If the question before was “what would you do if you had access to all the knowledge?”, the question now is “why aren’t you doing it yet?”.
That idea captures quite well several conversations I have had recently with clients and teams that are looking at artificial intelligence with interest, but also with a certain level of paralysis. There is enthusiasm, there is curiosity, and in many cases there is already access to tools. However, when the conversation moves from inspiration to execution, a barrier appears that is less technological and much more organizational: someone has to decide what to change, where to start, and who will be accountable for the outcome.
For a long time, technology was a real limitation. The tools were not always available, the data was not organized, specialists were scarce, budgets were high, and many ideas remained on hold simply because there was no internal capacity to bring them into practice. It was reasonable to say that progress was not possible because knowledge, people, or infrastructure were missing.
But that reality is changing. Today, many things that used to seem complex, expensive, or far away can be prototyped in days. A marketing campaign, an automation, a first version of an agent, a support assistant, a commercial proposal, a document analysis process, customer segmentation, or an initial model to prioritize opportunities no longer necessarily require starting from scratch or having the entire expert team hired from day one.
This does not mean that everything is easy, nor that AI replaces expert judgment, experience, or human responsibility. In fact, the closer AI gets to real business processes, the more important human judgment becomes. But I do believe that many of the barriers we used to justify inaction are starting to lose weight. Technology is no longer always the stopper. Knowledge is not necessarily the stopper either.
Then a more uncomfortable question appears: what do we really want to do?
Many companies already have access to ChatGPT, Copilot, Gemini, Claude, or other tools. Some are even paying for corporate licenses and have already delivered internal training sessions. But when you look at actual usage, AI often still operates as an individual and isolated tool. Someone opens a browser tab, asks for help writing an email, summarizes a document, prepares meeting notes, or generates ideas for a presentation. All of that can be useful, but it does not necessarily transform the way a company works.
Real change appears when the organization decides which process it wants to improve, which capability it wants to build, which problem it wants to solve, and what level of risk it is willing to assume. That is where, in my opinion, the new bottleneck is. It is not only in AI, nor in the tool, nor in the model. It is in human decision-making.
Because expanding capacity is not the same as creating value. Having access to artificial intelligence can accelerate tasks, but for that to become real impact, someone must define priorities, responsibilities, metrics, and boundaries. A company needs to decide which tasks are worth automating, which processes should be redesigned, which data can be used, which use cases make sense, and what will be done with the time or capacity that is freed up. AI does not solve that part on its own.
One of the most common mistakes I see is treating artificial intelligence as if it were a layer that can be installed on top of the current company without touching anything else. A tool is enabled, a training session is delivered, a license is purchased, and productivity is expected to appear magically. But productivity does not appear simply because someone has access to a technology. It appears when the way of working changes.
That is why I believe many organizations, at least at this stage, do not have a strictly AI-related problem. They have a problem of decision-making, focus, and ownership. There are pilots that do not scale because no one clearly defined what success meant. There are use cases that do not move forward because no one took responsibility for the full process. There is data that never gets prepared because everyone is waiting for another area to organize it. There are automations that are not implemented because they require changing a way of working that, although inefficient, is already familiar. And there are risks that are not managed because it is easier to keep the conversation at a general level than to make concrete decisions.
Meanwhile, many conversations continue to revolve around questions that are important but incomplete: which tool to use, which model is better, which license to buy, or what other companies are doing. These are valid questions, but they are not enough. Before choosing the tool, the company should ask itself what business decision it is avoiding.
If an organization knows what it wants to achieve, AI can help accelerate a lot. It can help analyze, synthesize, automate, generate alternatives, prepare documents, respond to customers, organize information, or discover patterns. But if the organization lacks clarity on the objective, AI can accelerate confusion. It can produce more documents, more ideas, more drafts, and more noise, without necessarily bringing the company closer to a concrete result.
It is also important to avoid the opposite extreme. Deciding does not mean automating everything without control or pursuing every use case just because it is technically possible. Human decision-making also means defining boundaries, acceptable risks, security criteria, supervision levels, and governance mechanisms. Adopting AI is not just about accessing a new capability; it is about deciding how that capability will be integrated into the business responsibly.
Leadership matters a lot in this process. AI can suggest paths, but it cannot define a company’s ambition. It can propose alternatives, but it cannot take responsibility for the outcome. It can accelerate analysis, documents, code, campaigns, or processes, but it cannot replace strategic clarity or the ability to prioritize.
That is why, when a company says it wants to implement AI, I am increasingly less interested in starting with the tool and more interested in understanding the decision behind it. What does it want to improve? What does it want to stop doing manually? What does it want to respond to faster? What does it want to sell better? What does it want to measure more accurately? What does it want to learn about the business that it is not seeing today?
AI can already do many things, and it will be able to do many more. But that does not eliminate the need to decide; on the contrary, it makes decision-making even more important. If knowledge starts to become available, if technology lowers the barrier to entry, and if new digital capabilities begin to be within reach of any team, then the competitive advantage will not only be in having AI. The advantage will be in knowing what to do with it.
Maybe that is the conversation many companies need to have now. Not a conversation focused only on tools, models, or trends, but a more direct conversation about priorities, decisions, and responsibility.
Because if AI can already help you do it, the underlying question is much simpler and much harder: what are you waiting for to decide?
