You’re Using AI — But Are You Using It Well? 3 Steps To Consider

The pace of AI development means even the well-informed feel left behind. Gen-AI can be transformational only if we learn to master it. Steps to take in your AI journey.

---

The other day I stumbled on a term I’d never heard before: SolidGoldMagiKarp. Originating from the anime world of Pokemon, it is the idea of something so rare to be almost unreal. The phrase was making rounds due to ability to cause glitchy behavior in AI when used as a prompt. And, I wondered if I had missed yet another concept in the ever-expanding universe of artificial intelligence.

That feeling — of brushing up against something new, strange and must-know — is felt by many and more often now than ever before. The rapid transformation of AI innovation is rivaled only by its pervasiveness in our lives. According to McKinsey’s AI in the workplace 2025 survey, nearly all employees (94%) and C-suite leaders (99%) use gen AI tools. And yet, just last year’s Gallup’s poll asking employees on frequency of AI use at work had nearly seven in 10 saying they never use AI, while only one in 10 employee reporting weekly use.

The dramatic change in AI in the workplace comes at a cost. Workers are more worried than hopeful about AI adoption; those who feel hopeful are mostly young professionals and those too scrambling to keep up.

Almost 80% of users are bringing their own AI tools to work (BYOAI). Lacking clear guidance on what constitutes acceptable use of AI, more than half are unwilling to admit using AI at work. If you are feeling like you are chasing a mythical carp, you are not alone — 77% of employees report being lost on how to use AI in their jobs.

AI Use ≠ AI Mastery

Symptom 1: Surface-level productivity

The advent of AI was meant to usher a transformation of how we work. And yet, Digital Work trends report found two-thirds of employees use AI primarily for cross-checking their work. Reason? It is a precarious time at work. Despite hopes of post-pandemic balance, meetings and after-hour work dominates a typical workday. Almost 70% report struggling with the pace and volume of their responsibilities, while nearly half say they feel burned out. A Microsoft 365 study found 60% of users spend their time on applications such as Outlook and Teams — responding, coordinating, and keeping up. In this environment, it's no surprise that employees reach for generative AI tools — not to innovate, but to simply keep up. If AI has to fulfill its promise, the scope of human bandwidth will determine the extent of AI mastery.

Symptom 2: AI Dependency without discernment

There is a tendency to trust AI generated work without verifying facts, citations, or data sources. Seldom do users critically evaluate if the content is accurate, complete, or unbiased. Educators have been sounding warning bells about plausible-sounding answers that are incorrect — hallucinations in AI-speak. But this issue is not limited to academic world. With work pressures and looming deadlines, AI provides a welcome short-cut to keep up with deliverables. Pick any AI tool. Trained to display a ‘know it all’ tech-authority while spitting out abundance of information, it lulls us to forget that it does not “know” things and is simply predicting plausible output based on training data. Is this behavior a reflection of our own tendency to “satisfice” — make decisions that are good enough rather than optimal, bound by the limits of our finite cognitive capacity, and real time constraints? AI’s limitations are architectural, not cognitive — but like humans operating under bounded rationality, it produces output that will suffice. The danger? Unlike humans, it doesn’t admit when it is wrong — and never signals uncertainty. Can AI companies take steps to have their tools acknowledge when a query pushes at its predictive boundary? They can and they should if hallucinations are to be managed. Without such guardrails, our AI dependency without due diligence will remain a liability.

Symptom 3: Ethical blind spots

It seems like we are living through the Wild West era of AI — where tools are advancing faster than the legal and ethical frameworks needed to govern them. Just like in the 1800s American frontier, this new territory shows promise — but also peril. Businesses are rushing to stake their claims, often with no rules of governance in sight. Consider the case of an HR manager who gets the go-ahead to integrate AI into the firm’s recruitment platform to streamline resume filtering. In a short period of time, it goes live and deemed a huge success as it slashes time-to-hire by half. Months later, it’s discovered that the AI systematically downgraded women and minority applicants based on biased training data — leading to legal exposure and public backlash. Such blind spots are rampant due to lack of clarity on questions such as ownership of the training data including liability for IP violations for any embedded copyrighted material or steps for auditing its algorithms and many more.

Why This Happens: AI’s Flooded Learning Curve

Generative AI tools are constantly ‘learning’ and the pace of advancement is such that even tech-savvy professionals are struggling to keep up. Jargon like multi-modal, tokenization, neural networks, retrieval-augmented generation abound and even playful terms like SolidGoldMagiKarp seem like secret passwords to a club we didn’t know existed. Many feel a growing pressure to sound competent with AI, even if they’re privately unsure of the whats and the hows. Others, excited to explore the new features and capabilities, end up with hours of wasted effort going down rabbit holes. Watching colleagues use custom bots or showing off complex prompts may make one question one’s techyness, falling prey to the classic imposter syndrome. Could it be that rather than drowning in data, today’s professionals are drowning in expectations? When the learning curve becomes a tidal wave, even the best talent is likely to tread water.

From Confusion to Control: Stepping up your AI journey

Step 1. Identify Your AI persona

Begin by assessing your relationship with AI. Reid Hoffman, in his recent work on AI and its future, introduced four personas to categorize how people think and feel about artificial intelligence: Doomers, Gloomers, Bloomers, and Zoomers. He describes Doomers as those who view AI as an existential threat to humanity; Gloomers with less extreme views but still fearful of AI’s role in deepening inequality, misinformation, and job disruption; Bloomers as cautiously optimistic of AI; and Zoomers, the early adopters embracing all things AI and viewing it as a force for growth and innovation. Reflecting on one’s relationship with AI is increasingly important, especially for professionals and leaders navigating rapid technological change. Being overly skeptical or overly optimistic will determine how you engage with AI. For instance, Doomer may shy away from valuable learning opportunities, while a Zoomer may be susceptible to overlooking ethical red flags. Leaders need clarity and insight on their own stance before fostering discussions around AI strategy, policy, and implementation for their organization. There is no ‘One size fit all’ AI strategy, but auditing your AI persona can help guide what works best for you and your organization.

Step 2. Audit Your Current AI Use to Go Deeper

AI can accelerate your workflow, but if you’re only using it to get things done faster, you may be missing an opportunity to learn and grow. Ask yourself: Am I using AI to save time — or to think better? Is there a way to pose this query from a different angle? Even if your current AI use is predominantly drafting emails, how about learning to automate this task by creating templates with AI’s help!

Next, reconsider how you typically treat AI output- review or copy and paste. Doing latter means you are risking errors or misjudgments or worse your job and reputation. Workplaces are increasingly adopting internal AI governance tools such as Microsoft’s Purview or plagiarism software such as Copyscape and Grammarly Business.

Instead consider the concept of Collab score that in essence is about well a human and AI collaborate on a task, whatever the task may be. Did you just accept the AI’s first output or did you revise it, critique it, build on it or even better sough follow-up questions? Collab score is a measure of interaction quality with higher score reflecting engaged, iterative use leading to high quality outputs and lowered risk, and thus, getting at the heart of meaningful human-AI teaming.

Step 3. Learn the Tool – but Also Understand the System

You don’t need to become a machine learning expert — but you do need to understand that effective AI use is an iterative process. The real value comes not from a single prompt, but from refining, questioning, and building on what AI gives you. Where to begin? Try LinkedIn Learning that offers concise videos on AI that targets different professions. Subscribe to credible sites such as MIT Technology Review’s AI section or newsletters such DeepLearning.ai’s. Brave New Words by Salman Khan of Khan Academy is another great resource. Though written for educators, it offers great insights on how to customize one’s learning journey, on using AI as an ethical guide helping navigate the web with one’s own filters of do’s and don’ts and finally, as an assistant willing to handle the mundane stuff, freeing time to innovate. And if all this seems effortful and time consuming, how about using AI as an entry point to AI. Reid Hoffman suggests navigating to your favorite gen AI model and starting a new chat with prompts that begin with, ‘Explain agentic AI to me like I’m five’, to ‘Explain agentic AI to me like I’m in high school’, and then graduating to ‘Explain agentic AI to me like I have a PhD’. Invest in learning about prompt engineering which in simple terms is about designing effective prompts to get accurate, useful, and consistent outputs from AI systems. The best news about learning AI from AI is that it does not judge — there are no dumb questions.

By Anjali Chaudhry, Contributor