

Visual hierarchy

• Are section headings clearly different from body text?
• Are important calls to action easy to find across the page, the navigation, the chat, and the footer?
• Is there enough spacing between sections?
• Does every section have a job, or are some only decoration?

Trust without fake proof

If you are not using testimonials, and you should not invent them, build trust by making the process concrete. Show what the deliverable includes, what the customer needs to provide, how the review works, and what happens after they receive the report. A new service with a clear, transparent process often reads as more credible than one padded with generic praise.

Copy

AI-written web copy tends to repeat itself. Watch for the same promise stated three ways across the hero, the services section, and the final call to action. Each section should add new information.

Step 5: Refine With Specific, Numbered Requests

“Make it better” forces the AI to guess what better means. Point to the part of the page and describe the outcome you want. This is the refinement prompt used in the video, lightly condensed:

Refine the homepage without changing the core offer. 1) Make the hero simpler and more credible; the headline should emphasize identifying what is limiting search and AI visibility. 2) Make “Get your audit” the strongest visual CTA. 3) Reduce repetitive copy across sections. 4) Improve spacing and visual hierarchy so the page is easier to scan. 5) Strengthen trust using process transparency. 6) Remove decorative elements that do not help comprehension or conversion. Do not add invented proof or claims.

The revised homepage had a direct headline (“Identify what’s limiting your search and AI visibility”), a dominant primary CTA with the consultation as a secondary option, fewer decorative charts and graphics, a simple sample report preview, and a clearer outline of scope, customer input, review stages, and report handoff.

Then you repeat the loop: build a complete draft, inspect it, find the highest-impact weakness, fix it, and inspect again. Work on one or two problems per round so you can tell what each change actually did.

Step 6: Connect the Business Systems Before Launch

A site that cannot take payment or deliver a form submission is a brochure. Before you send traffic to it, work through the launch checklist:

• Payments: connect Stripe and test a real checkout, ideally in test mode first.
• Forms: submit the contact or inquiry form yourself and confirm it reaches the right inbox.
• Accounts: sign up as a new user, log out, and log back in.
• Email: check that confirmation and follow-up messages send and do not land in spam.
• Placeholders: search the site for any placeholder text the brief told the AI to leave.
• SEO basics: page titles, meta descriptions, and a sensible URL structure.
• Analytics: confirm visits and conversions are actually being recorded.

Runable’s pricing page lists Stripe integration, a free database, free analytics, and a free subdomain under its paid plans for website and app building. Plan features change, so confirm what your plan includes before you count on any of them.

Common Mistakes to Avoid

• Skipping the brief. Without a source of truth, every revision drifts and the site loses its identity.
• Trusting research numbers as is. Competitor prices and market figures from an AI agent need to be checked at the source before they shape your pricing or show up in your copy.
• Vague revision prompts. “Improve the design” wastes a round. Name the section, the problem, and the outcome.
• Letting the AI invent proof. Fake testimonials and made-up stats can cost you credibility with a prospect, and in some places they can create legal risk. Use placeholders and fill them with real results later.
• Mistaking a pretty draft for a working site. Check payments, forms, and email by hand before launch.
• Over-decorating. Animations, gradients, and icons should support the offer, not compete with it.
• Ignoring credit usage. Long research and build runs consume credits. Watch your balance on the first project so you know what a full build costs you.

FAQ

What is Runable?

Runable describes itself as an AI general agent that plans and executes tasks end to end. Its feature list includes website and app building, document and research work, slides, and image creation, with access to several AI models. Check the current site for the latest feature set.

How much does Runable cost?

When this article was written in October 2026, Runable listed a free plan with daily credits plus paid Pro and Max plans with larger monthly credit allowances. In the video, I mention credit top-ups starting at $10 for 10,000 credits. Pricing changes often, so check runable.com/pricing before you budget.

How long does it take to build a site this way?

In the video, the market research took about 25 minutes and the first full site build took about 20 minutes, not counting the time spent writing the brief and reviewing the output. Your times will depend on the size of the project and how many refinement rounds you run.

Do I need to write code?

Not for the workflow shown here. Everything was done through prompts and the built-in preview. You will still need to handle configuration, such as connecting Stripe and email, and you should test those flows yourself.

Can I use this process with other AI website builders?

Yes. The master brief, the constraints, the review checklist, and the numbered refinement prompts work with any capable AI builder. Runable’s advantage in this example is that research and building happen in the same workspace.

Should I let the AI set my prices?

Use its research as one input. Verify competitor prices yourself, then test a price with real prospects. The agent’s suggestion is a hypothesis, not a decision.

Conclusion

An AI agent can take you from a business idea to a working first version of a service website in an afternoon. The quality of what you get depends far more on your process than on the tool. Write a master brief with clear constraints, research with a specific purpose, verify what the research tells you, review the draft against a checklist, and refine with specific requests before you connect payments and launch.

Your next step: write the six-part brief for your own offer (audience, problem, offer, key promise, conversion goal, and design direction) and add the “no invented claims, use placeholders” constraints. Then watch the full walkthrough above and run your first research pass.

If you’d rather not figure this out alone, join AI Income Lab for the courses, templates and help that go with these builds.