As I boarded a train to attend my first event of Boston Tech Week, someone boarding at the same time as me asked what event I was attending and then what I do for work. I told him that I started my own Consulting and Advising Practice and his only response was a question: "Why Not AI? If I was a founder, why should I choose to pay you instead of just using an AI?" As I answered his questions, I realized this would be an excellent article to write for my website.
To keep this concise, I will share my top 3 reasons why you should still hire experienced people instead of just relying on AI.
1. Using AI often feels like playing the game "2 Truths and a Lie".
For those unfamiliar, "2 Truths and a Lie" is a social icebreaker game for groups where you come up with 2 true facts and a lie about yourself, and the group tries to guess which one is the lie. Working with AI output can often feel like a more complex and consequential version of this game, where the truths can be genuinely helpful advice, but the lies can be anything from inconsequential mistakes or misleading half-truths all the way up to liability-inducing business advice or production-destroying technical direction.
Trying to identify the "lies" without experience and expertise in the subject matter can be difficult, time consuming, and risky. Hiring an expert to provide important context and help you avoid potentially disastrous decisions will pay off in the long run. Building a business is already a risky endeavor, why would you amplify the risk by choosing an AI over an expert?
2. Knowing what questions to ask and how to ask them is invaluable.
AI is decent at making precisely what you ask for, but knowing what to ask for and how to word the request is something that can take years of working with the topic to get right. A few misaligned requests usually won't cause noticable issues, but continuous misaligned additions to a tech stack builds technical debt and a pile of liabilities waiting to collapse.
From my personal experience working with non-technical people trying to write code with AI, I find that many of the issues with the output come from leaving out a detail in the request, describing a request in wording that is too similar to a different concept, or trying to use a framework in a way it was not designed to work. Sometimes, the AI notes potential mismatches between the request and the capabilities of the technology you are using, but not always; so hiring an experienced technical person to your team can help avoid much of the mismatch between your actual needs and what the AI generates for you.
3. AI rarely produces "new" material and is reliant on what it saw when training.
Without getting too technical, AI generates responses based on the material it was fed when training. As such, it struggles to generate truly "new" information, often providing an output that is an amalgamation of publicly available information with tweaks to make it appear to answer the prompt. In a lot of cases, this is perfectly acceptable; but when you are making decisions that impact the future of your business, relying on an assortment of automatically gathered information may lead to decisions that are not the right fit for your situation.
In software specifically, the nature of how AI is trained means its output can be impacted by a lot of issues found in publicly shared code. For one, many public codebases contain a large assortment of issues, poor practices, and outdated standards, not to mention how many abandoned, unfinished projects are available to browse online. The problems found in these sources can leak into the output you receive when working with an AI, leaving your business vulnerable. Further, AI has little access to proprietary codebases, meaning it has no reference point for most of the code that powers successful, established businesses. AI also routinely struggles with uncommonly used technologies, as there is little material available for it to "understand" how the tech works.
Software is a rapidly evolving field, with security vulnerabilities and errors found frequently and standards changing regularly. AI can be great at providing a foundation for experienced engineers to build upon, but choosing AI over a software engineer is an admission that your company's technology is largely comprised of publicly available components, may not reflect a high standard of security compliance, and is likely going to experience instability; all of which can have costly consequences.
Conclusion
If you want to gamble with your business while having a lack of clarity and no guiding hand, just use an AI. You will probably be okay for a while, but with time, you will likely run into a problem that no AI can solve and no human will want to sort out. If you instead want someone who knows when something is wrong, knows the details and intricacies of the real world, knows what costs and benefits are relevant to your business, and can help provide clarity and peace of mind, hire me.