Tools produce results faster, but they have also raised expectations, multiplied the available options and shifted much of the effort towards reviewing, integrating and making decisions.

We have never had so many tools capable of saving us time. And yet, we have rarely spent so much time reviewing, comparing, correcting and deciding.

Artificial intelligence is removing some of the most repetitive tasks from our work. But that does not mean the work itself has disappeared. In many cases, it has simply moved to other parts of the process.

How we worked before automation

Not so long ago, we already had tools that helped us develop projects more efficiently. Even so, many tasks still depended entirely on manual execution.

In design, for example, we had almost complete control over the result, but some tasks could take forever if you wanted to do them properly.

Selecting a person in a photograph, removing a background, refining edges or combining several elements in a photomontage required a great deal of attention. Sometimes it was difficult to explain to a client why a seemingly small task could take several hours.

Something similar happened in programming.

We often started by writing code from scratch and repeating very similar structures. The logic could be correct and the approach clearly defined, but the programme might still fail because of a comma, a quotation mark or a missing closing bracket.

Debugging could take hours before discovering that the entire problem came down to a single misplaced character.

These tasks consumed a lot of time, but they also forced you to understand every part of the process. Code reviews, error correction and decisions about how the project should be built took place continuously throughout the development.

The tasks that have apparently disappeared

Many of those tasks have been reduced considerably and, in some cases, have almost disappeared.

Selecting a person in a photograph can now be done by pressing a button. In the same way, an artificial intelligence tool can prepare the initial structure of an HTML project in a matter of seconds, provided it receives the right instructions.

Not long ago, something like this would have been difficult to imagine.

For designers and developers, these tools represent an extraordinary opportunity. They allow us to spend less time on certain mechanical tasks and focus sooner on other areas of the project.

The problem is that decision-making, reviewing, correcting and integrating are all still present.

Code can be generated in seconds, but we still need to check whether it works, whether it is secure, whether it has been properly planned and whether we will be able to maintain it later.

Execution has accelerated. The rest of the work is still there.

The new invisible tasks

It can now feel as though artificial intelligence is capable of doing absolutely everything.

And I have to admit that, in some situations, this is exactly the impression these tools give.

You ask for an image and they generate it. You describe a feature and they produce the code. You provide a few instructions and, within seconds, you have an initial proposal that would previously have taken several hours.

However, based on what I continue to see every day, we are still a long way from being able to delegate an entire project.

Working with artificial intelligence means reviewing, filtering and checking every result. It also adds a task that, for me, is essential: taking responsibility for the work being delivered.

Artificial intelligence will not be responsible for generating insecure code. Nor will it be responsible if an integration exposes sensitive information or if a technical decision causes problems a few months later.

The work we deliver has to be backed by the team that signs it off.

We are the ones who need to understand what we are delivering, how it has been built and what consequences it may have.

Tools help us reach the result. The responsibility remains ours.

Producing more does not always mean making more progress

The fact that a tool can generate a significant part of a project in a few minutes does not necessarily mean that the project is better.

I have seen companies and business leaders remove processes involving analysis, judgement and review because they believed artificial intelligence would directly replace professional expertise.

But a new tool does not automatically remove the need for knowledge.

With artificial intelligence, we are going through a period of social and professional readjustment. Many roles will change, while others will stop making sense in the way we currently understand them. New specialisms will also emerge, and we may not yet know what to call them.

Ten years ago, hardly anyone used the term content creator naturally, and twenty years ago very few people knew what a community manager was. Perhaps new professions related to artificial intelligence are already emerging, even though we do not yet know what their names will be.

The problem begins when we confuse producing more with making more progress.

Creating five websites in the time it previously took us to develop one does not mean that we are now doing work that is five times better.

In many cases, what has changed is the level of expectation. We are expected to produce more, faster and with fewer resources. This can lead us to sacrifice important aspects such as design, differentiation, research or prior reflection.

As a result, we are seeing more and more websites with almost identical structures, styles and messages.

This often happens not because the technical possibilities are limited, but because production began before anyone decided what actually needed to be done.

Judgement as the new bottleneck

We can now execute an idea almost immediately, but first we need to know which idea we want to execute.

We need to determine what we want to achieve, who we are addressing, what we want to communicate and what the project’s priorities are.

When there is a clear direction, the tools become valuable allies.

When there is not, we end up delegating decisions to them that we should have made beforehand. That is when the endless revisions begin.

Artificial intelligence interprets something differently from what we expected, generates a proposal that does not quite fit, and we start correcting it with new instructions. Then we correct the correction and enter into an endless conversation with the tool.

We can end up spending more time correcting the wrong direction than we would have needed to define the project properly from the beginning.

Our ability to generate has accelerated. Our ability to decide still depends on us.

A real example: the ISMF project

This difference between executing and deciding became especially clear during the development of the new digital platform for the International Ski Mountaineering Federation, the ISMF.

The starting point included creating a more modern website, but the project went much further than that.

The federation managed information about competitions, calendars, rankings, athletes, news, institutional documentation and resources for national federations. The platform also needed to work for athletes, organisers, the media, federations and fans from different countries.

Before development could begin, several questions needed to be answered.

How should all that information be organised? What did each user need to find first? Which types of content needed to be connected? How could the federation’s internal team manage everything? How could we prepare the platform to continue growing over the following years?

The work began by reviewing the existing structure, navigation, content organisation and user needs.

We then redefined the information architecture and developed a customised WordPress solution, with specific modules and a structure prepared for future expansion.

It was also important for the ISMF’s internal team to be able to manage and update the content independently.

Technology accelerated certain parts of the execution, but it could not decide what the right architecture was for the federation or fully understand the priorities of its users.

In a project of this kind, writing code is only one part of the work. Before that, you need to organise the problem and decide how to build a platform that will manage a large amount of information for many years.

Experience helped us make better decisions and identify potential problems before they appeared.

Tools help, but judgement is still necessary

Tools are becoming increasingly intelligent. They generate text, code, images and visual proposals at a speed that would have been difficult to imagine only a few years ago.

But the work has not become simpler. It has changed.

We have reduced the time needed to execute certain tasks, but we now need to pay more attention to defining, selecting, reviewing, integrating and taking responsibility for the results.

A large part of our work takes place before we write code or design a screen. It involves understanding the problem, organising priorities and preventing an early decision from negatively affecting the rest of the project.

Execution, tools and technology come afterwards.

We can now reach a result much faster. The difficult part is still knowing which result is actually worth reaching.

Before discussing tools or starting development, it is worth understanding the problem, organising the priorities and defining what the project actually needs.