How to Choose a Design Partner for an AI Startup

Choosing a design partner for an AI startup is one of the first big outside bets you make after a raise. Get it right and your product, website and sales story start pulling in the same direction. Get it wrong and you spend three months and a chunk of runway on screens your engineers will not ship and buyers will not trust.

Quick note on the term. In startup language, a “design partner” often means an early customer who co-develops the product with you. That is not what this guide is about. Here I mean the designer or small studio you bring in to design your product and how you sell it.

If you are an AI founder selling to enterprise, with three to six people and no designer, this is the decision guide I wish more founders used before they book calls with studios, including mine. I was founding designer at Almetra AI from the first prototype through a $19M Series A, and I now run Grauberg, a design studio for AI startups selling to enterprise. I have sat on both sides of this choice.

The short answer

A good design partner for an enterprise AI startup has three things:

1. Enterprise judgment. They know what procurement, IT and the daily user each need to see before they say yes.
2. AI output design skill. They can turn model output into something a busy person trusts and acts on.
3. A way of working that fits a team that changes its mind weekly. Fast loops, real files engineers can ship, and someone senior actually doing the work.

Portfolio polish matters less than all three. Plenty of beautiful work fails in a pilot.

Step 1: Decide what you are buying before you talk to anyone

Most founders start by browsing portfolios. Start with a sentence instead:

“In 90 days, design should have helped us ___.”

Good answers are concrete and tied to revenue or learning:

• “Run a sales demo that does not need the CTO in the room.”
• “Get our first enterprise pilot users to act on insights without training.”
• “Make the website credible enough that a VP forwards it internally.”

Vague answers like “look more professional” or “a full rebrand” are a warning sign for you, not the partner. Without a clear outcome, every studio will look equally good on a call and you will choose on vibe.

Then decide the model. If the brief has a clear end date, you need a project. If it changes every week, you need embedded senior judgment. I wrote up that choice in detail in founding designer vs agency, and the timing question in when to hire your first designer at a startup. This guide assumes you have made that call and now need to pick a specific partner.

Step 2: The criteria that actually matter

Here is the scorecard I would use. Rate each candidate from 1 to 3 on every row after your first call and a look at their work.

Criterion	What weak looks like	What strong looks like
Enterprise judgment	Talks about delight and visual trends	Asks about buyers, roles, permissions, rollout and security review
AI output design	Charts, scores and a chat box	Shows claim, evidence and next action, with clear human control
Shipped work	Concepts and case studies with no live product	Real products that shipped, with what changed and why
Who does the work	Senior sells, junior designs	The person on the call is the person in the file
Engineering fit	Hands off static screens	Works in your component setup and reviews what engineers ship
Speed of learning	Long discovery phase before any screen	Puts something testable in front of users in the first weeks
Range across surfaces	Product only, or marketing only	Can connect product, website, demo and decks into one story
Exit plan	Wants you dependent	Leaves a system your first design hire can inherit

You do not need a perfect score. You need strong marks on the rows that match your 90-day outcome.

Step 3: Look for enterprise judgment, not taste

Taste is easy to show in a portfolio. Enterprise judgment is not, so you have to probe for it.

Ask the partner to walk you through one product they designed for a business buyer. Listen for whether they talk about:

• The different people in the deal. The champion, the daily user, the admin and the person who signs.
• Messy states. Empty data, partial data, errors, permissions, the second shift.
• The path from pilot to rollout. What had to be true in the UI for a customer to expand.

If the whole story is about colours, motion and a big reveal, they are probably great at launches and less great at the slow grind of enterprise adoption. I wrote down what that bar looks like in enterprise-ready UI for AI startups and, for the pilot stage, in design for enterprise pilots. Use those as a checklist on the call.

Step 4: Test how they design AI, specifically

AI products have a design problem normal SaaS does not: the system is sometimes wrong, and users need to know what it did, why, and how to correct it. A partner who has not wrestled with that will default to dashboards full of numbers.

At Almetra, the first version of the product was a dashboard full of KPIs and filters. Research with shop-floor managers showed they did not have time to analyse it. We moved to plain-language insights per station with a clear next action, and resolving an insight became a single tap that also told the model what was worth flagging. That shift, from showing data to supporting a decision, is the core skill you are hiring for. More on it in designing AI software for manufacturing.

A simple test: give the candidate one real screen from your product and ask, “What would you change first, and why?” A strong partner will ask who uses it, what decision it supports and what happens when the model is unsure, before they say anything about layout. If you want a vocabulary for that conversation, see AI UX design patterns for enterprise SaaS.

Step 5: Ask these questions on the call

Bring these to every first call. The answers are more useful than the portfolio.

1. Who exactly will do the work, and how many hours a week? Names, not roles.
2. What do the first two weeks look like? You want specifics: interviews, a teardown of your current product, a first testable flow.
3. What do you need from us each week? A good partner needs access to customers, sales calls and an engineer. If they need nothing, they will design in a vacuum.
4. How do you hand work to engineers? Ask to see real handoff files or a component setup, not just final mockups.
5. Tell me about a project where the design was wrong. What did you do? You want someone who changes their mind when users push back.
6. How do you keep the product, website and sales materials consistent? This is where small teams lose trust without noticing.
7. What happens when we hire our first designer? The answer tells you whether they plan to make you independent or dependent.

Step 6: Red flags I would walk away from

• Bait and switch. A senior founder on the call, a junior team in the file. Ask directly and get it in writing.
• Long discovery before any screen. Weeks of workshops and a strategy deck before anyone touches your product. At seed, that is runway you do not have.
• No questions about your buyer. If they do not ask who signs the contract, they are designing for the wrong person.
• Only concepts, nothing shipped. Concepts are fine as a side note. They are not evidence that the partner can work with constraints.
• A full rebrand as the default answer. Most seed-stage AI teams need a clearer core workflow and a credible demo far more than a new logo.
• Files only a designer can edit. If your team cannot update the deck or the site without the partner, you are buying a dependency.

Step 7: Start with one scoped asset

The safest way to choose is to work together on something small and important before committing to anything long. Good first projects for enterprise AI teams:

• The core demo flow your sales team shows on every call. See how to design a sales demo for enterprise SaaS.
• The one dashboard or insight screen pilot users open daily. See SaaS dashboard design for enterprise buyers.
• The homepage and product story a champion forwards internally.
• The sales deck your champion uses to sell you up the chain. See enterprise sales deck for AI startups.

Pick the one closest to your 90-day outcome. Within a few weeks you will know how the partner thinks, how fast they move, how they take feedback and whether your engineers like working with them. That is far better evidence than any pitch.

What good looks like after 90 days

If you chose well, you should be able to point at concrete changes:

• Sales runs the demo without improvising around half-built features.
• Pilot users act on the product’s output without a training session.
• The website, decks and product tell the same story with the same screens.
• Engineers ship UI from shared components instead of inventing it every sprint.
• You understand your users better, because the partner put work in front of them early.

If after 90 days you mostly have beautiful Figma files, something went wrong, either in the partner or in the brief.

How I approach it at Grauberg

I am obviously not neutral here, so use the scorecard above on me too. The way I work comes from the Almetra years: one senior designer in your product and sales context, starting with the asset that moves your next deal, then installing the rest of the Design OS assets, from product UI to website and sales deck, as you head toward product-market fit. The goal is always a handoff to your first full-time designer, not a permanent dependency.

Bottom line

Choosing a design partner is less about finding the best designer and more about finding the right fit for your next 90 days. Write down the outcome, decide the model, score candidates on enterprise judgment, AI output design and who actually does the work, and test the fit on one scoped asset before you commit. Do that and the choice usually becomes obvious.