AI in My Design Workflow: Where It Helps and Where It Doesn't

AI tools have changed how I work faster than any software change I can remember. Some days they save me hours. Other days they produce confident nonsense that takes longer to fix than doing the job myself.

After a lot of trial and error across brand design, social media work, app interfaces and code, I’ve settled into a clear split: AI is excellent at some parts of the design process and poor at others. Here’s where I use it, where I don’t, and the rules I follow to use it responsibly.

Where AI genuinely helps

1. Turning messy input into a clear brief

Clients rarely send a brief. They send a voice note, three WhatsApp messages and a screenshot of a competitor. A language model is very good at turning that into a structured summary: goals, audience, deliverables, deadlines and open questions.

I never send that summary as-is. I use it as a first draft, correct it, and then confirm it with the client. It saves me twenty minutes of organising and makes sure I don’t miss a requirement buried in a message.

2. Research and first-pass understanding

When I start working with a new industry, AI tools help me get oriented quickly: what terms people use, what competitors typically emphasise, what questions I should ask. It’s a starting point for real research, not a replacement for it. Anything factual gets checked against primary sources before it influences a design.

3. Exploring directions quickly

Image generation tools are useful early in a project for exploring mood and direction: lighting, colour atmosphere, composition ideas for a campaign. I treat these images as sketches. They help a client react (“warmer”, “less busy”, “more local”) before I spend time building the real thing.

What I don’t do is present generated images as finished brand assets. More on that below.

4. Image editing chores

AI-assisted features in tools like Photoshop’s generative fill are genuinely useful for production chores: extending a background to fit a wider banner, removing a distracting object, cleaning up a product shot. These are tasks that used to take careful manual work and now take minutes, though I still check every edge closely.

5. Copy variations and translation drafts

For social media campaigns I often need many variations of a headline or caption. AI can produce options quickly, which I then edit for tone and accuracy. For Bangla, I’m especially careful: AI translations can be grammatically fine but sound unnatural, so a native reader (usually me) rewrites them. I covered the typography side of bilingual work in Designing Bilingual Brands.

6. Code and automation

This is where AI has changed my work the most. I use AI assistants to write and debug the scripts that generate my GitHub profile cards, to help build parts of my apps, and to automate repetitive tasks like resizing and exporting assets. For a designer who codes, an assistant that explains an error or drafts a function in seconds is a huge accelerator. I still read every line before it runs, and I test it.

Where AI doesn’t help (yet)

Strategy and judgement

AI can list options. It can’t decide which option is right for this client, with this budget, in this market. That requires understanding people, context and trade-offs, which is most of what a designer is actually paid for.

Logos and identity marks

I don’t use AI to generate logos. Generated marks tend to look generic, are hard to edit as clean vectors, can resemble existing marks without anyone noticing, and raise real questions about ownership. A logo has to be distinctive, ownable and built properly for every size. That still needs a designer and a construction grid, as I described in Brand Identity Is More Than a Logo.

Typography and fine detail

AI tools are still weak at typography: kerning, consistent letterforms, and especially Bangla script with its conjuncts and vowel signs. Any text in a generated image needs to be replaced with real, properly set type.

Knowing what’s true

Language models can state incorrect facts with complete confidence. In a design context, that means wrong statistics on an infographic, a made-up quote on a social post, or an inaccurate product claim. For anything factual, I verify.

My rules for using AI responsibly

1. Client data stays private. I don’t paste confidential briefs, unreleased product details or personal data into tools without permission and without checking how the tool handles data.
2. Generated images are drafts, not deliverables, unless the client has agreed and the tool’s licence clearly allows commercial use.
3. No generated logos or identity marks.
4. Every fact gets checked against a reliable source before it’s published.
5. Be honest about AI use when it’s a meaningful part of the work, especially for imagery.
6. Real people’s likenesses are off-limits for generated images unless they’ve agreed.
7. The final decision is human. AI can propose; I decide, and I’m responsible for what goes out.

How my workflow looks now

A typical brand or campaign project now looks something like this:

1. Brief: AI summarises the client’s messages; I correct and confirm.
2. Research: AI helps me get oriented; I verify and dig deeper.
3. Exploration: quick generated mood images and rough sketches for direction.
4. Design: I build the real work by hand: identity, typography, layout and hierarchy.
5. Production: AI-assisted editing for background extensions, clean-ups and resizing.
6. Copy: AI drafts variations; I edit for tone, accuracy and Bangla.
7. Review: I check every detail as if no AI was involved.

The time saved goes into the parts that matter most: thinking, crafting and talking to the people the work is for.

The skill that matters more now

If AI makes producing something easy, then the valuable skill is knowing whether that something is good. Taste, judgement, an understanding of design principles and the willingness to throw away a plausible but wrong idea are what separate a designer from a generator.

AI is a powerful assistant. It’s not the designer. In my work, it makes me faster at the parts I don’t enjoy, so I can spend more time on the parts that need a person.