How We Use AI Sourcing To Amplify Hiring

When it comes to recruitment, there’s a lot of data that simply doesn’t show up on a candidate’s profile. It’s not a question of whether you can parse the data, or how well you can process the data. If you look at the skills your colleagues have on their profiles, you’ll see that many of them don’t list a lot of the skills you know they have.

This is why we decided to approach the problem from a completely different direction.

An Amazonian Lesson In Algorithmic Approaches

AI Sourcing has its merits, but it’s the human touch that seems to produce true talent advantage.

In 2014, Amazon attempted to mechanise the search for top talent. It trained computer models to vet candidates by observing patterns in resumes submitted to the company over a 10-year period. Automation has been key to Amazon’s success within ecommerce; this wasn’t their first rodeo.

This approach produced a star-ranking system, similar to ecommerce, but perhaps lacking the ‘wisdom of the crowds’. Candidates were assigned a score from one to five stars, reduced to a single value.

Make no mistake about it: Computers process numbers – not symbols. We measure our understanding (and control) by the extent to which we can arithmetize an activity.

Epigrams on Programming, Alan J. Perlis

Of course, people aren’t that one-dimensional. Figuring that out early on, the team at attract.ai decided to take a different approach. We realised that candidate fit is more likely to be found, by looking at their environment, and other indirect factors.

A Unique Perspective

Rather than being a hard data play, attract.ai focused on letting humans do what they are great at, and leveraging computers for what they do well.

Our first iteration was a hybrid model of AI sourcing and a human-in-the-loop strategy. It allowed users to gain a unique slice of candidates in any given market. From a talent acquisition standpoint, this means we provide an advantage over everybody else who are using the same LinkedIn search box and getting pretty much identical results.

We found that on traditional platforms, algorithms often bring up candidate profiles that have been optimised for discovery, resulting in the same profiles being shown, in a similar order. You can imagine then, how often those candidates near the top of the results get messaged, and conversely, how many candidates are quietly sitting undiscovered.

We search for the profiles that are hidden gem candidates that are more likely to respond positively, because they are not optimised for discovery and thus aren’t getting messaged regularly. Many great candidates don’t have good search presence because they are too busy excelling in their day job and do not fit within one-dimensional ranking systems. We took the lessons from our original hybrid model and began to apply them to our long term goal: to build an AI sourcing platform which uncovers the candidates who are being overlooked by everybody else.

Optimising, Not Outsourcing, The Hiring Process

In the first iteration of attract.ai we used AI Sourcing as a tool to power our hybrid model. This allowed us to cut through the market in ways others couldn’t and refine our algorithms. Over time, the tool sped up our own sourcing process. Transforming it from a tedious, manual process into a quick & relatively painless system. We have built thousands of hours of human nuanced sourcing experience into our systems, cutting through the buzzwords, and getting to the heart of whether a profile is worth reaching out to or not.

We’re not done re-writing the AI Sourcing playbook just yet. Now that our sourcing tool is available to the public, we’re able to incorporate the nuances of more companies and users, making our AI Sourcing solution even better at finding the candidates you should be talking to, not the candidates with the loudest voice.