We are constantly told that we need to learn how to prompt and that knowing how to write instructions for artificial intelligence is the new essential skill of the 21st century. Training courses are popping up, people are displaying titles that are as new as they are original, and some even see it as a profession with a bright future.
But we need to face reality: no one wants to prompt. Not employees, not managers, not customers. And if we want AI to become a permanent feature of our professional lives, we must accept that this stage is only a transitional phase and by no means an end in itself.
In short:
- Prompt engineering is an overrated and transient skill that is unattractive to most professionals, who are primarily looking for simplicity and time savings when using AI.
- The history of technology shows that its widespread adoption depends on simple, integrated interfaces, not on uses reserved for technicians or enthusiasts.
- The hidden cost of prompting, in terms of time and efficiency, can negate the expected benefits of AI, especially in a professional setting where productivity is paramount.
- The main challenge is not to train all employees in prompting, but to embed AI into existing business tools and processes in a fluid and contextual way.
- The adoption of AI will come through its invisibility in daily workflows, directly serving the task at hand, rather than through explicit mastery of the tool.
The illusion of prompt engineering
For several months now, we have been told that to take advantage of artificial intelligence, we need to know how to prompt, and this has even become a discipline in its own right, known as prompt engineering, as if tomorrow everyone were going to wake up with the task of writing instructions for a machine.
Let’s start by looking at the bright side: if people, in order to talk to machines, start making the effort to be precise and clear in a way they never wanted to be with humans, and maintain this (good) attitude when talking to their fellow humans in the future, that will still be a good thing (No one should be promoted to manager if they don’t know how to use ChatGPT).
Wishful thinking aside, let’s be clear: curious people and other early adopters, myself included and perhaps you too, are naturally drawn to it. It amuses geeks and feeds a few micro-businesses that make a living from training people to write prompts, but for the majority of employees, it is neither a desire nor a skill they wish to develop. What people want is not a new cognitive burden but simplification. AI will only become truly useful, usable, and used when it is embedded into everyday tools and practices in a virtually transparent way.
This has been one of the basics of introducing any technology into the world of work since time immemorial, and things are not going to change anytime soon. However, little or no mention is made of this because those who most often have authority on a subject are not the average users who will then have to use the technology. This is not a problem (I am the first to admit that I am often one of them), provided that we have the clarity of mind to admit it and do not adopt the common stance of saying that “it is the others who are not up to speed and do not want to make an effort”.
From command lines to interfaces
The history of computing has already taught us this lesson. Command lines were a tremendous productivity tool for some, but consumer computing only exploded with the arrival of graphical interfaces. As long as operating machines was reserved for those who knew how to type commands, their use was limited to a technical elite.
The same can be said for mobile phones and smartphones.
Today, AI is still at this stage. It questions, it amazes, it interests, but it remains constrained by the need to manipulate language as a series of instructions. As long as users have to learn to “speak machine” rather than the other way around, we will not reach the stage of mass adoption.
The cost of prompting
We often forget the hidden cost of this practice. In a business, if an employee spends ten minutes testing and retesting a prompt to get the right answer, the expected benefit may already have evaporated. You might say that we learn by trying, but recently a specialist showed me that, on a business scale, the difference between employees who prompt effectively and those who don’t can vary from 1 to 10! This may be a variable to take into account.
The impression of productivity masks a reality that is often more nuanced. Yes, the tool is impressive, but if the time saved is swallowed up by the complexity of the interaction, then the economic equation can be negative. And that’s not even mentioning the barrier to adoption for users who may become discouraged or even turn away from AI.
The job before the machine
The real challenge, the next step, is therefore not to train battalions of employees in prompting, but to design systems that integrate into their jobs. Jocelyne’s job in accounting is not to dialogue with AI, but to produce reliable financial statements. Robert’s job in marketing is not to string together prompts, but to understand his customers and align a relevant message.
It’s the old story of the gold rush. Seekers weren’t there to perfect the use of shovels, they were there to find gold, and often the only ones guaranteed to get rich were the shovel sellers. Today, those who thrive on learning the prompt are replaying the same scenario, but they are not the ones who will create lasting value. At a given moment, during a phase of initiation, discovery, and maturation of tools, yes, but not in the long term.
From technical interface to “stealth” companion
The true value of AI will not be captured by those who learn to manipulate the machine with incantations, but by those who know how to integrate it almost invisibly into workflows, where it is forgotten and increases performance without conscious effort, without the user even thinking about it. In short, where technology disappears behind business use and is no longer talked about (Technology is a word that describes something that doesn’t work yet (Douglas Adams)).
We must move from technology reserved for tinkerers to universal technology, adopted not because geeks excel at taking advantage of it, but because it is useful to everyone, with no skill barrier to entry. That will be the moment when we stop talking about AI and start talking only about work, value produced, and tangible benefits.
Embedded AI vs. standalone AI
In the workplace, the difference between AI embedded into work tools and AI offered as a standalone solution is essential. Generalist assistants such as ChatGPT or Claude may appeal to certain curious profiles, but for the majority of employees, they remain outside the workflow. You have to open a separate environment, come up with prompts, transfer content, and then reintegrate it elsewhere. This is an experimental approach, which is of course very useful for testing ideas and for certain profiles, but which struggles to find its place in the daily routine of jobs organized around well-established processes and tools.
Conversely, players who bring AI closer to business uses are changing the game. Microsoft with Copilot in the 365 suite (sometimes), Salesforce with Einstein, and ServiceNow in operations management do not require users to learn how to interact with a machine. AI is embedded into HRIS CRM, or support tools, and it intervenes contextually on specific tasks: preparing a presentation, qualifying a lead, writing a meeting note, or documenting a ticket. Users don’t have to worry about prompts; they can stay focused on their work and benefit from specific help at key moments. They don’t control the tool; the tool takes them by the hand when necessary.
This integration is the real condition for adoption. As with IT in the past, it’s not the power of the raw technology that has made the difference, but its ability to fit into everyday use. It is therefore understandable why players such as OpenAI, which do not have this business proximity, must rely on partners to penetrate businesses on a large scale. Without Microsoft’s support, their technology would remain largely confined to individual and exploratory uses.
Bottom Line
If artificial intelligence is to become a permanent fixture in business and society, it will have to follow the same path as IT did in the past: moving from an interface that needs to be mastered to a virtually invisible companion.
The day we no longer talk about prompts but only about work done better, we will know that AI has left the age of tinkering behind and entered the age of maturity. At that point, we will have stopped asking users to be engineers and they will once again become what they should be: professionals focused on their work, not on the machine.
To answer your questions…
Prompting requires cognitive effort that employees and managers are unwilling to expend. They expect simple, integrated tools. As with other technologies, the manual learning stage is temporary. Eventually, prompts will disappear in favor of invisible, practical interfaces.
Testing multiple prompts takes time and reduces productivity gains. On a large scale, the gap between effective employees and others can represent a significant cost. This slows adoption and argues for integrated AI rather than AI that depends on the quality of prompts.
Computing became widely accessible with graphical user interfaces, not with command lines. Similarly, AI will only become widespread when it is simple, integrated, and accessible, without requiring users to “speak machine language”.
AI such as ChatGPT remains outside the workflow. Conversely, AI integrated into Outlook, Word, or a CRM acts directly within business tasks. It provides assistance at the right moment, without additional training, which encourages adoption.
When we no longer talk about prompts but only about tangible results: better-written documents, faster tasks, clearer analyses. AI will become an integral part of business processes, rather than a technical skill to be mastered.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
Why the widespread adoption of AI by consumers says nothing about its future in the workplace
Why enterprise AI can’t keep up with consumer AI: beyond ChatGPT, a more complex reality
Will AI replace juniors? The false debate that’s only the tip of the iceberg
Generative AI in the workplace: revolution or illusion?
AI heading for an economic dead end?
Work and AI: a transformation that is more organizational than technological
