Blog
Generative AI is not a catch-all drawer: it is a tool of the trade
AI arrived all at once and many people see it as something that does everything. Sorting each use, accepting the learning curve and putting craft first is what turns it into real productivity.
"Can't AI just do this?" I hear it in almost every meeting, whether it comes from a developer, an architecture practice, a shop or a restaurant. It rarely refers to one specific task. It refers to everything at once.
I understand why. Generative AI has changed the way I work more than any other tool in recent years. That is exactly why I know what lies behind it: choosing each use carefully, learning how each model behaves and reviewing everything it produces.
Generative AI multiplies people who master their craft and know what they want it for. Without that order, it is a catch-all drawer you pull random results out of.
What has changed in the way I work
Today AI is part of most of my working day. It helps me optimise processes, improve results and produce more in less time. But the important change is not speed: it is where my attention goes.
Tasks that used to eat hours without calling for judgement, such as first drafts, variations or repetitive work, I now prepare with AI and then review. The time I free up goes back into what I actually want to develop: design, user experience, visual direction and the conversation with the client. I described this when writing about vibe coding and web mini-games: once you stop wrestling with code, that time goes into design, UX and UI.
None of this happened on day one. It came after trying a lot of tools, discarding most of them and learning where each one fails.
Overwhelming, but not yet mainstream
Generative AI arrived at the right moment, but in a rush. Almost overnight, entire sectors were told they could multiply their productivity or cut costs with hardly any effort.
The numbers tell a different story. According to Spain's National Statistics Institute (INE), in the first quarter of 2025, 21.1% of Spanish companies with 10 or more employees used artificial intelligence; among those with fewer than 10, the figure was 13.4%. In construction it stood at just 11.4%. Adoption is growing fast, but most companies still do not use it.
The European picture is similar. Eurostat put AI use at 20% of EU enterprises in 2025. And among those that considered using it but did not, 70.9% cited a lack of relevant expertise. The brake is not the tool: it is knowing what to do with it.
The catch-all drawer: when AI "does everything"
It happens with my clients all the time. Businesses with a lot of digital process tend to see it more clearly: their goal is to optimise workflows and automate tasks. But people who sell products, run a restaurant or, like many of the people I work with, are architects or developers, usually start from a different idea: that AI will do everything for them.
That "everything" is the problem. It lumps together uses that have nothing in common and that call for different tools, risks and levels of review. Sorting the drawer starts with separating them:
- Ideation: exploring variations, references and moods. Nothing that comes out is final.
- Communication: explaining a proposal, preparing a presentation, answering a client's questions.
- Editing: retouching an image, adapting formats or polishing a text built on verified facts.
- Animation: bringing a shot to life to present a proposal.
- Automation: summarising, classifying or extracting data in repetitive digital processes.
- Experiences: AI as part of what the audience lives through.
That last drawer is the one we work in most at trade fairs. In the TransformIA Photo AI for Red.es, which has run in 2024, 2025 and 2026, visitors take a photo and in under 20 seconds see themselves in futuristic settings. There, AI does not save hours: it is the product, and its value depends on an interface and a booth designed to be understood instantly.
One architect, four different uses
What good is it for an architect to run their images through AI to "improve" them if they do not know how to treat them properly? Details get lost, a window frame changes colour without anyone asking, or a window appears that is not in the drawings. In two clicks, they have damaged their own work.
Now imagine a practice that separates each use:
- For ideation, it explores materials, planting or atmospheres in the early stages. The AI proposes; the architect decides.
- For the client experience, it shows alternatives during the meeting itself and turns the conversation into a more powerful creative flow.
- For editing, it retouches details without touching the architecture. Models such as GPT Image 2, released by OpenAI on 21 April 2026, promise stronger editing and better detail preservation. Even so, OpenAI's own guide warns that the model uses the mask as guidance and may not follow its exact shape with complete precision. Knowing the model means knowing how to ask, what to isolate and how to check every edit against the original.
- For animation, it brings a short shot to life, with little camera movement, and presents the proposal in motion. With judgement: checking that no line bends and no material changes from one frame to the next.
We explain how we combine 3D and AI without betraying the design in 3D architectural visualisation and AI.
The learning curve nobody advertised
It was sold to us as something that appears out of nowhere and works without any learning. It does not. You need to know which model suits each task, write precise instructions, prepare what you feed it (references, masks, data) and review with a professional eye, because AI rarely tells you when it gets something wrong. And you need to keep up: models change every few months.
That curve is not a flaw. It is normal for any powerful tool. The strange thing was believing this one would not have it.
Joaquín's jigsaw
I always give my clients the same example. My dear father-in-law, Joaquín, is a carpenter who works wood professionally. A jigsaw lets him improve his work and raise his productivity. If I pick it up, I will most likely cut my hand.
The tool is the same. What changes is the craft of the person holding it: Joaquín knows what wood is in front of him, where to cut and when not to. The jigsaw does not give him that knowledge; it amplifies it.
AI works in exactly the same way. Professionalism still comes before everything else. A good professional with AI achieves far more than before. Someone without the craft, using the same AI, just reaches the mistake sooner.
Where to start sorting the drawer
If AI still feels like a catch-all drawer in your company or practice, this is how I suggest starting:
- Write down the real tasks in your week, not the ones you would like a machine to do.
- Group them by goal: ideation, communication, editing, animation, automation or experiences.
- Pick a single use to start with, with a result you can compare against what you already do.
- Decide what must not be touched: the architecture, the brand, your clients' data or any figure you have not checked.
- Set aside time to learn, test and make mistakes on something that is not an urgent job.
- Decide what you do yourself and what you hand over to a professional who already masters that tool.
If you want to work out what AI can really do for your practice, your development or your business, tell us what you need to solve. At Viseni we work with AI across 3D visualisation, app development and trade fair experiences, and we can help you separate what adds value from what is just noise.
