That’s Not the AI You Think You’re Looking For | That’s Not the AI We Think We’re Looking For
Unless, of course, it’s written on PowerPoint. Then: it’s the AI the vendor says you’re looking for.
I’m stealing a bit of that. Well, a lot of that. The first bit I’m thinking you might recognize from a small, really inconsequential bit of fantasy that came out in 1977. The second bit I stole from Microsoft’s Mat Velloso. As he put it:
There’s oodles of truth to that. Oodles. Also, I don’t know what oodles means. But I think it’s like googles. Which means it should be a software company. Maybe it already is.
Here’s the thing: we haven’t invented true AI yet. And, by “we”, I don’t just mean the recruitment tech industry. I mean all of us. Everybody. It ain’t happened yet, and maybe it never will. Humans haven’t even figured our natural intelligence. How in the world are we going to make an artificial variant?
What you’re typically being offered, dear friends, is some level of process automation, that is powered by some pieces of AI technology. Which sounds way less sexy on a PowerPoint/ demo. Marketers, being, well, marketers, are naturally averse to unsexy.
They want to tell you that their CRM/ RMP/ sourcing silver bullet/ chatbotatscrmwamalamdoodle is magical. Saying it’s really just “process automation that uses a bit of machine learning and natural language processing to get the basic job done”? Nah – that’s both accurate and terrible! Better to say it’s an “AI powered recruiting chatbot”. Ooooh, ahhh…. so exciting. So, so sellable. So much better on a PowerPoint.
But… you need to buy tools. Not magic.
AI is basically a bucket. One that holds a variety of interesting tools, each of which can be used in equally interesting ways to solve problems. Sometimes you combine them. Sometimes they’re stand-alone. The stage we’re at right now is called “Narrow AI”. Basically, it barely works as a whole, but the components do (ie, they’re narrowly focused). Once we figure out how to get them to sing in harmony we get to the next level. The point where the AI can pass a Turing Test & prove that it’s sentient. We call that Artificial General Intelligence (or AGI, or “holy crap my computer just told me to eff off!”). The next level is Superintelligence. Or, in some cultures “SkyNet”.
For now, the tools that you’re getting told are “AI” are generally making use of some combination of (or, the entirety of) this:
They may not have all of that, but it’s pretty typical. For example, that “AI chatbot”? It’s running along a schema. Some data comes in, the chatbot turns that query into data. It thinks about the questions it’s getting, translates the answers back into human, and hopefully says something reasonable. Along the way, it learns from mistakes, remembers patterns, and gets stronger. It’s still not “intelligent” – simply mimicking intelligence. It’s using parts of AI to automate a process that’s redundant and easier for a machine to do. But it’s not thinking about it. Not really.
It's Okay to be Cynical
I know, I know – we live in an age of exaggeration, and alternate facts. That said: you’re buying for your business, and you’ll be judged not on how well you buy what marketing’s shoveling at you, but on how well you can see past their, umm, bovine excreta. It’s not hard, really.
The first thing to think about is what problems you’re trying to solve for. From there, you can start to work through which of those problems seem the banalest. Think scheduling. Volume sourcing. Publishing jobs to Craigslist. High-level resume screening. Heck, you can even think about automating some of your job description creation. You’re essentially looking at the tasks that take up the most time but aren’t core to the job of recruitment. Work that junior-level coordinators tend to get to do. Start with those.
There are tools for that. Seriously – and, they tend to involve different AI technologies. Semantic search. Rules-based scheduling that adapts as it learns your business (ie, machine learning). They’re awesomely useful – just… not AI. Per se. They use pieces of what will – if we ever get out of our own way – will eventually come together to form our first AI.*
Humans Still Matter
Here’s the thing: process automation tools aren’t going to do away with recruiters. Similarly to the way SalesForce and Marketo didn’t destroy sales and marketing, the tools that are coming online (and they are coming online fast – personally I’m hoping we’ve achieved peek chatbot), will enable us to do what we do so well. Attract, recruit, and hire.
Investing in automation makes sense from that perspective. If it incorporates AI tech to do the job, and that makes the tool more efficient? Cool. That’s a great thing. But the main goal is a business one: free up your staff to do work that matters and gain efficiencies in terms of scale and cost. A nice result? In times of hyper-growth, there’s no need to add on a slew of extra resources to handle scheduling, building pipelines, and other mundane tasks – you just turn to your “AI” powered thingawhmutsit and adjust the volume to eleven for the duration. Headcount doesn’t spike – it’s SAAS baby.
…you need automation. It makes sense to turn to it. Workflow automation, pipelining, scheduling – there’s a solid case to make to your internal stakeholders (ie: Finance) as to why you should go there. Just be careful of the buzzwords, and don’t be afraid of AI. Because it ain’t the AI we’re looking for. But process automation? Yeah. That’ll do.