Things happen fast.
At first, computerized technology could only do algebra. A few short years later, it was dominating Jeopardy!. Nowadays, it’s better to ask what this technology can’t do — especially considering the advent of generative AI.
In the last year, generative AI has become ubiquitous; ChatGPT and Bard are now household names, and every week a new company releases a new product with a new generative AI feature that will “change everything.”
For recruiters looking to upgrade their tech stack, it can all be a little overwhelming, especially if you’re someone who’s still learning how to make their technology do algebra. We understand that you may not be familiar with how generative AI works, and that’s okay. It’s not your job to be an expert — it’s ours.
And the thing is, you don’t have to be an expert on generative AI to use it effectively. You just have to know the basics and how it can be used to solve specific problems. Your problems.
How is generative AI different from other implementations of AI?
Generative AI is a subset of artificial intelligence that sets itself apart by its capability to create new, original content, with minimal or even zero direct human guidance. Previously, many AI systems, including those based on natural language processing (NLP), would interpret user inputs and match them to the closest predefined response or action.
Think of this like your typical chatbot. If you asked it what the weather was outside, it would understand your intent (what’s the temperature?), and respond with a pre-programmed output (it is XX degrees outside). But if you then asked “what else?” things would get tricky. Without contextual awareness, this simple question would actually be too complex for NLP-based AI to understand.
In contrast, generative AI leverages models that can understand and produce content by learning from vast datasets, allowing it to generate responses or outputs that are not explicitly programmed into it. With generative implementation, the AI ditches templated responses and instead replies within the context of the conversation, effectively simulating human-like dialogue. So if you asked the AI “what else?”, it would remember the context of the conversation (it already gave you the temperature), and instead opt to give you something new, like advising you to wear some sunscreen.
Generative AI leverages models that can understand and produce content by learning from vast datasets, allowing it to generate responses or outputs that are not explicitly programmed into it.
Why generative AI works for TA.
In the context of recruiting, we can replace these weather examples with job-related questions. Organizations can train their generative AI assistant on their “Ground Truth” — a collection of documents and information regarding current openings, benefits, company values, etc. When confronted with a question, the AI will understand the candidate's intent, retrieve the relevant content from the organization's Ground Truth, and automatically generate a dynamic answer.
Say a candidate asks for the benefits of the job they’re interested in applying for — the AI assistant can generate a response based on the job description and career site. And if the candidate seeks clarification, or asks a follow-up, the assistant can elaborate and respond in a more human way than an NLP-based AI. Since the answer is dynamic, if 10 different candidates ask the same question in 10 slightly different ways, the assistant will create a different, personalized response to each of them.
Generative AI responses feel less like individual answers to questions and more like interconnected pieces in a whole conversation. Because generative AI can understand previous context gained in conversations, its responses feel bigger, fuller, and more personalized — similar to a text series you’d have with a human recruiter.
This experience gives the AI a higher sense of competence, which in turn translates to a candidate experience with less friction. Candidates can now get their questions answered and elaborated on faster and smoother than ever before.
What type of AI is right for me?
With generative AI, candidate conversations become 100x more powerful — and that’s really just the start. As we develop generative AI capabilities, conversations will become more and more personalized. More and more human.
As for right now though, generative AI is still a relatively new frontier of technology. And like all new technologies, there will always be some level of uncertainty. Generative AI’s lies in its lack of total control. Unlike NLP-based algorithms, where you instruct AI on how to respond to prompts, generative AI algorithms dynamically produce new responses on the spot. They generate something new. And in that sense, you can’t guarantee word for word what the AI will say.
But that doesn’t mean you can’t “teach” the AI to respond in a certain way that’s suitable to your standards. By setting up parameters from which the AI can pull, you can steer generated content toward desired outcomes. So when a candidate asks about your brand values, the AI will always pull from a consistent knowledge base: your career site, job description, etc. While you won’t be able to predict the sentence structure your AI uses, the content will be accurate.
Ultimately, your AI preference will likely align with your opinions on control.
For organizations willing to sacrifice 1% of control in exchange for significantly more complex, human conversations, generative AI will be a powerful part of any hiring solution. For organizations that value total control over AI responses, NLP-based AI will likely be a better solution. And for organizations in the middle, you can implement generative AI, but then define certain intents (like legal questions) that when encountered will reroute the assistant to always respond with predefined templates.
Using NLP-based AI doesn’t mean you will be providing your candidates with a bad experience. Generative AI isn’t necessarily smarter or better than NLP — it’s just a different implementation of AI. NLP-based AI assistants still provide customized experiences and conversations towards candidates, which altogether equate to high approval ratings and conversion rates.
Sure, generative implementations are certainly a leap closer to smarter, human conversations with AI. But think of it less as a leap forward from NLP and rather a leap upward. No matter which form of AI you implement, the candidate experience is going to improve — it’s just going to look a little different algorithmically.
