THE SCALES LAB
Designing an AI product to feel like a real person, not the bot-growth services it's competing against.
B2B
AI AUTOMATION, BRAND & WEB DESIGN
Role
Product Designer
Timeline
10 weeks
team
1 Developer, 1 PD, me
platform
Web
The Brief
The Client, founder of The Scales Lab, reached out needing a website that could do double duty: position an AI automation agency as credible and professional, and actually convert visitors into booked calls.
He came to the project with a starting point, not a finished plan. A Miro board laid out a rough page structure, and his brand direction was two words: black and lavender purple. The site would be built in Framer, with Zapier webhooks already handling the landing page automations behind the scenes.
There was also an existing website, one client had built himself, and he was happy to leave it behind entirely rather than have me match it, which gave me real room to rebuild the structure and system from scratch.
Underneath the visual brief was a harder positioning problem. The Scales Lab runs multiple offerings under one umbrella, Amplify for individual businesses, Amplify as a white-label product for agencies, and a separate AI Automation service, each solving a different problem for a different audience. Amplify itself also sits in a category full of disreputable, bot-driven growth services, and client wanted no association with that kind of fake-growth positioning. Getting that distinction across visually, not just in copy, was the real design problem underneath the surface brief.
The UX Process
I started from the base client gave me: a Miro wireframe and a two-word color direction. No brand guidelines, no finished copy, no component library. My first deliverables were a strategy document mapping out positioning and a sitemap, before any visual design started, so we were aligned on structure before I touched Figma.
Developing the brand direction, not just applying it. Black and lavender was a hint, not a palette, so I proposed a specific dark, lavender-to-violet system, typography, spacing, and a card component library. Client loved the direction and started using the updated colors in his own social content before the site had even launched.
Restructuring the sitemap after real feedback. My first sitemap pass didn't give AI Automation its own page, it read as secondary to Amplify. Client flagged that it needed to stand alone as its own service, so I added it as a dedicated page, kept deliberately small and simple since it's a secondary offer, not the main event.
Splitting Amplify into two audiences. Amplify serves two very different buyers, individual businesses running it on their own account, and agencies white-labeling it across multiple clients. After reviewing the sitemap, client asked for two clear sub-paths: land on /amplify, see what it is and how it works, then branch into a dedicated Business page or a dedicated Agency page, each with its own pain points, benefits, and pricing logic. That branching structure became one of the site's core navigation patterns.
Designing trust signals to counter a real brand risk. Since the biggest positioning threat was reading like a bot-growth service, I leaned on patterns that make the product feel human and credible: real-looking dashboard and analytics previews instead of vague icons, named testimonial cards with titles attached, before-and-after comparison blocks that make the value concrete rather than promotional, and a client logo marquee placed high on the homepage, a trust signal client wanted visible early in a visitor's journey.
Designing ahead of the copy. Client wanted to write the final copy himself and was still refining positioning as we worked, so most of the site was designed and structured against placeholder text first, with his real copy dropped in later once it was ready.
What Actually Happened
I wireframed the new flow and tested it with our internal clinical expert and customer success team (who talked to therapists daily). That's where I caught my mistakes.
For example: I'd grouped 'treatment goals' and 'session interventions' together because they seemed related. But therapists think about them at totally different points—goals get reviewed before the session, interventions get documented after. Keeping them separate made more sense.
The auto-save design also evolved. My first version showed a brief 'Saved' notification that disappeared quickly. Beta testers said they didn't notice it and still felt anxious. I changed it to a persistent 'Last saved [time]' indicator that updated live. Small tweak, massive difference in confidence.
We shipped to a small beta group first. The response was immediate—therapists noticed the progress bar and auto-save within minutes. We adjusted some section labels based on feedback, then rolled it out to everyone.
What Changed
The site is live now at thescaleslab.com. I don't have hard site metrics to report yet, but the clearest signal of impact came directly from the client. Client's reaction to the design system was immediate, he called the colors spot on and started using the palette in his own social content before launch, and he'd already signed off enthusiastically on the strategy and sitemap work before visual design even began.
The strongest outcome, though, wasn't on the site at all. Off the back of this project, Client offered to pass my resume on to business owners in his network for full-time roles, and started recommending me directly to contacts for freelance UI/UX work, several of whom have projects coming up in the next few months.
What I'd Do Differently
I'd try to lock the two-audience split on Amplify (Business vs. Agency) earlier in the sitemap phase rather than after the first review round. It was the right call once we got there, but starting from a single Amplify page meant reworking navigation and page structure mid-process instead of designing for it from day one.
Designing most of the site against placeholder copy worked, but it meant a few layouts had to flex once Finn's real copy landed. Next time I'd ask for even rough, messy draft copy earlier, just to stress-test whether the structure actually held.
What I Learned
When a product sits in a category people are already skeptical of, like AI growth tools next to bot-follower services, design has to do as much trust-building work as the copy. Real testimonials, real-looking product data, and a restrained, premium visual tone aren't decoration, they're the argument.
The best client input isn't always a finished brief. Finn gave me two words and a Miro board and stayed genuinely open the whole way through, which meant real progress depended on me proposing strong opinions early rather than waiting to be handed direction.