AI scales whatever you give it
One of the most obvious benefits of AI is that it can produce an enormous amount of work very quickly. There’s no disputing that it can be an effective tool at helping your organization do more. Send more emails. Post more social content. Write more articles. Plan more programs. But AI doesn’t just scale the output. It scales the quality too, and that’s not always a good thing.
Give it a good strategy with plenty of detail and it will help you find ways to execute it. Give it a bad strategy with no direction? It will help you execute that too.
AI doesn’t care what it's scaling. It doesn’t care if you’re wrong, and it wouldn’t have the guts to tell you if it did. Its purpose is to give you what you ask for. That makes it both highly useful and incredibly dangerous.
More is not always better
A 100-word draft becomes a 1,000-word article. A rough budget becomes a detailed financial plan. A few notes become a complete strategy. Who wouldn’t be amazed by this? Suddenly, instead of spending 15 hours writing a report, you can write a single page of highlights and generate 10 pages of polished content.
I once saw a company commit $50,000 to support another organization's new initiative. With that much money on the table, they asked for a detailed implementation plan. The organization developed an outline, then used AI to expand it into several pages of polished content. But polished content is not always meaningful content.
The investing company described the situation plainly: there were a lot of words, but it didn’t say anything. The document was revised, again with the help of AI, but the underlying problems still existed. The partner ultimately reduced its commitment from $50,000 to $5,000.
This was not a failure of the AI platform. The tool delivered exactly what it had been asked to produce: more content. AI can add volume without adding substance. If you give it an empty container and ask it to build something, it will produce a larger empty container.
In many organizations, doing more is confused with progress. This can make AI feel necessary, useful, and impactful. But doing more is not the answer to every problem.
Disconnected departments can use AI to become more efficient independently, but each of them may be moving in a completely different direction. The gaps between departments become even larger, and the organization itself stagnates. More things are being done, but the results pull in different directions rather than toward an overarching goal.
This may help explain why so few organizations are seeing meaningful value from AI. BCG’s 2025 research found that 60% of companies were generating minimal value from AI despite significant investments. Only 5% were achieving AI value at scale.
The problem is not necessarily AI. In many cases, organizations and individual users have no clear plan beyond: “I need this right now.” AI is an attractive tool because it can deliver an immediate result. Whether that result is beneficial is another question.
Best practices are not a strategy
Ask any AI platform how to improve marketing, increase donations, develop a new service, or make a team more efficient, and it can give you plenty of reasonable advice. Most of it will sound familiar because similar advice already exists in books, articles, videos, and countless other sources.
Most of the general-purpose AI platforms organizations use today are built on large language models trained to recognize patterns across enormous amounts of content. They are generating a likely response based on the input they’ve been given and the context you provide.
That means they can produce real recommendations, but they do not automatically understand how those recommendations fit your organization. That does not make the advice wrong, but it may not be deep enough. It may not even be practical.
AI does not automatically understand why customers choose you, what your employees are capable of, how internal limitations shape your work, or why something that succeeded elsewhere might fail in your environment.
That matters especially for smaller organizations. A small organization cannot compete with a global company by copying the global company’s strategy. The larger organization has more people, more money, more data, and more capacity to execute the same “best practices” at a much larger scale.
The smaller organization needs a different advantage. It may understand a particular audience better. It may have stronger relationships, deeper local knowledge, or a more focused service. When AI is used strategically, it can enhance those advantages. When used arbitrarily, it can flatten them into generic corporate thinking.
We built our "About us" copy generator to demonstrate how easily that can happen. Give AI a company name and almost no meaningful context, and it can still produce polished, plausible language about innovation, people, growth, and the future. The words sound professional. Whether they actually say anything about the organization is another question.
AI scales mistakes too
Being wrong occasionally is part of running an organization. No employee, leadership team, or consultant is infallible. Not even us, unfortunately. But AI can produce and repeat mistakes at a scale that was previously much harder to achieve.
An incorrect claim enters one document. That document becomes the basis for another. The information is summarized into a report, added to training materials, or used to guide a strategic decision. The original mistake becomes nearly impossible to find because it is now surrounded by pages of polished, confident language.
Research into AI accuracy varies depending on the platform and task, but factual errors remain common enough that organizations cannot treat verification as optional. Expertise is a non-negotiable part of working with AI. It is not a replacement for talent, experience, or skill, it is a tool that can enhance them.
The person using AI needs to be competent enough to identify when the AI is wrong, the amount of context that is needed, and much more. In The Expanse, pilot Alex Kamal uses his ship's computer to plot courses, run simulations, and understand situations that would be impossible for one person to calculate quickly.
He is not an unskilled layman relying on AI to do his job. He is a highly trained pilot using AI to extend something he understands deeply. He also does not assume that access to an AI system makes him a competent engineer or mechanic. The tool enhances his expertise, it does not give him someone else’s.
Expertise is not enough
Expertise helps you recognize when AI is factually or technically wrong, but it doesn’t guarantee that your original concept, idea, or plan is good.
Responsible AI use requires the ability to edit yourself. You have to be willing to step back and ask yourself difficult questions. Does this plan have real substance? Does the output truly address my need? Is this information practical and useful?
Self-editing requires separating the quality of an idea from your personal attachment to it. You have to want the work challenged more than you want it affirmed. That is difficult. Most people don’t enjoy hearing that their plan is unclear, their argument is weak, or several hours of work have produced something unusable. But criticism is almost always necessary for improvement. Very few people produce exceptional work without having their ideas questioned, revised, and occasionally rejected.
AI makes self-editing even more difficult because it does not create the same social discomfort as a colleague pushing back. It will continue developing your idea without stopping to say, “There isn’t anything here.” You can ask AI to challenge your premise, identify weaknesses, and argue against your recommendation, but you have to be willing to initiate and accept it.
AI can enhance your organization's pre-existing editing and refinement processes, but it cannot replace them or serve as a substitute for human expertise and perspective.
Scale deliberately
In the coming years, AI will almost certainly become more common, whether people are excited about it, afraid of it, or dismiss it. For organizations using it today, the more important question is not simply whether AI can do a task, but whether the organization is prepared to use the tool effectively.
That means setting clear goals, training employees, establishing parameters, and tracking what actually happens. These are not policies to add after employees start using AI. They are the foundation for using it effectively, and that foundation requires ongoing attention. Goals shift, employees change, and tools evolve. Without continued oversight, different parts of an organization can quickly begin pulling in different directions.
AI can strengthen a well-prepared organization, but maintaining the benefit takes constant effort. In a disorganized one, AI can scale existing problems with remarkably little effort.