The Problem With How We Talk About AI
In conversations with my clients and at conferences, I keep encountering two very different perspectives on AI. On one side are people who believe AI is the single most important thing in today’s economy and that every organization needs to rapidly adopt it or risk being left behind. On the other side are people who resist using AI entirely because it cannot perform work in the same way a human can.
Personally, I think the answer lies somewhere in between those extremes.
AI is a tool, and like most tools, its usefulness depends heavily on how it is used. It can absolutely help people move faster, organize information more effectively, and reduce repetitive work. At the same time, it can also create confusion, amplify mistakes, and generate misleading outputs when the information being provided to it is incomplete or inconsistent.
Lately, I’ve been thinking about AI in comparison to a much older invention: the knitting machine.
What Knitting Machines Can Teach Us About AI
When knitting machines were first introduced, they transformed textile production by allowing clothing to be produced much faster than it could by hand. Sweaters, stockings, shirts, and countless other knitted goods could suddenly be created at a scale that had never been possible before.
But despite how revolutionary the machine was, the quality of the final product still depended entirely on the inputs.
The machine could only produce based on the pattern it was given. If the pattern contained mistakes, the finished product would contain mistakes as well. If the wrong pattern was loaded, the wrong item would be produced entirely. The yarn mattered too. High quality yarn produced durable sweaters that lasted through many seasons, while poor quality yarn created garments that quickly fell apart. Knots or tangles in the yarn could create imperfections or even jam the machine altogether.
In other words, the machine itself was powerful, but it was never independent from the quality of the materials and instructions being fed into it.
AI operates in a very similar way.
Why Context Matters in AI
The outputs we receive from AI are deeply connected to the quality and structure of the inputs we provide. The data, documents, terminology, rules, and prompts we give to AI shape the answers we receive back.
If organizations provide conflicting policies, AI may blend those policies together into confusing guidance. If terminology is inconsistent across departments, AI may misunderstand the meaning of key concepts. If prompts are vague, the responses will often be vague as well. Sometimes the output can sound polished and confident while still lacking the precision or context that the organization actually needs.
One of the most common responses I hear is that AI already has access to the internet and has been trained on an enormous amount of human knowledge. People ask how the inputs could possibly still be the problem.
But I think the more important question is whether AI has access to your organization’s context.
AI Still Requires Human Judgment
AI may understand general information, but it does not automatically understand your company’s vocabulary, policies, business rules, priorities, or decision-making processes unless those things are intentionally provided. It does not inherently know the nuances that make your organization unique.
Too often, we expect AI to function almost like a mind reader. We hope it will somehow understand the full context behind the question we are asking, including all of the assumptions and organizational knowledge that exist in our own heads.
But even highly advanced AI systems cannot reliably infer context that was never shared with them in the first place.
And honestly, they probably should not.
Without organizational context, AI defaults toward generalized information pulled from broad public knowledge. Over time, that can flatten an organization’s unique expertise, voice, and competitive advantage into something generic. At the same time, when organizations provide disorganized or logically inconsistent information, AI tends to reflect that confusion right back to us. To reduce this confusion, AI requires a human to create clarity for it.
In many ways, this feels similar to every major technological shift we have experienced before. The technology itself is powerful, but the surrounding systems, processes, and human judgment still matter deeply.
AI does not remove the importance of context, vocabulary, rules, or expertise. If anything, it increases their importance.
The machine was never the entire story. The materials, the pattern, and the people using it always mattered too.