If the phrase “Artificial Intelligence” makes your eyes glaze over, you’re not alone. The term “AI” gets thrown around a lot these days, often by people who don’t actually know what it means.
What exactly is AI and how can it affect your approach to recruiting? As Head of Data Science at Uncommon, I have some thoughts about this that I’d like to share.
Understanding Artificial Intelligence
AI is fundamentally about doing something at the same level or better than a human can do it. Automation alone (e.g. self-driving trucks, assembly line QA) itself can be considered AI, but is it sufficient? In some industries, sure. An AI logistics system can understand patterns in historical supply and demand, combined with real time conditions to deliver a car, groceries or other services on demand.n In this case, I’d say that automation combined with the speed and economy of scale afforded by it is a huge value add.
In recruiting, I believe we need to be a little more thoughtful. Hiring is not simply about efficiently identifying and allocating human capital. It’s far more nuanced and demands the seamless cooperation between intelligent data driven systems and hiring professionals. Does simply speeding things up or affording economies of scale alone really change the industry in the way that many of us want to? I think the answer is no. Performing the tasks that a recruiter normally performs at a faster rate isn’t enough—we need to use technology to enable fundamentally better outcomes for both job seekers and employers. Furthermore, we should use technology to democratize best practices in the industry, increase fairness and transparency, and accelerate the rate connecting employers to talent. This means breaking though personal and organizational bias, qualifying candidates based on transparent and objective merits, and empowering everyone to carve through the noise of the hiring marketplace with best practices in their pocket. This opens entirely new pools of talent and a vastly accelerated pace than available today.
Using AI to revolutionize recruiting
Let’s look at an example of how AI can be applied in the recruiting world. It’s possible to design technology that reads the text on a résumé faster than a human would and infer things from it. But if all it’s doing is just a fancier version of a keyword search, then it’s not really AI. Furthermore, if this black box system is making decisions that aren’t transparent to the human users, how can we be sure it’s not injecting bias into your process? In order to be considered AI, the technology would need to improve upon what the human recruiters could do. For example, if it could surface candidates that human recruiters would miss or enable automated, highly personalized messaging through high precision AI driven candidate screening. qualification.
To drill down a bit further, résumés are very noisy. There are hundreds of ways of saying the same thing. If you’re doing a keyword search for a blog writer, people who could do that job might describe themselves as “content marketing manager,” “marketing manager,” “copywriter,” etc. If you just look for the set of titles that you know about, you’re going to miss a lot of people that you should probably be talking to who have relevant experience.
At Uncommon, we built a model that understands all the complexity in job titles. If you give us one title, we’re able to expand that and bring back people that have relevant experience, thereby greatly expanding the pool of addressable talent. Experienced recruiters have that map in their head, and they can build complicated queries that get at those people. We are using technology to democratize to teams of all levels and capabilities. Anybody can gain best of industry-practices by getting ahold of our technology. We can provide this expert-level domain knowledge that only a subset of recruiters have today to all recruiters—and we can do that in a way that’s 100% consistent.
Another way AI can be used to transform recruiting is through messaging and outreach—and we’re just beginning to see this. Recruiters now understand that when messaging prospective candidates, it generally takes several touch points to build a relationship over time. If you have 100 candidates and you’re managing multiple sequences, it quickly becomes something that a single human can’t handle very easily, but it’s no problem for a machine. And when machines handle this task, we can easily track success and understand how factors like time of day, day of the week, intervals between emails, subject lines, etc. can impact effectiveness.
While some of this technology already exists in the marketing world, in recruiting there’s a much smaller margin of error. If you receive a marketing message that’s not right for you, you’ll probably just delete it and move on. But if it’s an email that’s coming from a specific recruiter, you’ll be much less forgiving if it’s clearly not a match. Companies are rightfully protective of their brand in the employment marketplace and are very mindful of both the time of hiring managers and job seekers.
Becoming an informed consumer of AI
If you’re evaluating potential AI solutions for your company, there are a few things to keep in mind. As I mentioned earlier, AI means that you’re approaching or surpassing human capability in some way. To do that well, it typically requires large volumes of data. This is why it’s important to understand where the data is coming from and what type of data they’re using.
I’d also urge you to be wary of any “black box” AI functionality. Some companies offer closed-loop systems that don’t give you any control over how they make decisions or rankings. You shouldn’t just be receiving output—you should have the ability to shape the outcome. One of the things we really believe here at Uncommon is that AI should be transparent and interpretable.
Finally, it’s important to understand that machine learning systems have a huge potential to propagate bias. If you’re training a model to rank candidates based on previous hiring success, you run a very real risk of propagating historical bias that you may be tasked to improve. That’s why at Uncommon we ingest data on people and companies about the world, but we never train our data on outcomes. We don’t analyze based on people who have been successful at your company before. We believe that previous outcomes are an important signal, but they shouldn’t be the only signal.
To be an informed consumer, don’t just accept scores and rankings. Try to understand how the technology is reaching these conclusions. And keep checking in to confirm that it’s making the same types of decisions you would while also bringing you insights you wouldn’t have had otherwise.