120+ B2B Teams Tried Softworker Before the MVP Existed

The Problem: Nobody Could Picture the MVP

We were a new AI startup with no demo, no customers and no logos. The website came first. We rebuilt it three times, and each version explained the idea better. But the response stayed weak.

When we talked to people, the reason was clear. They could not picture what the MVP would actually give them. And with AI agents, the question that matters is not “what can it do?” but “what will it do without me?”. No headline answers that.

Why a Tour, Not Just a Video

We first planned only a product video. A video is quick to make, but you sit back and watch. You never make a decision, so you never feel in control. For an agent product, control is the whole point.

A tour lets people do the important part themselves. I leaned on a few UX and business principles to shape it.

Choosing the First Use Cases

We wanted to win one vertical first. Our research pointed at finance and procurement: lots of repeat work, clear rules and many approvals, which is exactly where agents with approval gates help most.

Then we talked to friends and ex colleagues working in startups. One thing kept coming up: trusting AI with money is a hard first step. Nobody wants an agent to be the first thing that touches payments.

Marketing and sales felt different. The work is easy to see, a bad draft costs little and is easy to undo, and teams already use AI for copy. So marketing and sales became the first focus for the MVP.

The tour still has three use cases. Marketing is where most people start.

We kept finance and procurement in the tour on purpose. They show the approval gates on high stakes work, which builds trust for every team. For other kinds of work, Softworker comes with templates and skills for different workflows.

Inside the Tour

Here is the Marketing tour, start to finish. A request comes in from the growth team, an agent picks it up, shows its plan, stops before a risky step, and finishes once you approve.

1. 

   A request arrives. Work lands in a shared inbox, the way it would from Slack or email.

2. 

   An agent picks it up. It writes as it works, and a teammate can step in with a note.

3. 

   Its plan is visible. You see what it will do before it does it.

4. 

   Every action is logged. Each step shows where its information came from.

5. 

   It stops and asks. Before the risky step, a person reads the reason and decides.

6. 

   Done, with proof. The approval and the evidence stay in the thread for later.

Key Design Decisions

Decision 01

A guided tour, not an empty sandbox

A blank workspace asks people to work out where to start. Each step has a short tip that points at the next click, with Back and View All Tours if you want to look around.

Decision 02

The approval gate is the hero moment

Every tour leads to one moment: the agent stops and asks you. The card shows the reason, the evidence, and what happens if you approve or reject. This is where people stop worrying about AI acting alone.

Decision 03

A busy company, honestly labelled

An empty product feels fake. The tour has 8 workspaces, 10 agents, 10 skills and 15 knowledge files, with believable tasks and risk levels. Every action also says “Prototype action only; no external system is contacted”, so nobody feels tricked.

Decision 04

The waitlist is always one click away

Join Waitlist sits in the top bar on every screen of the tour. Whenever someone is convinced, they can act. On the website, “Experience Product Tour” sits in the menu and the hero.

How I Built It

I mapped each use case and its steps in Figma, explored visual ideas in Flowstep, and wrote the scenarios and built the tour with Claude. We went through a few rounds of changes before landing on this version.

• The actual product, sample data. The tour uses the same screens the MVP would use, filled with made up companies, people and tasks.
• Scripted agents. Messages stream in as if the agent is working, so it feels alive without a backend.
• One link. demo.softworker.ai/product-tour works on its own, so we could share it anywhere.

Before the Tour: The Website, in Short

The website came first. Each version taught us something, and the last one became the front door to the tour.

Version 1

Nobody joined

Big enterprise words, no proof and too much to read. After a week, the waitlist was empty.

Version 2

Clearer, still flat

One honest page, built in 3 days with AI. People could read about Softworker, but they could not see it.

Version 3

Let them try it

A plain headline, the product itself in the hero, and “Experience Product Tour” in the menu and the hero.

The whole website story, from version 1 to version 3, is in this short film.

We also made a separate 90 second product video for LinkedIn and emails, for people who would not click through a tour.

Try It Yourself

Here is the tour, live. Pick a use case and follow the tips. It takes about 3 minutes.

Start the live demo

Works best on a laptop. On a phone it opens in a new tab.

Results

From an empty waitlist to 120+ B2B teams in 4 months. Many of them joined after the tour went live. Two became design partners, and both went through the tour before they signed. The company names stay private.

We did not wait for traffic. We found the right people on LinkedIn, wrote to them one by one, and shared the tour by hand. That is why our visitor numbers are small and our waitlist isn't.

Microsoft Clarity for the website, 1 May to 27 September 2026

What the Tour Kept Doing

The tour did not stop at the waitlist. It kept working long after people joined.

• Sales calls. We click through the use case that fits the team, live, instead of showing slides.
• Investor conversations. It shows the product vision more clearly than a deck.

What I Learned

• Let people try it. For AI agents, trust comes from pressing Approve yourself, not from reading about it.
• You can ship the experience before the product. A tour with fake data taught us what to build first.
• Pick the first use case people are ready to trust, not only the biggest one.
• A great demo still needs the right people to see it. Outreach one by one filled the waitlist.