What Happens When You Use AI to Run an Entire Design Process. We Found Out
A conversation we are hearing in every boardroom right now, and it usually sounds something like: “Can’t we just use AI for this?”
It’s a fair question. And we think it deserves an honest answer – so we decided to do the work and find out for ourselves.
We spent over 500+ hours running structured R&D across real design and build project – using the full ecosystem of AI tools available today. LLMs like ChatGPT, Claude and Perplexity for research and strategy. AI design tools like Figma AI and Relume for ideation and wireframing. AI build tools like Cursor, v0 and Bolt for front-end development. The whole stack, applied to actual project work. We actually used a lot more tools, but the mention here is a good example of a few.
So let’s dig deeper and discuss where AI genuinely delivers, and where it quietly misleads you.
First: what problem are we actually trying to solve here?
The promise of AI in design is seductive. The idea of faster delivery and lower cost is a good motivation to invest in certain tools. To be clear – some of what is promised is real in reality.
The teams getting genuine value from AI in the design process are not the ones who bought a tool subscription and pointed it at a blank brief. They’re the ones who had already done the hard strategic work, understanding their customers, defining their problems, aligning their teams and then used AI to move faster from there.
If you’re considering AI as a way to skip discovery, stakeholder work, or customer research – it won’t work. It will just produce confident-sounding outputs that miss the actual problem. Faster.
The thing nobody warns you about: AI is very, very agreeable
This is the finding we talk about most internally and the one we think decision-makers most need to hear. Every AI tool we used has the same underlying characteristic: it is optimistic, persuasive, and naturally inclined to tell you things are going well. It will validate your brief, affirm your direction, and generate outputs that feel compelling and complete.
Every solution we suggest is met with “That sounds great…” We have to accept this is just how these systems are built.
This creates real risk. It’s genuinely easy to move fast through an AI-assisted process, feel the momentum, see strong-looking outputs, and mistake all of that for validated progress. We fell into this ourselves. Multiple times.
This is why human judgment at every decision point isn’t optional. It’s the thing that makes the whole process credible.
On time and cost savings – read the fine print
If the business case you’re building for AI is primarily about saving time and money, the savings are real in some places, and they’re genuinely offset by new costs in others.
Yes, AI speeds up early-stage work significantly. Very significantly in fact. But at some point in every project, you still need to test your design with real customers. That step does not go away. If you have not done the groundwork upfront, it will cost you later.
If you’re working across internal teams, external partners, and client stakeholders, the alignment and communication work is still time-consuming. AI doesn’t shortcut it. You need to be transparent and communicate with the teams, allow for error and agree on the experimental path you are adopting.
Another biggie to note, once something is built with AI assistance, updates, ownership, and iteration become harder to manage than most teams anticipate. It’s an excellent toolset for set-up, validation, and prototyping. It is much less straightforward for the long haul and scale.
Timelines don’t bend. Be honest about whether now is the right moment to experiment
One of the more uncomfortable findings from our R&D: if you make a significant misstep with an AI-led approach mid-project, you may need to backtrack and restart. And your deadline will not move to accommodate that.
There is a real and important question that every project needs to answer upfront: is this the right moment to experiment with AI, or do we need to deliver? Both answers are valid. What’s not valid is assuming that using AI removes the risk of getting things wrong, or that it provides a safety net if you do.
People will push back. Be ready for it.
As AI-generated research and outputs become more common, the scrutiny around them is increasing – and rightly so. CEOs and CMOs who are signing off on strategy informed by AI research are asking harder questions. Boards want to know how insights were validated. Clients want assurance that what they’re looking at reflects reality, not a well-structured hallucination.
The quality of the brief, the methodology behind the planning, and the human judgment applied at every decision point: these are what make AI-assisted work credible and defensible. Without them, you’re producing fast work that informed people will (correctly) question.
So – should you invest in AI for design? The honest answer
Yes. But not as a replacement for strategic thinking, discovery work, or the expertise that comes from doing this for a long time across many different clients and industries.
The businesses and teams getting real value from AI right now are the ones using it to move faster within a process that is already well-structured. They’re not skipping steps. They’re compressing them.
And the agencies, including us, who are integrating AI into their workflows aren’t just bolting tools onto the way they used to work. They’re rethinking the model. What skills matter now? Where is human expertise irreplaceable? How to build processes that use AI’s speed without inheriting its blind spots.
AI rewards businesses that have already done the hard work of knowing what they’re doing and why. For everyone else, it mostly just makes the existing problems arrive faster.
500+ hours in, that’s the most honest thing we can tell you.
All the best,
LDN AI Lab
AI LAB is the experimental AI division of Studio LDN, where we’re actively building and testing AI solutions that enhance customer research, UX, CRO, and creative content.
