Summarize with AI
A design partner program page for a startup should state the exchange in plain words: what a partner gets, what it costs them in time and data, and how many seats there are. Its job is to recruit five or so named companies and to turn everyone else away, so it needs qualifying questions, a seat count and a closing date, not an email field.
Here a design partner means an early customer who shapes the product before launch, not a design agency. For a pre-launch AI startup those few companies become the first paying customers, the first references and most of the next pitch. A page that collects everyone hides the five you need inside a list of people you will never sell to.
A Get early access button and 340 strangers
Picture an AI tool for licensed customs brokers. A brokerage receives a commercial invoice from an importer, and someone has to give every line on it a ten-digit tariff code before the entry is filed. This tool reads the invoice, suggests a code for each line, and shows the past ruling it relied on. A licensed broker approves every line. The product and its page are made up for this example.
The founders need five brokerages to use it on real shipments for four months. Their site has a navy hero with a slowly turning globe. The headline reads AI for global trade. The subhead says Built for the future of customs. One button sits under it, Get early access, and it opens a single email field. After someone submits, the field is replaced by a line of green text: You are on the list!
Nothing on the page says the word broker. Nothing says what an early access user would do, what they would get, or how many places exist. In seven weeks the list reaches 340 addresses.
What the founder did with the spreadsheet
He exported the list to a spreadsheet and sorted it by email domain. Most rows were personal addresses. A few dozen came from universities. Some came from software companies, and two came from startups building something close to his. He highlighted the rows that looked like brokerages in yellow. There were eleven.
He wrote forty personal emails anyway, because each name might be someone. Six calls got booked. One was a logistics student writing a thesis. Two were sales reps who wanted to pitch him their own tools. One was an importer who files her own entries and wanted it free. His Fridays went to these calls for a month. His investor update still said 340 on the early access list, and when an investor asked how many of them file entries for a living, he changed the subject to the model.
The cost is time, and time is runway. The median gap from seed to Series A is around 616 days across all sectors, and roughly 70 percent of seed companies never raise a Series A at all. The seed bar for product-led companies is now roughly 300 to 500 thousand in ARR. Design partners are where that first revenue and those first references come from. A month of Fridays spent on the wrong calls is a month the five right brokerages are not using the product.
Why a design partner page should turn most people away
Most advice treats the design partner page as a softer waitlist: a friendly headline, a short form, and as many signups as possible. Some advice says skip the page and recruit only through your network. Both miss what the page is for. The waitlist page is built to grow a number, and our guide to the waitlist landing page for an AI startup covers how to make that number mean something. A design partner page is built to shrink one.
The founder cannot see this from inside. He knows the tool is for brokerages, so AI for global trade reads to him like a description. A broker reading it cold sees a globe and a slogan and cannot tell if this is a freight app, a trade finance product or a news site. He knows his product and nobody else does, and the page was written for the person who already knows.
Saying no on the page is not rude. It is a kindness to the people you would otherwise book calls with and then disappoint. And it is the only signal a busy brokerage owner trusts. A program with five seats, a start date and a list of who it is not for reads like a team that knows its customer. A counter reads like a team still looking for one.
The exchange, written out where a buyer can check it
The heart of the page is one section with three parts, side by side on desktop and stacked on a phone. Here is the customs tool's version.
What you get. Use of the tool on your real entries, at no cost for the four months of the program. A direct line to both founders, with a reply the same working day. Your shipment types handled first: if your brokerage files mostly apparel, apparel rules get built before anything else. A founding rate locked for the first year after launch.
What it asks of you. One 45 minute call every two weeks. Twenty commercial invoices from last month, with client names removed, so the tool is tested on your real lines. One named person who will use it at least once a week. Honest feedback, including telling us when a suggested code is wrong and why.
Seats and dates. Five brokerages. Two seats are taken. Applications close on January 31. The program runs February to May.
Founders often leave out the middle part, because they worry the ask will scare people off. It does the opposite. A partner who sees exactly what is expected can say yes knowing what they agreed to. A program that hides the ask looks like either a sales trap or a team that has not thought it through, and a brokerage owner handing over client invoices will assume one of the two.
Four questions that do the filtering
The button under the exchange reads Apply for one of five seats. It opens a short form, and each question is there to sort, not to collect.
- How many entries did your brokerage file last month? Four ranges to tap, starting at under 100. The founders know the tool earns its keep above a certain volume, so the smallest range gets a polite note instead of a call.
- Which ports do you file at most? A short list plus Other. It tells the team which rules and ruling sets to load first.
- What do you use to file entries today? A free text field. It shows how hard the setup will be before anyone books a call.
- Who on your team would use it every week? Name and role. If nobody can be named, the brokerage is not ready, and both sides learn that now.
After submit, the page does not say You are on the list. It says: We read every application. One of the two founders will reply within three working days, yes or no. Then it shows the dates again. That one line sets the tone of the whole relationship, and it costs nothing to keep.
Fewer people will apply than joined the old list. That is the result you want. If the page brings in fifteen applications and nine of them are brokerages in range, the founders have a week of useful calls instead of a month of the wrong ones.
What to show when the product is not ready yet
A pre-launch page has no customer logos and no usage numbers, so the proof has to come from the product itself and from the people behind it. For the customs tool, that means one worked example. On the left, a real invoice line with the client removed: Mens cotton knit pullover, 60 percent cotton, 40 percent polyester. On the right, the tool's suggestion, the ruling it relied on, and a grey button labelled Broker approve. One sentence under it says the tool never files anything without a licensed broker's approval.
Under that sits a short section on the founders: who they are, what they built before, and which of them has worked inside a brokerage. A broker trusting a two person team with client data wants to know there is a person to call. This is also where the look of the page does real work. A stranger decides in about four seconds whether a company is serious, and a careful, specific page reads as a team that will still be here in a year. When we designed the Mi-VAD website, for a coin-sized heart pump, every page had to hold up for investors and clinicians who form a view before they reach the science. A broker reading a design partner page is that same kind of expert reader.
If launch is close and the rest of the site needs the same care, what launch-ready means for an AI startup lists the pages that carry the first weeks after you go live.
Rewrite yours this week
You can do most of this in a few days, without hiring anyone.
- Write the headline as a recruiting line: how many partners, of what kind, starting when. Five customs brokerages, starting February, beats any slogan.
- Name the job in the first sentence under it, in the buyer's words. Read it to one person in that job and ask what the product does.
- Write the exchange in three parts: what partners get, what you ask of them, and the seats and dates. Put the ask in hours and documents, not adjectives.
- Add a short list of who the program is not for. Two or three lines are enough.
- Replace the email field with three or four questions that sort applicants by fit. Make the first one about volume or size.
- Change the confirmation message to a promise: who replies, by when, and that the answer may be no.
- Show one worked example of the product on a real input, and one short section on the founders.
- Update the seat count by hand each time one fills. A page that says two of five taken in March and still says it in May tells a stranger more than you meant.
Studio Maydit designs websites for AI founders in the US, UK and Europe, in Framer, Webflow or custom code, and the work often continues into the product. A pre-launch page like this one is a fixed scope of three to four weeks: we write the exchange with you, design the page so an expert reader takes it seriously, and build it so you can change the seat count yourself. If you are still comparing studios, here is a ranking of design agencies for pre-launch AI products. If you are recruiting partners now and the page still says Get early access, book a 30 minute call.
