Individual, department or company: the right way in depends on your company's moment, and the most common mistake is walking through one door without coordinating the other two.
AI adoption has become widespread. It is hard to find a company that has not bought licenses, run a pilot or announced a generative AI initiative. The results, however, still arrive unevenly: in a recent survey by Wharton with the Harvard Business Review, 45% of executives report a significantly positive return from AI, while among middle managers, where the work actually happens, the number drops to 27%.
Our reading of these numbers is simpler than it seems, and even more familiar. Adoption already generates real returns at both levels; the gap between them is not disagreement, it is the size of the result left on the table when the levels are not coordinated. And that gap does not describe a limit of the technology, but a market looking only at the technology, when adoption happens in the processes, the culture and the infrastructure of each organization. Adjusting that view, balancing context, method and technology, is what gets past the adoption barrier.
Adopting AI is a project that demands agility, but it is also a long-term one. It asks you to think not only about the next step, but about what comes after it. Before committing to high-budget solutions or licenses, you need to test and get to know the technology up close, understanding how your company and your team can generate results with it.
The three levels
Adopting AI is not just incorporating a new technology. It is working with a technology that makes it possible to rethink countless business problems, new and old. That is why adoption does not happen in a single place: it can happen in the individual, in the department or in the company.
At Taqtile, we start our projects by reading each company's scenario and interpreting its challenges through these three levels of adoption:
- Level 1, the individual. Training people to use well the AI tools they already have access to. Often the license has been bought and the tools have been built; what is missing is the method to structure agents, delegate what is repetitive and keep reviewing with judgment.
- Level 2, the department. Transforming with AI the processes of a team that take too much time, depend on rework or could yield more. One bottleneck at a time: the process gets mapped, partly automated and measured, and the team gains room for what only it can do.
- Level 3, the company. Redesigning processes end to end around what the technology made possible: journeys in which AI runs entire stages, products and services that did not exist, and new sources of revenue for the company.
These levels are not only about adoption; they are also about delivering results. And this is where a good part of the challenges lie: the initiatives that fail do not stumble on the technology, they stumble on design, when the expected result and the chosen adoption process do not match.
The right way in depends on your company's moment and challenges. That is why it is essential to define beforehand where the result will be measured: in the trained employees, in the efficiency of the chosen process, or in the new revenue that an AI solution opens up?
In practice: if the internal demand is for new ways to generate revenue with AI, it makes sense to start by prototyping, showing the business the potential of a new solution. In that case, the way in is level 3 and, if there are no trained people in-house yet, it is the moment to bring in a partner that already builds with the technology. Now, if what you want is autonomy in development, the path is to train the teams first, and only then ask for new revenue.
There are many ways to start. In our projects, we often start with prototyping, where AI training goes along with building a business solution instead of coming before it. That is how it went with the agent that now runs car subscription sales for one of the leaders in mobility in Brazil: the client's team corrected the agent while it was being built.
In other words, speeding up AI adoption does not mean training everyone first, then designing processes, and only then proposing new uses. One of the most valuable traits of this technology is that it lets you think and act fast: a focused prototype proves value in weeks and grows from the result. It is a logic we stand by in our projects: start smaller than feels comfortable, prove value, expand.
And if your company's challenge is a different one, the combination flips. A company with no urgency for revenue, but with an appetite for a culture of innovation, can come in by transforming a department process and educating the team along the way, so that the next implementations are born from within. The variety of combinations is wide, and it is what lets you locate your company on this map instead of copying someone else's path.
The silent mistake
The mistake we see most often is not choosing the wrong door. It is walking through one door without coordinating with the other two. In AI adoption, an isolated choice, however impressive, does not hold when the other parts are left out: licensing tools, building solutions or hiring training needs to come with a design that ties people, processes and results together.
The signs are familiar. The technically flawless solution, built without educating the people who will operate it, ends up on the shelf within months. The process redesigned without bringing the team along goes back to the old design at the first deadline pressure. The broad training program, launched while leadership still doubts the investment, spends the initiative's political capital without producing the proof it needed. In every case, the technology moved and the context was left behind.
The combination that sustains a transformation has two ends tied together: what the company needs right now, and the people who need to get there with it.
Context before technology
That is why the moment we pay the most attention to in any conversation is not the tool demo: it is the design of the proposal. Speed, size and depth vary from company to company, and a proposal that ignores this is generic from the start. Some contexts call for a prototype in weeks to unblock a decision, and others call for a longer program, where educating the team matters more than the first delivery. Reading this before proposing is what turns the three levels into a path, rather than a menu.
The MIT study on GenAI in companies brings a figure that helps put this in perspective: projects run with external partners are twice as likely to succeed and reach production three times faster. We are biased here, we know. But the reason behind the number interests us more than the number itself: the gain does not come from a single move. Partners who adapt to the right front, training a team today, prototyping an agent tomorrow, integrating it into the workflow after that, are the ones who show up in that 2x success and 3x speed.
Where does your company come in?
If your company is in this conversation, the question that matters is not "which tool should we adopt?", nor "where is everyone else starting?". It is another one: what do I need to prove right now about the technology (potential, applicability, revenue, efficiency, capacity to innovate), and how can I prototype that proof as fast as possible?
The answer locates your way in.
If you want to test this question in your context, start a conversation with us, and we will find out together where your challenge lies.