Most people who struggle with AI are not struggling with the technology. They are struggling with communication.
That distinction matters, because communication is something you already know how to improve.
Here is the pattern I see repeatedly, across organizations and job functions: someone opens an AI tool, types a quick request, gets something back that is technically correct but practically useless, and concludes that AI is overhyped. What they actually produced is workslop — output that is polished on the surface and hollow underneath. Technically generated. Practically worthless.
Workslop is not an AI problem. It is an input problem. Garbage in, garbage out still holds in the AI age.
The three-part prompt most people skip
When you sit down to use AI, you already know what you want done. That is the task, and it is the only piece most people include in their prompt. What they leave out are context and intent — and those two omissions are responsible for nearly every frustrating AI interaction.
Think of it this way. Context is orientation: who is this for, what is the situation, what constraints matter here. Intent is purpose: what does success actually look like, and why does it matter.
The running shoe example makes this concrete. Walk into a running store and say "I need running shoes," and you get a wall of options. Walk in and say "I have low arches, a $200 budget, I hate neon colors, and I'm training for a marathon" — now the person helping you can actually help you. The second version is not more words for the sake of it. It is the right words. Context narrows the space. Intent points at the outcome.
Context is the foundation. Get it wrong, and your intent does not matter. If the AI does not understand who this is for and what situation they are in, it cannot serve the purpose you have in mind.
Why this feels harder than it should
There is a cognitive trap here, and it has a name: the curse of knowledge. Once you know something, it is almost impossible to remember what it was like not to know it. Your situation feels obvious to you. You assume the person you are talking to has the same shared history, the same institutional memory, the same understanding of what "good" looks like in your organization.
With other humans, you often get away with this. Shared history fills in the gaps. With AI, there is no shared history. There are no corridor conversations, no meeting after the meeting, no years of working alongside your team. AI is the most low-context entity you will ever interact with. It has to be told explicitly what a human colleague would pick up without being asked.
As the CEO of Shopify put it, running a company is just context engineering. Every strategy document, every brief, every alignment meeting — all of it is an attempt to give people the context they need to make the right call. When that context is missing, you get politics, missed deadlines, and work that misses the point. The same is true with AI. The machine cannot read what you do not explicitly write.
This is not AI's limitation. It is a communication skill gap that the arrival of AI has made suddenly, immediately visible.
What goes wrong when intent is misaligned
Intent is the part that surprises most people, because misaligned intent does not always look like failure. Sometimes it looks like results — just not the right ones.
Consider what happened at Klarna. They replaced a significant portion of their customer service team with AI and gave the agents a clear directive: keep call times low. The AI was extraordinarily good at it. And it destroyed customer relationships in the process. The explicit measure was call time. The actual intent — building trust, solving the real problem, making the customer feel heard — was never stated. The AI had no way to know what was left unsaid.
This is not a story about AI going wrong. It is a story about intent engineering failure. The same failure mode exists in every organization where people are measured on a proxy instead of the purpose it was meant to represent.
When you work with AI, you have to state the intent explicitly. Not "keep call times low" — but "resolve the customer's actual problem in a way that builds confidence in us." Not "draft this email" — but "reschedule this meeting without losing the client's trust, and be specific enough about timing that they know we are serious."
Intent is what separates work that gets done from work that gets done right.
The two-sentence check
Before you use any AI output, ask yourself three questions. Would a senior colleague look at this and recognize that you understood who it was for? That you understood the situation you were in? That you understood what you were actually trying to achieve?
If the answer to any of those is no, the context or intent was missing before you started.
The fix is not complicated. Before any AI task, add two sentences to your prompt. One sentence for context: who is this for, and what is the situation. One sentence for intent: what does success look like, and why does it matter here. That is it. Two sentences. Most of the rework you are doing right now is the cost of skipping them.
This is not a new skill. You have been doing context engineering every time you briefed a new hire, wrote a strategy document, or explained to someone why a decision mattered. AI did not create the communication gap. It just made the cost of that gap immediate and visible, every single time.