The AI skills your team actually needs

Prompting: the skill nobody talks about properly

‍

Prompting isn't typing questions into a box. It's thinking clearly about what you want and asking for it in a way that gets results.

Good prompters know how to be specific. Instead of "write an email to my client," they say "write a follow-up email to a client who missed our meeting yesterday, friendly tone, suggest two alternative times next week, keep it under 100 words." The difference seems small but the results are completely different. The vague prompt gets you a generic template that sounds like it came from 2005. The specific prompt gets you something you can actually send with minor tweaks.

Good prompters also know how to give context. They don't just paste a document and say "summarize this." They understand the AI needs to know why you're summarizing and what matters. So they say "this is a vendor contract for cloud services, we're deciding between three vendors, pull out the pricing terms and any cancellation penalties." Now the AI knows what to look for. Context shapes everything about how the AI approaches your request.

Good prompters know how to iterate. Their first attempt rarely works perfectly. They look at what the AI gave them, figure out what's wrong or missing, and adjust their approach. "Make it shorter." "Use more specific examples." "Focus on the risks, not the benefits." They treat it like a conversation where you're refining what you want through back and forth, not like a one-shot request where you either win or lose.

This sounds simple. It's not. Most people type vague requests, get mediocre results, and conclude AI doesn't work. They never learn that the problem isn't the AI, it's how they're asking. The people who get it understand that AI is a conversation, not a vending machine. You don't get what you want on the first try. You refine. You experiment. You learn what works for your specific tasks and your specific way of working.

Critical thinking: knowing when AI is wrong

‍

AI hallucinates. It makes things up. It sounds confident about information that's completely wrong. Your team needs to know this viscerally, not just intellectually, and they need to act accordingly every single time.

Good AI users treat output as a draft, not a finished product. They check facts. They verify sources. They know which parts need human review and which parts are safe to trust. This judgment doesn't come from a checklist. It comes from understanding your domain well enough to spot when something feels off.

Here's what this looks like in practice. Someone uses AI to draft a proposal. They don't send it directly to the client. They read through it carefully, fix the parts that sound generic or templated, add specific details the AI couldn't know about the client's situation, remove the section where it made up a statistic about industry growth rates. The AI saved them two hours of writing time, but they still spent 30 minutes making it actually good. That's the right balance.

Someone uses AI to analyze data. They don't present the results in a meeting without understanding what the AI did. They check if the analysis makes sense given what they know about the business. They spot when the AI misunderstood the question or made assumptions that aren't valid for their specific situation. They use the AI analysis as a starting point for their own thinking, not as the final answer.

The people who get it have a healthy skepticism. They use AI to move faster, not to stop thinking. They know AI is a tool that amplifies their judgment, not replaces it. They're comfortable saying "the AI got this part wrong" and fixing it without feeling like they failed. The people going through the motions copy and paste AI output without reading it carefully. They trust everything the AI says because it sounds authoritative. They get burned when something goes wrong and then they either stop using AI entirely or they keep making the same mistake.

Workflow design: figuring out where AI actually helps

‍

Not every task benefits from AI. Your team needs to recognize which parts of their work AI can handle and which parts it can't. This requires understanding both what AI is good at and what your actual work involves at a granular level.

Good workflow designers break down their tasks. They look at a big complicated process and identify the repetitive, time-consuming pieces that follow predictable patterns. Then they test if AI can handle those pieces. They don't assume. They try it, measure the results, and decide based on evidence.

Example: Creating a client report. The AI can pull together data from multiple sources, write the first draft of explanations for what the numbers mean, format tables consistently so everything looks professional. It cannot make strategic recommendations about what the client should do next based on their specific business context and competitive situation. That needs human expertise, judgment, and relationship knowledge. So a good workflow uses AI for the mechanical parts and reserves human time for the parts that actually require expertise.

Good workflow designers also know when to stop using AI. If it takes longer to fix AI output than to do the task yourself, stop using AI for that task. If the AI keeps getting something wrong and you're spending all your time correcting the same types of errors, that's not a good use case. Some tasks just don't fit AI's strengths. Recognizing this quickly saves massive amounts of wasted effort.

The people who get it experiment with different approaches. They try AI for a task, measure if it actually saves time or improves quality, and adjust based on real results. They're honest about what works and what doesn't. They'll tell you "we tried using AI for customer onboarding emails but it kept missing the nuances of each customer's situation, so we went back to templates." That's not failure. That's learning.

The people going through the motions try to force AI into every task because they're supposed to be "using AI" and they want to show adoption metrics. They waste time on bad use cases and never measure if it's actually helping. They confuse activity with progress.

Pattern recognition: spotting what works

‍

After a few weeks of using AI, patterns emerge. Certain types of prompts work better. Certain tools fit certain tasks. Certain workflows consistently save time while others consistently create more work than they save.

Good AI users notice these patterns. They pay attention not just to individual successes and failures but to what those successes and failures have in common. They document what works and share it with others. They build a mental library of "this type of task works well with this approach" that gets more sophisticated over time.

Example: Someone discovers that Claude is better at analyzing complex documents with nuanced arguments while ChatGPT is faster for quick summaries of straightforward information. They start routing tasks accordingly. They share this insight with their team so others don't have to figure it out from scratch through trial and error. Now the whole team is more effective because one person noticed a pattern and shared it.

The people who get it are always learning. They pay attention to what works for others. They adapt successful approaches to their own tasks instead of insisting on doing everything from scratch. They contribute to the collective knowledge instead of hoarding what they learn. They treat every experiment, successful or not, as data that helps everyone get better.

The people going through the motions do the same thing every time, whether it works or not. They don't learn from failures because they don't really analyze why something failed. They don't share discoveries because they're not paying enough attention to have discoveries worth sharing. They exist in isolation and they stay stuck at the same level of effectiveness for months.

Collaboration: learning together, not alone

‍

AI skills develop faster when people learn together. The lone wolf approach doesn't work here because the solution space is too large for any individual to explore efficiently.

Good AI users share their prompts. They show others their workflows, including the messy parts and the failures. When they discover something useful, they don't keep it to themselves like it's a competitive advantage. When they hit a problem they can't solve, they ask for help instead of struggling in silence. This isn't about being generous. It's about being smart. Your team's collective learning compounds when people share.

They participate actively in the shared Slack channel or Teams group. They attend the weekly sessions and they come prepared to show what they're working on. They demonstrate what they're doing, failures included, because failures teach as much as successes. They treat AI learning as a team sport where everyone's experiments make the whole group smarter.

The people who get it understand that everyone's experiments create value for everyone else. Your failure saves me time because now I know not to try that approach. My discovery helps you because you can adapt it to your work. We're all exploring different parts of the same territory and sharing maps as we go. This creates a multiplicative effect where the team's collective capability grows much faster than any individual could grow alone.

The people going through the motions work in silence. They don't share because they're either embarrassed about their failures or protective about their successes. They don't ask questions because they don't want to look stupid. They struggle alone with problems others have already solved. They miss out on insights that could transform their work because they never see what others are discovering. They stay isolated and they stay stuck.

How to spot who gets it

‍

Watch for these behaviors in your team because they reveal who's actually developing real AI skills versus who's just checking boxes.

The person who gets it is experimenting constantly. They try new approaches. They iterate on prompts to see what produces better results. They test different tools for different tasks because they understand that one size doesn't fit all. They measure results, not in a formal way necessarily, but they pay attention to whether something actually helped or just created busy work. They adjust based on what they learn. You see them getting better week by week.

The person going through the motions uses AI the exact same way every time. They found one prompt that kind of works and they never change it. They don't explore alternatives. They don't adapt based on results. They're stuck in a local maximum because they stopped experimenting once they found anything that worked at all. You see them doing the same thing in month three that they did in week one.

The person who gets it talks about specific results. "I cut my report writing time from three hours to one hour by using AI to generate the first draft and then spending my time on analysis instead of formatting." "This approach failed because the AI couldn't understand our specific terminology, so I had to create a glossary to include in my prompts." "Here's what I changed to make it work for our use case." They speak in concrete terms about concrete outcomes.

The person going through the motions talks in vague terms. "I'm using AI." "It's pretty helpful." "I think it saves time." They can't articulate what's actually working or why. They can't explain their process. They can't teach others because they don't really understand what they're doing themselves. They're going through motions without building real understanding.

The person who gets it helps others. They share what they learned without being asked. They troubleshoot when teammates are stuck. They build on other people's discoveries and credit them. They see the team's collective success as their success. They invest time in helping others because they understand it creates a better environment for everyone, including themselves.

The person going through the motions hoards knowledge. They see AI skills as a competitive advantage they need to protect. They don't contribute to the team's collective learning. They help others only when forced to. They miss out on the collaborative benefits and they end up learning slower as a result. Ironically, their attempt to maintain an advantage actually holds them back.

These skills can be learned

‍

Here's the good news. None of these skills require technical background. You don't need to understand how AI works under the hood. You need to understand how to work with it effectively, which is a completely different type of knowledge.

Some people pick this up in days. Others take weeks or months. The difference isn't intelligence or technical ability. It's mindset and approach.

The people who learn fast are curious. They're willing to fail and they don't take failures personally. They see AI as a tool to explore, not a test to pass where there are right and wrong answers. They share openly because they're more interested in learning than in looking smart. They learn from others because they're not too proud to admit someone else figured out something useful. They treat the whole process as an adventure rather than a chore.

The people who struggle are waiting for perfect instructions. They want to be told exactly what to do in every situation. They're afraid of looking stupid so they don't experiment. They work alone because asking for help feels like admitting weakness. They treat AI as something to master through formal training rather than something to explore through practice. They're waiting for certainty in a space that's fundamentally about experimentation.

Your job as a leader is to create an environment where the first group thrives. Where experimentation is celebrated, not just tolerated. Where sharing is expected as the default, not praised as exceptional. Where failure is learning, not career risk. Where people feel safe trying things that might not work. Where the person who discovers that something doesn't work is valued as much as the person who discovers something that does.

Because these skills, the real AI skills your team needs, they don't come from courses or certifications or formal training programs. They come from doing. From trying things and breaking things and fixing things and sharing what you learned along the way. From working together and building on each other's discoveries. From treating AI as a tool that amplifies human capability rather than as a replacement for human thinking.

That's how people actually get good at this. Not through training. Through practice, experimentation, collaboration, and continuous learning.