The game is afoot in the world of artificial intelligence, but contrary to popular belief, AI isn't playing the role of the great detective. Instead, it's serving as our faithful companion, which is more Dr. Watson than Sherlock Holmes. And for those of us in technical fields, this distinction isn't just a literary metaphor; it's the key to understanding how we can best leverage AI in our daily work and why there's no "autopilot" for vibe coding complex systems anytime soon.
As the CEO of GitHub, Thomas Dohmke, recently stated on “The MAD Podcast with Matt Turck,” depending solely on automated agents could lead to inefficiencies, and it's oftentimes easier to correct AI-based code than try to explain in plain English how to get the AI to do the work itself. Yes, AI is incredibly capable, but it's not a Vending Machine.
When ChatGPT burst onto the scene, followed by a parade of increasingly sophisticated AI tools, the narrative quickly became one of replacement. "AI will automate away jobs," the headlines proclaimed. "The machines are coming for your career." But this fundamentally misunderstands the relationship between human expertise and artificial intelligence. Yes, jobs are going to be replaced in droves, and companies like Salesforce are already announcing they will no longer hire developers this year, but GenAI isn't a silver bullet, and experienced developers still have quite a bit to offer (and likely will, on into the future).
Consider how Holmes and Watson operated. Watson wasn't a lesser version of Holmes—he was a different kind of intelligence altogether. He documented cases, offered medical expertise, provided a sounding board for theories, and occasionally spotted details Holmes missed. Most importantly, he never replaced Holmes; he amplified him.
Today's AI tools operate in remarkably similar ways. They're exceptional at certain tasks such as pattern recognition, data processing, natural language generation, and rapid information retrieval. However, like Watson, they lack the intuitive leaps, contextual understanding, and creative problem-solving that define genuine expertise (leaving AGI out of this for the moment), as well as the creativity and intuitive problem-solving capabilities that come from years of experience in a technical field.
In software engineering, this Watson-like role manifests in countless ways. AI can generate boilerplate code faster than any human, suggest function implementations, and catch syntax errors. But it's the engineer who understands the system architecture, knows which edge cases matter, and makes the crucial decisions about trade-offs between performance and maintainability.
I've been working a lot with senior developers and seeing how they use AI with tools like Cline to build complicated micro-service-based systems. The AI would suggest code snippets, completed routine functions, and even propose test cases. Every suggestion, though, required the developer's judgment because there are cases where the solution won't scale, or it could violate a security protocol, or perhaps the code needs to handle async operations differently. Given that it can only program on what it knows, and developers like me aren't going to explain EVERYTHING the AI needs to consider, there will be a process in which Vibe Coding puts AI as the assistant but isn't going to simply go off and build everything.
The AI is much like Watson, then to Holmes - capable, helpful, even occasionally brilliant, but always requiring the detective's discernment.
Holmes famously claimed to use pure deduction, but a fan of the characters knows his greatest insights came from combining logical reasoning with intuition, experience, and keen observation of human nature. He could deduce a man's profession from his hands, but understanding why that man might commit murder required something more.
Similarly, AI excels at pattern matching and logical operations but struggles with the "why" behind the "what." An AI can identify that your application's response time degrades every Tuesday at 3 PM, but it takes a human to realize that's when the marketing team runs their weekly bulk email campaign on the same infrastructure.
This limitation isn't a flaw, but a feature. By handling routine deductive work, AI frees us to focus on higher-order thinking that defines expertise, which is where vibe coding excels today. Handling the parts that are wrapped around grunt work, commoditized coding, or procedural work, but leaving deeper context-rich thinking, creativity, and broader awareness of where the project will live in production to the engineer.
Watson's most valuable contribution often came not from his answers but from his questions. His queries helped Holmes clarify his thinking, explore alternative theories, and occasionally stumble upon the crucial insight.
Modern AI tools excel in this Watsonian role. They can help you explore possibilities you hadn't considered, challenge assumptions, and provide alternative perspectives. But like Watson, they need Holmes to direct the investigation.
Think of a data scientist using large language models not to find answers but to refine the questions, asking something like "I'll explain my problem to the AI and ask it to suggest what additional information might help solve it. It's like having a tireless research assistant who's read everything but needs direction on what matters."
It's a librarian who never gets tired of all the questions, available any hour of the day. It's an intelligent assistant that can act as a sounding board for your inquiry. And it's the helpful programmer who can handle the parts of your software you don't want to waste your time on. It can also be the chief of staff to whatever domain expertise you're looking to get support on, which may have nothing at all to do with technology.
Holmes trusted Watson implicitly, but he also knew Watson's limitations. He wouldn't send Watson alone to confront Moriarty or expect him to crack a complex cipher without help. This calibrated trust is precisely how we should approach AI.
Today, you can trust AI to:
Process large volumes of information quickly
Identify patterns in data
Generate first drafts and templates
Suggest solutions based on documented patterns
Handle routine, well-defined tasks
However, for now, don't trust AI to:
Make critical business decisions
Understand nuanced context or company politics
Take responsibility for outcomes
Innovate beyond its training data
Replace human judgment in complex situations
AI is a moving target evolving quickly, so this article may not have a long shelf life :) But it's safe to say, as AI becomes more helpful, more people will have more time to do the things they are most capable of doing (and likely the work they enjoy the most, but lack the time to do).
One of Watson's primary roles was as a chronicler of Holmes' adventures. He transformed complex investigations into understandable narratives. Today's AI excels at this same function, turning messy codebases into clear documentation, complex analyses into executive summaries, and technical specifications into readable prose.
But just as Watson's chronicles required Holmes to provide the actual detective work, AI-generated documentation needs human expertise to ensure accuracy, relevance, and appropriate detail. The AI can write the story, but only after the human solves the case.
"Hallucinations," for example, are often spoken about as a tragic flaw of AI, but if you approach the tool with the right coaching up front (rules, roles, and "pre-prompts") and guide it as it goes, you'll be amazed at the insights you can uncover. LLMs know practically everything but require the right approach to maximize their value and get the insights and support you need executed effectively and it's easier and easier to feed your agent first hand information and knowledge you yourself have on hand, to further hone in on what insights and direction are going to be most valuable to you.
The most successful Holmes stories weren't those where Watson was absent but those where the partnership functioned at its peak. Watson's medical knowledge complemented Holmes' detective skills. His military background provided practical experience. His humanity balanced Holmes' cold logic.
As AI capabilities expand, the Watson role will become more sophisticated. Future AI assistants might anticipate our needs better, handle more complex subtasks, and provide increasingly valuable insights. But the fundamental relationship remains: AI is a capable assistant to human expertise.
This isn't a consolation prize or a temporary state before AI "takes over." It's the optimal configuration. The combination of human creativity, judgment, and accountability with AI's speed, consistency, and processing power creates something greater than either could achieve alone.
Yes, AI is going to be a massive disruptor for current jobs, as people predict AI taking 50-100% of all white collar jobs globally in the years to come, but new jobs will be created at the same time, which will rely on people gaining domain expertise in an area and leveraging AI to maximize the services and capabilities that individual can provide. AI won't replace your HVAC system or build your house (we will see what happens with the AI-powered robot revolution), so blue-collar jobs will undoubtedly grow in popularity in the coming years as our economic situation shifts.
However, AI today isn't a silver bullet (nor will it likely be any time soon), as I mentioned earlier, and understanding where it is and where it's headed means gaining a sense of where you fit into the picture, with the opportunity to leverage it immediately. Sure, there are Waymo taxis that drive themselves through downtown San Francisco, but there are still race car drivers who can handle tight corners and fast speeds, as well as semi-truck drivers who can maneuver large trailers through difficult roadways and restrictive alleys. Cars will evolve, but the types of individuals behind the wheel will become more specialized and operate in ways AI won't be able to for some time.
The professionals thriving in the AI age right now aren't those trying to compete with machines at machine tasks. They're the ones leveraging AI to become better versions of themselves—more productive, more insightful, more creative. The common saying goes that it won't be AI that replaces you, but someone like you using AI.
Today, great use cases for AI include:
Eliminating drudgework and focusing on high-value problems
Exploring more solutions in less time
Validating ideas against vast knowledge bases
Communicating complex ideas more clearly
Learning new domains more quickly
But never forget that they're not Holmes in this story; you are. You set the direction, make the connections, take the responsibility, and ultimately solve the case.
The tech industry loves its disruption narratives, but the AI revolution looks more like an evolution, one that will change how we work and what we work on, not whether we work at all. Artificial Sherlocks are not replacing us, but rather displacing us in the coming years while we discover new work, roles, and jobs that perhaps don't exist yet. We're being given artificial Watsons to help us solve cases faster, in ways we didn't think possible before, and use this newfound assistant to evolve your own career into a niche you're best suited for (and enjoy doing the most!).
And that's genuinely exciting. While Holmes was brilliant alone, he was legendary, with Watson by his side. The same potential exists for every developer, analyst, engineer, and technical professional willing to embrace AI as their partner rather than fear it as their replacement.
The game is indeed afoot. But we're not competing against AI; we're investigating the future together. And like Holmes and Watson walking into the London fog to solve another case, the real adventure is just beginning. Elementary, perhaps, my dear Watson (I had to fit that in there somewhere), but sometimes, the simplest truths are the most powerful. AI is our Watson. We remain Holmes. And together, we're capable of solving problems neither could tackle alone.
Ready to lean in more with AI? Here are some links to start, or go deeper.
Wharton professor who tests AI on real business problems daily. His posts show exactly how to use AI for strategy, writing, analysis, and decision-making. Start with his "Prompting Guide for Work," as it's the clearest explanation of how to get AI to actually help rather than frustrate.
Upload your documents, reports, or research papers and have an AI that understands YOUR specific context. Like giving Watson your case files before starting an investigation. It can answer questions, find connections, and even create study guides from your materials.
Thoughtful conversations with AI researchers and critics about what this technology actually means for work and society. His episode "A Skeptical Take on the AI Revolution" with Gary Marcus provides essential balance, while his interviews with Anthropic researchers explain why AI works best as an assistant, not an autonomous agent.
AI search that shows its sources and reasoning. Unlike regular search, you can ask follow-up questions like "but what about..." and it maintains context. Essential for research, fact-checking, and exploring complex topics. The free version is limited; Pro ($20/month) is worth it for serious work.
Four-week course designed for non-technical professionals. Covers what AI can and cannot do, how to identify AI opportunities in your work, and how to work with AI teams. No coding required, just practical understanding.
Create specialized AI assistants that know your business context. Upload your style guide, past proposals, or industry reports. Your AI "Watson" then understands your specific domain when helping with analysis, writing, or strategy.
Designer and strategist documenting unusual ways to use AI as a thinking partner. Recent examples: using AI to explore "what-if" scenarios for product launches, creating alternative perspectives for stuck problems, and turning messy meeting notes into actionable plans.
See AI tutoring in action. Even if you're not studying, watching how Khanmigo guides rather than gives answers demonstrates the Watson approach perfectly. It asks clarifying questions, provides hints, and helps you reach conclusions yourself.
Harvard Business Review's series on AI as a colleague, not a replacement. Features case studies from consulting, healthcare, and finance showing how professionals amplify their expertise with AI. Episode 3 on "AI as Thought Partner" directly parallels the Holmes-Watson dynamic.
Transform rough ideas into polished presentations. You provide the expertise and key points; AI handles design and flow. Perfect example of Watson handling the documentation while you focus on the investigation. Used by consultants and executives who need professional outputs fast.
Start with one tool that matches your biggest daily friction. The best AI partnership begins when Watson helps with something that currently slows down your best work, or helps clear out the things getting in the way of the work you really want to be doing.
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.