Summarize with AI
A B2B website that generates qualified demos is built to turn people away. It names who the product is for and who it is not for, gives a price signal, asks one question on the form that sorts buyers from browsers, and offers every wrong-fit visitor a door that does not lead to the founder's calendar.
For a sales-led AI startup this is the main job of the site. Nobody signs up and tries the product alone, so every visitor who wants to know more ends up on a call. You know within a minute whether a caller can buy. The page has never been told, so it books them all, and you do the sorting yourself, thirty minutes at a time.
Sixty demo requests in October, twenty worth taking
Picture an AI month-end close tool, used here as an example. It pulls the trial balance from every legal entity a company runs, matches the invoices those entities send each other, flags the pairs that do not net to zero and drafts the elimination entries for the controller to approve. It is built for finance teams at companies with five or more entities, usually after an acquisition or a move into a second country. It is sold on calls, because it only shows its value on the buyer's own ledgers.
The homepage hero reads The AI finance team for modern companies. The nav has four items: Product, Customers, Blog and Book a demo. There is no Pricing link, no Careers page and no Partners page. The demo form has three fields, Name, Work email and Message, and after it comes a calendar with every open slot. The most read page on the site is a blog post called Accrual vs cash accounting: a simple guide, and like every post it ends with a purple banner that says See the AI finance team in action, with a Book a demo button.
In October, sixty people booked. The founder fills in a column called Fit in the CRM after each call. Here is how the sixty sort:
- 20 controllers and finance leads at companies with five to forty entities. These are the calls the company exists for.
- 14 founders of single-entity startups looking for something cheaper than their bookkeeper.
- 9 accounting firms and outsourced bookkeeping shops asking about partner pricing or a white-label version.
- 7 students and course learners who arrived from the accrual post.
- 6 job seekers, because Book a demo is the only form on the site.
- 4 vendors pitching their own services: an SEO agency, two dev shops and a lead generation firm.
Two thirds of the calendar was never going to buy. None of them did anything wrong. Each one read the site, found nothing that said this is not for you, and pressed the only button it offered.
Why the founder takes all sixty himself
Now look at the founder's month. He takes every call, because his cofounder is building the product and nobody else can run the demo. Before each one he opens the booker's LinkedIn profile. He has a saved email reply titled Not a fit, kind version, and he pastes it after most calls with a link to a free template. The accounting firm calls run the longest, because the questions are good, and he has agreed to a partner call next month that nobody on the team owns.
On Mondays he opens a dashboard with one big number on it: demos booked. It has gone up three months in a row. He tells his cofounder that demand is strong. Meanwhile the next open slot on his calendar is eight working days out, so a controller who books today waits more than a week for the call that could close.
None of this looks like a website problem from his chair. It looks like a busy pipeline. But he is the only person who knows that single-entity companies cannot use the product, and the site has never said it.
What forty wrong calls cost a sales-led startup
Count the hours first. Forty calls at thirty minutes, plus ten minutes each for the lookup and the follow-up, is close to twenty-seven hours. That is more than three working days a month, taken from the one person who can close the other twenty. It is also the reason the twenty wait, because the slots they need are full of people who cannot buy.
Then look at how the pipeline reads to an investor. Sixty demos and a handful of deals looks like a product that does not close, when it is really a page that does not sort. For AI startups the Series A bar is around 3.5 million in ARR, up from roughly one million three years earlier (Carta, Q1 2026). Across all sectors, roughly 70 percent of seed companies never raise a Series A. A close rate dragged down by students and job seekers is a weaker story to tell in that room, and the cause sits on the website.
Most advice counts demos. Count the people you turned away
Most advice on B2B websites treats demo volume as the score. Cut form fields, add more buttons, put Book a demo on every blog post. That advice comes from companies with a sales team to absorb the volume. A seed-stage founder has one calendar, and every extra booking is paid for in his hours.
So here is the claim. For a sales-led startup, a site change that takes demo requests from sixty a month to thirty, while the twenty buyers keep booking, is a better site, even though the number on the Monday dashboard fell by half. The score that matters is the share of booked calls that could buy, and the time from booking to call for the people who can.
Turning people away does not make the company look smaller, either. A page that says who it does not serve reads as a company with enough demand to choose. Established vendors write like that. Early ones tend to write as if any visitor might be the one, and buyers can tell the difference.
Five doors that do not lead to the founder's calendar
The fix is not one better headline. Each of the forty came in through a different page, so each group needs its own sentence and its own exit. Here is the close tool's site rebuilt as a filter.
- Name the buyer by the thing they count. Finance teams know how many entities they close each month, so the first lines say month-end close for companies running five or more entities, with a short not-for line under it. Writing that line is covered in how to show who your product is for on the homepage. The SATHI launch site we designed does the same in its first lines, which name its buyer as Shopify brands turning affiliate into a primary revenue driver. A brand that does not sell on Shopify knows at once the page was not written for it.
- Give a price signal, even without a number. A Pricing page that says priced per entity, from five entities, billed yearly turns away most of the fourteen single-entity founders before they book. Add one honest line: if you run one entity, your bookkeeper is the better tool for now.
- Ask one question on the form that predicts fit. How many legal entities do you close each month, with four choices: 1, 2 to 4, 5 to 20, 21 or more. One entity goes to a free close checklist. Two to four goes to a recorded walkthrough and an offer to talk when they add a third. Five or more sees the calendar. The rest of the form is covered in demo request page best practices.
- Give the people who are not buyers their own door. A Partners page with a short form that goes to a separate inbox, for the accounting firms. A Careers page, even if it only says how to send a note when there are no open roles. A footer line for vendors and press with an email address. And on the accrual post, swap the demo banner for the close checklist, which is what a student or a learner actually wants.
- Write the pages the buyer forwards. The controller does not buy alone. Her CFO, her auditor and her IT lead each have one question, and a call is the slowest way to answer it. A page that walks through a close day by day, and one that shows an approved elimination entry with who approved it and when, travel by link. The pages your first sales hire needs covers the same idea from the sales side.
Notice that every one of these says no politely. The student gets a checklist. The accounting firm gets a partner form. The single-entity founder gets an honest reason and a cheaper answer. Nobody is insulted, and nobody needs thirty minutes of the founder's day to hear it.
A one-week sort of your last sixty requests
You can do the first pass this week without redesigning anything. You need your CRM, your analytics and an afternoon.
- Export last month's demo requests, with whatever notes you wrote after each call. Sort them by who they were, not by fit or no fit. Students, agencies, too small, job seekers and vendors are separate piles with separate fixes.
- For each wrong pile, find the page they came from. The first page of the session in analytics, or the referrer field in the CRM, is usually enough. Expect one or two pages to explain most of a pile.
- Write the sentence that would have told that pile no, kindly, and where to go instead. Put it on that page, near the button they pressed.
- Pick the one count that predicts fit for your product, such as entities, sites, seats or monthly volume, and make it the single routing question on the form.
- Change the Monday number from demos booked to qualified demos held, and add the days from booking to call. Tell your cofounder the first number will fall on purpose.
Recount after a month. If the buyers still book and the other piles shrink, the site is doing its job. If the buyers drop too, the not-for line is too wide, so loosen it before you touch anything else.
The reason this is hard to do alone is the same reason the site books everyone. You know who fits, so you read every line on your own pages with that knowledge filled in. Studio Maydit designs and builds websites for AI founders in Framer, Webflow and custom code, on a fixed scope of three to four weeks. For a sales-led company the site is the whole revenue surface, so its brief should start from who should book, who should not, and where each of them goes instead. If your calendar is doing the sorting right now, book a call with us.
