.png){kind=small}
You're Using AI. But Are You Using It Well?
- Mar 12
- 14 min read
Updated: May 14
What to Automate, What to Protect, and What to Question: A Leader's Honest Guide
Three years of watching leaders adopt AI tools taught me one thing: having access to powerful technology and knowing how to use it strategically are two very different things. This is a framework for closing that gap — honest, practical, and built from what I've actually seen in the field.
About three years ago I started noticing a pattern.
Leaders were adopting AI tools quickly and enthusiastically. That part was expected. What came next wasn't: the sprawl. Unstructured delegation to tools that hadn't been given clear parameters. Over-reliance on outputs that hadn't been critically examined. A kind of sloppy fluency where speed became the only metric that mattered.
Underneath all of it, something that concerned me most: a fundamental misunderstanding of what these tools actually are. Generative AI outputs are only as good as the models they're built on. Many tools marketed to leaders are narrowly trained within limited ecosystems, producing outputs that sound authoritative but reflect a fraction of the context required to make good decisions. Leaders who don't understand that are operating with tunnel vision, optimizing within a ceiling they don't even know exists.
I felt the pull myself. My journalism and communications training is what kept me honest. There's a specific instinct you develop when you've spent years working with language professionally. Something sounds off. The rhythm is wrong. The words are technically correct but nobody actually talks that way. I recognized it fast and corrected course.
Not everyone is expected to have that instinct. And when outputs disappoint — when the AI produces something generic, or wrong, or just not quite right — many leaders draw the wrong conclusion. Rather than interrogating how they were using the tools, they retreated from them altogether. That retreat carries its own risk. The adoption window is real and it won't wait for anyone to feel ready.
The conversation about AI is loud and moving fast. This is a small attempt at slowing it down just enough to think clearly.
The adoption curve tends to follow a predictable arc. Initial skepticism. A moment of genuine surprise at what the tools can do. Rapid integration into existing workflows. Then, somewhere around month three or four, a nagging sense that something is off but difficulty naming what it is.
What is off, in most cases, is not the tool. It is the absence of a strategy for using it.
When you adopt any powerful capability without a deliberate framework, you default to using it wherever it creates the most immediate relief. That's rational in the short term, but it creates a specific problem that doesn't show up immediately but later, when you realize that certain kinds of thinking feel harder than they used to.
I call this cognitive outsourcing. Letting AI do the thinking you would have otherwise done yourself. Each decision is small enough to feel inconsequential, but the cumulative effect is not.
So the problem isn't that any single one of these choices is wrong - it's that they weren't choices at all. They were defaults. And defaults, in the absence of strategy, optimize for speed and convenience rather than for the capabilities you need to lead effectively over time.
This became apparent when a leader I worked with copied AI-generated content directly into a legally binding document. They didn't set parameters and they didn't conduct a critical review of the language. To be fair, the output was dense, authoritative-sounding, and incorrect only in ways that were genuinely difficult to detect. The AI had done the thinking and the leader had signed off on work they hadn't actually done. That is not efficiency. That is exposure.
Now this is not to say that you have to do ALL the work you sign off on, what is the point of that? Same way you trust that highly capable people producing work for you will deliver reliably good products that you then review and verify - that's how you should think of AI, but with an even keener eye. Capable of quickly producing fairly high quality work but definitely needing review, validation, and the human lens.
In my work with founders and leaders on strategic communications, one of the first things we sort out is how to actually use AI well. Not whether to use it. How.
The framework I come back to consistently is simple. Think about everything you do in your professional role and sort it into three categories.
Execution work covers tasks that require a competent process but not judgment. Scheduling, formatting, research, first draft production, data organization, meeting transcription. The quality of the output is verifiable. The thinking required is mostly procedural. This is where AI earns its place. Use it freely, prompt it specifically, and review the output with a critical eye.
Judgment work covers decisions and communication that require the synthesis of context, nuance, relationship awareness, and values. How you respond to a difficult stakeholder. How you frame a strategic recommendation. How you read a room and decide what to say. These tasks require YOU. Not a version of you mediated by a language model. Do this work yourself first. Then, if useful, use AI to stress test your thinking or identify gaps. The thinking must originate with you.
Development work is the thinking, writing, and problem solving that builds your capability over time. The strategic memos you write that capture your ideas, your approach to a problem, or your insight on a subject. The analyses you do manually not because automation is unavailable but because doing the work yourself keeps you close to the information and data. Writing these sharpens your judgment. Protect this category deliberately. Validate or improve upon your own work with AI once it exists. Do not let AI replace the act of creating it.
The question is not whether AI can help with any of these categories. It can. The question is what parts of your work you let AI touch, how much, and in what order.
WHAT TO AUTOMATE
Let's start here.
First draft production for low stakes documents. Meeting summaries, status updates, routine client communications, templated proposals, research briefs, agenda preparation. These are documents where the goal is accurate, clear communication of information that already exists. If you have been writing these from scratch, you have been misallocating your time. Let AI produce a working draft and spend your time editing for accuracy and voice rather than generating from zero.
Gathering research and resources on well defined questions. This is one of the areas where I use AI consistently in my own work. When you need to understand a topic, a market, a competitor, or a regulatory landscape, AI is genuinely useful for accelerating the aggregation phase. Where you apply discipline is in verifying sources and treating AI generated summaries and information returns as a starting point for your own analysis, not as the analysis itself.
I want to be specific about where this can break down. I worked with a team that used AI not just for data aggregation and pattern identification but for the high level insight that was supposed to follow - what you get paid the big bucks for. The result was an analysis that looked thorough and read smoothly but felt completely canned. The observations were generic and the recommendations could have applied to any organization in any industry. Nobody in the room could defend the conclusions because nobody had actually done the thinking. The data had been processed and that's it. There was never an actual analysis that provided the insight that is most valuable - and there is a big difference between the two.
Meeting preparation and follow up. Pre-meeting briefings, capturing post-meeting action items, follow up email drafts. These are high visibility, action-heavy, low judgment tasks that consume a disproportionate share of our time. Automating them is straightforward, the time savings are real, and there is no danger of losing meaning or insight in this work.
Editing and polish. This is another area I rely on personally. Proofreading, formatting consistency, adapting content for different audiences or channels, refining tone and flow. AI performs reliably here and the cost of occasional errors is manageable with a thorough review. What I protect is the original thinking and voice. What I use AI for is making sure it lands clearly.
I am particular about how I say things (my voice), what sounds right to me, and so I challenge the output all the time. Because I always begin with my own work and treat AI as an editor, I will frequently decide to keep original writing that it suggests I change or simply decide not to use added language. It's like the early days of Grammarly - as a writer I took small pride in running a doc through and getting a 97% accuracy score at first pass. It meant my grammar and syntax were spot on. But I double-checked and there was almost always an improvement that could be made.
Engineering and technical support. For those working in or alongside technical development, AI has become a genuinely powerful tool for code review, documentation, debugging support, and accelerating build cycles. The important caveat is the same as everywhere else. You need enough domain knowledge to know exactly what your asking for and challenge the output. AI in the hands of someone who cannot evaluate what it produces is a liability, not an asset.
The common thread here is that AI can be readily applied when the quality of the output is verifiable, the thinking required is procedural, and your judgment is applied in review rather than in generation. That is the right ratio for execution work.
WHAT TO PROTECT
This is where most leaders underestimate what's at stake. AI tools can produce some version of almost everything on this list. And the version they produce is often good enough that the temptation to let them is real. Resist it.
High stakes communication that carries your credibility. When you write a board presentation, a difficult message to a senior stakeholder, or a position paper on a complex issue, that document carries your name and your judgment. If AI wrote the first draft, the second draft, and you edited for tone, what is actually in that document? You need to be able to answer that question clearly. For high stakes communication, writing from your own thinking is critical - It is the work itself.
Decisions that require reading people and context. AI has no access to what you know about the specific people in a negotiation, the organizational dynamics at play in a restructuring, or the history that makes a particular piece of feedback land differently for one person over another. These contextual, relational, real time judgments are where experienced leaders create real value. No AI model can replicate what you know from being in the room and being human.
Feedback and performance conversations. If you are using AI to script difficult conversations in advance, you are practicing the wrong skill. The skill that matters in a performance conversation is the ability to listen, adapt, and respond to what is actually happening in real time. Have the conversation. It will be more difficult and more valuable than anything a script could produce. Creating an agenda to make sure you touch on certain points can be valuable, just don't spit back scripted pre-thoughts. It can come off as robotic and not at all dynamically responding to the live inputs you're receiving.
Your own thinking about your own role and direction. When you are working through a significant professional decision, a transition, a strategic pivot, a question about what you actually want, asking AI to analyze your options and hand you a recommendation is asking the wrong thing of the wrong tool. The output will be coherent and generic. The decision deserves more than that. This one is tricky though - there can be real value in stress testing options. One way this can be useful is asking for statistics on a particular scenario that may help you in your decision making.
What to do instead: write it out before you open any tool. Get your own thinking on paper first, however messy. Deliberate journaling — putting your actual reasoning, concerns, and instincts into words before any external input — is one of the most underused leadership practices I know. Once your thinking exists, you can bring it to a trusted advisor or coach who knows your specific context. Use a structured decision framework to test your conclusions. AI comes last in that sequence, as an expander, enhancer, and validator of thinking you have already done.
WHAT TO QUESTION
There is a third category that sits between automate and protect. We'll call it "other". In this "other" category of work, AI use is neither clearly beneficial nor clearly harmful. It just depends on how you use it.
Using AI for thinking out loud and problem exploration. There is genuine value in using AI as a thinking partner for exploratory conversations. My own approach looks something like the scientific method. I describe a problem statement or hypothesis, ask for a structural overview of what areas to develop, determine if any areas are missing or not needed, do my own writing, and may use AI for expansion where useful. Then I stress test, research, and fact check. The thinking originates with me. AI may accelerate the process and sometimes challenge it and, in turn, challenge the AI. The primary risk here is that AI's confident, articulate responses can create the illusion of having resolved something you have actually just described. Exploration is not resolution.
Using AI to prepare for high stakes conversations. There is value in using AI to think through how a message might land or to identify angles you haven't considered. The risk is substituting that preparation for adaptive real time thinking. Use AI to sharpen your understanding of the situation, not to produce language you will then deliver from memory. The goal is a clearer mind going into the conversation, not a script.
Using AI for professional development and learning. Having AI explain a concept or walk you through a framework is a legitimate use case. The risk is that passive consumption of AI generated explanations can create the feeling of having learned something without the actual cognitive work of integrating it. The test is simple. Can you explain it to someone else in your own words, without the AI, with concrete examples from your own experience? If not, you have been informed. You have not yet learned.
Using AI to accelerate your writing process. Having AI help you structure an argument or identify gaps in your reasoning is different from having AI write the piece. The former develops your thinking. The latter substitutes for it. The distinction matters and it requires honesty with yourself about which one is actually happening.
Okay, now that we know when we should use AI and how, I want to point out a pattern I have started seeing more consistently and that concerns me more than any of the others.
When leaders consistently outsource their thinking to AI, over time something shifts in how they communicate. Not all at once, but gradually. The most visible sign is not that their work gets worse in a technical sense. It is that their voice starts to disappear. They begin to sound like the AI instead of themselves. And not even a sophisticated version of the AI, but a generic one. The distinctive perspective, the tone, the way of seeing a problem that made them worth listening to in the first place — all flattens out.
This becomes genuinely concerning anywhere real time thinking is required: debate, extemporaneous speaking, responding to unexpected questions in a board meeting or a client presentation. These situations require agile thinking, deep subject command, and the ability to put it all together in a coherent and articulated message - under pressure. What is said here cannot be scripted in advance nor handed off. It requires a mind that has been doing the work consistently enough to draw on it instantly.
So, if you have been letting AI do the heavy cognitive lifting, your capacity to think in this way weakens. Not because you became less intelligent, but because you stopped practicing the things that keeps your thinking sharp. If used well, leaders' minds tend to get sharper over time because they are using AI to challenge and expand their thinking. Leaders who use AI as a replacement tend to get more dependent and less distinctive.
Are you losing your own voice?
It is worth sitting with that question honestly before moving on. For leaders it shows up in the boardroom and the inbox. For those who write creatively or communicate as a core part of their professional identity, the stakes are even higher. Voice is not a nice to have or a descriptor. It is credibility, authority, and distinctiveness reflected in language. Once it flattens, it is genuinely difficult to recover without deliberate effort.
Here is a simple audit worth doing.
On the automation side: are you still manually producing work that could be reliably automated without meaningful quality trade-offs? If so, why? Is it habit? A belief that doing it yourself is more trustworthy? Genuine strategic value in maintaining that skill? Each of these is a different answer and not all of them are wrong. Keeping certain things manual is a legitimate and often wise choice — if the deliberate practice creates a desired habit, prompts deeper thinking, or strengthens a capability you want to maintain. The point is not to automate everything, but to make an informed, strategic choice about what you automate and what you protect. Be intentional.
On the protection side: are there categories of work you have started outsourcing that you have not explicitly decided to outsource? What would you lose if you needed to do that work without AI tomorrow? How confident are you in your own judgment on the decisions you are making, independent of the AI generated analysis that informed them?
On the dependency side: when you sit down to write something difficult from scratch without AI assistance, what happens? Do you feel the same clarity and capability you would have two years ago? Or has something shifted? This is not a trick question. The answer is informative regardless of what it is.
On the development side: in the last six months, what have you written, analyzed, or thought through manually without AI assistance that developed your capability - rather than just produced an output? If the answer is very little, that is worth paying attention to.
If these questions surface something you want to understand better, the Leadership Focus Check-In at Quadrosa Coaching is a free, five minute diagnostic that identifies the patterns most likely affecting your focus and performance right now.
WHAT THE TOOLS CANNOT DO
AI cannot know what you actually want. It can help you articulate it and explore it, but only if you already have some version of it. And that is precisely where most leaders struggle.
Clarity is not something AI can give you. It is something you arrive at through the right questions, asked in the right order, by someone who understands where you are trying to go. The most meaningful insights I see leaders reach are almost always organic, produced through structured prompting, reflection, and the kind of guided thinking that allows their own conclusions to surface rather than being handed to them.
AI cannot substitute for the relationships that make leadership effective. The trust, the credibility, the understanding of specific people and contexts are built through real interactions over time. A language model that has processed millions of documents about organizational dynamics still has less relevant information about your specific team and stakeholder landscape than you have from three months of direct experience.
Nor can it replace the judgment that comes from having lived with complexity in a specific domain over time. Pattern recognition at scale is not the same as wisdom in context. The models are impressive at the former. They have no access to the latter.
And it cannot do the work of integrating what you know. The kind of integration that produces genuine insight about a complex situation does not happen in a context window. It happens in a human mind that has been shaped by years of serious engagement with hard problems.
WHERE TO START
Here is a practical starting point. Do the audit above, and then one week of deliberate practice of the following:.
Before using AI for any professional task, take thirty seconds to classify it. Is this execution work, judgment work, or development work? If it is execution work, use AI freely and review the output. If it is judgment work, do the thinking yourself first, then optionally use AI to pressure test or refine. If it is development work, protect the thinking but use AI as a resource, for research, reference, and exploration, while keeping the conclusions entirely your own.
At the end of the week, notice what you learned. Where were the decisions easy? Where were they hard? Where did you find yourself rationalizing judgment work as execution work because AI was convenient? That last question is the most informative.
The leaders who will navigate this well are not the ones who adopt AI fastest or resist it longest. They are the ones who decided early that how they use it is a strategic choice, not an accident. The tools will keep evolving. Your relationship with your own thinking is worth protecting. That distinction changes everything.
———
Jenny Fernández is a strategic advisor, leadership coach, and founder of Quadrosa and Quadrosa Coaching. She works with founders, executives, and organizational leaders on the clarity, structure, and thinking capacity that drives performance. quadrosacoaching.com
Comments