
Design agency pricing models: fixed price or retainer?
Design agency pricing models: fixed price, sprint or monthly retainer, when each fits a founder, what goes in the contract and the risk of each.
17 Nov 2026
AI SaaS website design: what buyers check after an AI recommends you, the four screens to show, the limits to state and why gating the demo costs you.

Good AI SaaS website design shows the real product: four actual screens, one demo under two minutes, the model’s limits stated plainly, and a price. Buyers arriving from an AI answer are checking whether the recommendation was right. Stock robot imagery, vague claims and a form in front of everything send them away.
An illustrative scene, though a common one. Direct traffic climbs with no campaign behind it. Someone asked an AI assistant for tools like yours and came straight to your homepage to check.
Semrush surveyed 622 US B2B professionals in March and April 2026. Of the 519 who use AI at work, 71% visit a vendor’s website after an AI mentions it and 63% search the company on Google. The assistant gets you on the list. Your site decides whether you stay there.
And they arrive sceptical. In TrustRadius’s 2026 B2B Buying Disconnect, a January 2026 survey of 1,862 verified technology buyers, 47% trust online resources less than they did a year ago, against 11% who trust them more.
They go looking for things they can check themselves: your product, your reviews and someone who has used it. An AI answer is a lead, and buyers treat it like one.
In the same TrustRadius survey, 74% of buyers consulted user reviews and 53% spoke with a peer who had used the product. Semrush found 38% go to G2 or a similar review site after an AI recommendation. HG Insights, which publishes that research, reports that 72% of buyers who used AI fact-check its output always or very often, up from 58% the year before.
So your site’s job is narrow. It confirms what the assistant said, with evidence a sceptic will accept. Our note on what the GEO paper says about getting quoted by AI covers getting named in the first place. Getting believed is the other half.

What should an AI product page actually show? A summary of this section.
Four real screens, a demo you can watch without booking anything, the model’s limits and a price. Put those ahead of anything clever.
Real screens, with real data.
A browser frame with a gradient in it doesn’t count. Show the screen your best customer looks at on a Tuesday, with labels a buyer recognises.
A demo under two minutes.
Consensus, reporting in May 2026 on more than 6 million buyer interactions with its demo platform, found the average demo runs 15 minutes 14 seconds while the average buyer watches 5 minutes 14 seconds. Cut yours to the part people actually watch.
What the model does and doesn’t do.
Where it gets things wrong and what a human still checks. One honest paragraph beats a page of adjectives.
A price, in the visitor’s currency.
A number, a range or a starting point. If you sell in the US, the UK and Europe, show dollars, pounds and euros.
Proof a stranger can verify.
Named customers, review links, a security page, a status page. Logos nobody can click are decoration.

What to show and what to cut on an AI product page. Framework by Dtail Studio.
Pictures of robots, claims with no number attached and a page where every link opens a form. Roughly in that order.
Stock AI imagery is the cheapest sign there’s nothing to show. Glowing brains and blue heads made of dots suggest the product wasn’t ready for its photo.
Vague claims cost more than they seem to. Semrush found 33% of buyers say AI recommendations are too generic for their use case, so visitors land already hunting for specifics. Gartner’s survey of 632 B2B buyers, fielded in late 2024, found 69% see inconsistencies between what a supplier’s website says and what its sellers say. Every unsupported line widens that gap.
Gating everything is the expensive one. The same survey found 61% of B2B buyers prefer a rep-free buying experience. A form in front of the demo asks them to do the one thing they’ve already decided not to do.
Our advice, which makes some founders wince: publish what your model gets wrong, and what happens when it’s unsure. Buyers who find the limits on their own assume you hid them. Buyers who read them on your page start believing the rest.
Enough that a buyer can decide you’re worth a call without talking to anyone. TrustRadius found free trials or accounts were rated a most influential resource by 77% of buyers, and product demos by 73%, ahead of every sales tactic. The change buyers most wanted from vendors was transparent pricing, at 59%.
Leave the tour, screens, limits and pricing open. Keep the form for a sandbox loaded with the buyer’s own data, which is a fair exchange. Our post on why your pricing page is your second homepage says more about the money page, and most startup websites explain too much covers what to cut.
Complex data, shown as something a person can act on. Cinelytic is an AI-powered platform that helps the entertainment industry turn complex data into smarter decisions. Teams use it to explore talent, evaluate projects and forecast a film’s potential, and its case study page says Warner Bros., Sony Pictures and Lionsgate trust the platform. The design work covered complex workflows, dashboards and data visualisations, with a flexible design system that keeps things consistent as the product adds capabilities and use cases. “Effective workflow, smooth communication, and remarkable talent across multiple disciplines,” says Dev Sen, Founder of Cinelytic Inc.
Those dashboards are the asset. A product that reads clearly in one screenshot is a product you can sell from one screenshot.
We’d start from the questions buyers ask after an AI names you, not from the homepage layout. Usually like this:
Collect the real questions.
Sales calls, support tickets and post-demo follow-ups. That’s your page outline.
Pick the four screens.
The ones that prove the product works, filled with data that looks like a customer’s, not a test account’s.
Cut the demo to the watched minutes.
One path, one outcome, no narrator warming up.
Write the limits section.
What the model does, what it doesn’t and what a human reviews.
Open the pricing.
A number or a range per currency, and what moves it.
Bring your demo recordings, your pricing logic and the objections you hear twice a week. More on how we work.
Do we need a self-serve demo if we sell to enterprise?
Yes. Enterprise buyers read the same pages before they book, and far more people see your site than reach a call.
Should the homepage say we use AI at all?
Say what it does for the buyer first, then how. “Drafts the first version of every claim review” lands better than “AI-powered intelligence platform”.
Won’t publishing our limits help competitors?
Your competitors already know your limits. Buyers don’t, and they’re the ones deciding.
How do we know whether AI answers send us traffic?
Watch direct and referral visits from assistant domains, and ask on your enquiry form how people heard about you.
Open your homepage on a phone and give yourself 90 seconds. Write down what you learned about what the product does, what it can’t do and what it costs. Anything missing is next week’s work. Then ask an AI assistant your buyers’ question, see who it names and visit those sites with the same timer.
If you want a second pair of eyes on the page buyers land on, let’s talk.
About the author: Rori Thompson is a Content Writer at Dtail Studio and writes about product design, UX and the details that make an app easy to use.
Sources: Semrush, How AI tools shape the B2B buying process (Margarita Loktionova, July 2026), TrustRadius, 2026 B2B Buying Disconnect: beyond the hype (July 2026), HG Insights, 94% of B2B buyers fact-check AI research (Grace Wells, July 2026), Gartner, Sales survey finds 61% of B2B buyers prefer a rep-free buying experience (June 2025), Consensus, 2026 B2B Buyer Behavior Report (May 2026)

Design agency pricing models: fixed price, sprint or monthly retainer, when each fits a founder, what goes in the contract and the risk of each.

Energy management dashboard design for commercial buildings: show today’s decisions, savings in money and alerts that matter, not charts of kWh.

Dementia-friendly app design: one task per screen, high contrast, big labelled buttons and a clear “done”, plus a carer view that respects the person.