Designer + AI: BASIC Checklist

Freelancing is a business. That means you need to act like a professional to be treated like one. Don’t just send a “paypal link” for the money from a client. Don’t use Gmail.

This screams “I’m working from my Mum’s basement” and most clients will treat you accordingly.

Here’s what you should do:

• Buy your own domain name and use your email on it.
• Have a contract template and an NDA template that’s yours.
• Some clients will send theirs, but you should be able to provide yours at all times.
• Understand taxes. Understand the NET price. Be clear about pricing and taxes in your contracts.
• Get any professional accounting/invoicing software and create real, legal-binding invoices. Specify payment terms, bank accounts etc. on them. Without a legal invoice you won’t be treated seriously as clients need an invoice from you to account your work as a “cost” and pay less taxes.
• Have a clear set of rules of engagement in a document. How much money do you take upfront. What are the acceptable delays from your side. When do you handoff the source files.

AI Quality debt is similar to design/programming debt. It means not being able to determine whether an AI made part was good or not, yet using it anyway to save time.

After a while the debt builds up, and you lose track of which part was made in which AI tool, how and what was the reason behind it. If you’re smart and thorough it will create both a sense of professionalism in the eyes of your clients and it will make it easier for you to upgrade the project.

• Create an “AI usage” document template for yourself
• Make sure to outline all AI tools that you used in the project in the document.
• Write which part was created with which AI.
• Add when it was created and which version (eg. GPT 4o) Add the prompt that you used.
• Annotate the potential issues and quality gaps. You can use a system (I have one) or create your own system that’s consistent so when coming back to it you understand it.
• Create versioning if you ever update an element when an AI tool becomes better. Show the before and after. Note out which quality gaps were bridged and which still exist. 

When generating any kind of AI outputs you need to be critical. Especially with LLM’s, keep in mind that they are there to make you satisfied with an answer. That doesn’t always mean the answer is right, good or even true.

That’s why you need to work on your ability to spot mistakes, issues and shortcomings WHILE generating stuff with AI. It’s based on logic and critical thinking combined with a distrust towards ALL ai outputs. It has to earn your trust.

• With photos or illustration look for mistakes (extra fingers?)
• With text look for signs of “AI writing”. It’s usually pretending to be “smart”, using smart sounding but meaningless words (streamlined, seamlessly) and — em dashes.
• With interfaces (even moodboard ones) look for misalignments, wrong copy and logical mismatches.
• With graps and charts look for when a chart doesn’t make sense, or presents data where the numbers don’t match what’s on the graph.
• With animation look for odd transitions, missing frames, glitches.
• With icons look for style mismatch and imperfections - different stroke width or cap endings. 

Don’t ask an LLM to validate your idea. It will jump through flaming hoops to validate any BS concept because it wants you to be happy with the answer. Before AI gets access to any research databases, don’t ask it for market viability either. Always talk to real people to find that out.

• Ask an LLM to find flaws or weak spots in your idea.
• Give it a perspective relevant to who would be evaluating your idea in real life (i.e. you are an electric car user and here’s my idea for an EV charging app - what would annoy you the most?)
• You can ask about similar ideas from other brands but always fact-check the results. AI tend to hallucinate and make stuff up when doing online research.
• Ask the LLM for a set of questions you can ask people to verify that idea, but always give them an open question at the end about their biggest problem that they want solved.
• To speed things up create an idea evaluation system. Add where in the process you ask AI, when it’s time for human feedback, when do you parse it with AI and how
• Mark which AI recommendations matched what the users later said. I use a traffic light system for this. 

AI averages content from the internet. Remember this. It takes the best alongside the worst of what online creators have to offer. Which means you get an average result most of the times.

Unless it's guided.

• Create a list of designers for whom you’re sure they’re doing real, client work, not eye-candy for clicks.
• Add a list of industrial designers, architects and engineers and create a curated inspiration board of their best work.
• Good design is timeless, so update your inspiration board regularly and look at it often.
• Look at and always evaluate all interfaces (and/or websites) you use. Even if for a brief moment, see if it’s a good experience or a bad one and make a mental note of it.
• Know the main design styles and trends by name and characteristic. You can then prompt for a “swiss style” typography for inspiration.
• Make a list of words or prompt structures that are more likely to get you higher quality results. Consult that list before new work. 

AI removes a lot of critical thinking and logic from people. Which means people using AI a lot get less critical. They often don't understand what they generated and why.

Learn annotation techniques to not only train your analysing muscle, but also show to clients that you understand.

Use paper annotations and flows to stand out + add the human/personal touch. Alternatively use a tablet.

• Learn creating color-coded “expanding” flow diagrams.
• Learn how to make a mind map on paper.
• Practice hierarchy strips technique for handoff documents.
• Do typeframes before website design.
• Create a system to annotate and explain your design decisions. I use a system based on three colors (black, red, green) and numbers to signify importance and correlation.
• Add the extra annotations to most of your deliverables, as the perceived value of your work will strongly correlate with your ability to explain it.
• Learn how to talk about your design decisions in a clear, no-jargon way to go through the annotations during meetings. 

Just making stuff with AI is pretty lazy. You get an output that’s sometimes good enough, but with a little extra work it can be elevated to great.

Learn the techniques to merge what AI generated with some extra work from your side. Don’t just generate some random “pretty” image and BS copy and call it a “hero section design".

These is the exact type of designers that will be getting cheaper and cheaper.

• Create a workflow for using AI photos - upscale, background removal, color correction, noise.
• Make a workflow to shorten and clarify the copy you write. Specify how short you want the outcome to be, and the “grade” understanding difficulty level to make it simpler.
• Merge hand-drawn sketches with external inspiration for more unique outcomes.
• Play with animating specific parts of a visual with AI video tools - learn to identify how to only animate specific parts and describing the animation.
• Add a badge to your annotation documents near main elements (like key visuals) to show the work proportions.

99% + 1%

I know. I know. We’re all introverts and we don’t want to do calls and meetings. But with the disassociation that AI brings, meeting people (whether online or IRL) will become a big part of earning more as a designer.

I’ve been to hundreds of meetings, including ones with big CEO’s of Fortune500 companies and there are a couple of things to remember.

• Learn to listen before you talk. And when listening make sure to UNDERSTAND.
• Don’t be afraid to own your design decisions. Explain them without jargon or hard words, but with growth and strategy (or conversion) in mind.
• Don’t interrupt people, but be clear to show when you want to chime in.
• Disagree even with the CEO if you have a good argument for it - be sure to back it up.
• In smaller meetings, with a more relaxed atmosphere you can jump to non-work subjects for brief moments, but make sure that when talking about yourself, you pick some interesting parts the other party can relate to. That will make them more friendly towards you. Also remember what they talked about last time and mention it. 

Finding new context in existing data will be a big skill of the future. With AI combining data sources is now faster than ever.

Learn to look at data as potential. What more can I do with what I get?

• Ask AI about correlations between various data points based on your user research - provide context for the question, otherwise it will hallucinate.
• Create your own algorithms that calculate and present new information from multiple data sources
• Validate your custom data with users and do manual checks to see if the data is correct.
• Learn to think about data sets as building blocks that can make new building blocks. Don’t just accept the data you get - create new data out of it.
• Plan for user-contextual data points to merge with external/general ones. You can ask LLM’s how to do it, but make sure to start at a point beyond statistical error.
• Not all data needs to be presented to the user. Some can be parsed, most can be hidden. Learn to avoid user overload and only pick what’s actually needed. 

If you “just” sell a design, you’re doing it wrong. The output of a design (even if great) is a commodity. There are great “design templates” out there. But even if the quality is good, the value of a templatized approach is always low in the eyes of clients. Instead focus on selling strategy.

• Always create a “strategy guide” document and attach it to your project. Do it even if the client doesn’t ask for it.
• In the document outline the current status of the design (state A) and use annotations (6) throughout.
• Divide elements into how likely they need to be changed. For example a section showing logos of clients is unlikely to be greatly improved. But a section about the offer is likely.
• Starting with state A, plan for a state B (and/or C) of what else can be tweaked and how. This can be new copy, a different visual or exchanging/replacing whole parts of UI.
• Create a timeline for user testing of state A, when state B (and others) should be introduced and how many results are viable for a revision.
• Emphasize your strategic approach as a path for the client to make more money because of working with you. You sell them a constant set of small improvements instead of a “one time design". 

Read case studies. Create personal databases of information, inspiration and patterns. Ask users - even through social media surveys. Gather data points.

Live and breathe design, but also (more importantly) live and breathe strategy. You’re a strategist now, which is what designers were always supposed to be.

They just forgot about it in the craze of moving colorful boxes filled with gradients.

Now moving those boxes will become cheap. The deep thinking, problem solving and ability to assure the client to trust you are the skills that will allow you to earn more.

As AI progresses, some of these skills may be replaced to a larger extent. That’s ok. With a long-term, strategic approach and a high focus on providing value you will be able to adapt.

Taste and quality will become your differentiator, but for that you need to first UNDERSTAND what good design is and WHY it is good. Otherwise you’ll join all the “prompters” in the race to the bottom of a price chart.

Designers are there to solve problems. Lower design cost is a problem that you can solve with this checklist. Combine it with the much needed “designer curiosity”, learn new things and go crush 2025 and beyond!

Where was AI used in the making of this article and how?

99% + 1%

This article was 99% human made. AI was used to shorten less than 1% of the paragraphs to make them more readable.