Latest Posts

How Automation Makes Bad Decisions

Man working on laptop at a table

There’s a lot of fear regarding implementing automation, artificial intelligence, or machine learning. 

What if it makes a wrong decision? 

Can I trust it to work well? 

Is it going to replace me?

How can I stay in control?

We see stories about Siri searching things incorrectly, facial recognition software with dangerous bias, and autopilot cars in accidents. It’s natural to fear the failures of technology. But instead of running from the tools, what if we focused on making them better? When we implement new tools, we need to confront our fears and concerns head-on.

How and Why Automation Makes Bad Decisions

To fix things, we need to know why they happen. When automation and artificial intelligence doesn’t work, what or who is at fault?

Bad Data

Bad data encompasses a lot of problems when it comes to the efficacy of automation. Unstructured data is a particular culprit. We talked recently about the effect of unstructured data, but let’s recap. When applications intake unstructured data (i.e., a fill-in-the-blank box) rather than structured data (i.e., a checkbox or drop-down selection), your database is left with information that it can’t necessarily parse correctly.

Bad data can also include problems like misspelled words. Misspellings happen, but if your database can’t be contextually searched, you’ll have difficulty surfacing those profiles again. Outdated data can also pose a significant problem to automation making a less-than-stellar decision. Do you know the last time your team updated their candidate data?

What about missing data in a profile? Does everyone has a resume or work experience attached?

We process a lot of data here at HiringSolved, and we recently found that:

Image with colorful graphics that shares: "51% of candidate records are missing a resume, 37% of candidate records are missing any education data, 1 in 4 records are missing work experience, 5% of records are missing any kind of contact information."

Incomplete records affect automation abilities and the ability to provide a quality candidate experience and provide all the needed information to hiring managers. 

In short, bad data, whether that’s unstructured, outdated, or just plain missing, is the number one reason why automation, AI, or machine learning won’t work well.

Inflated Expectations

The Gartner Hype Cycle of 2021 recently highlighted that AI in Talent Acquisition is currently in the Peak of Inflated Expectations stage. While AI can be an incredible addition to a recruiting workflow, “Users must set realistic expectations on what functionality these tools provide. AI should be used to support recruiters through enhanced automation of recruitment operations and candidate outreach to increase engagement and productivity, and to assist human decision making.”

While the abilities of AI and automation in recruitment may feel endless, the reality is that we’re a far way from C3PO telling us the odds.

GIF of C3PO behind Han Solo with Han Solo saying, "Never tell me the odds!"

Undedicated Change Management

Whenever a new tool is introduced, the team needs to accept the change and begin using it. Dedicated change management can help to make that process easier.

Ask yourself:

  • What will our recruiters get out of this change?
  • What resources do they need to adjust to this tool successfully?

Once you align expectations to the product’s abilities and the strategy for implementation, your team will be unstoppable!

The Fixes You Need to Make Automation Work

While the above items can cause roadblocks to a successful implementation of automation, they aren’t a dealbreaker! Here’s how to tackle them head-on.

Data Normalization

Normalized data is the foundation of automation, AI, and machine learning, but what is data normalization?

Click to learn more about how data normalization supports automation

Since data enables or denies the capabilities of tools like automation, normalizing your existing data and creating a new strategy to only bring in structured and normalized data in the future is the perfect strategy for automation success.

Audit your existing systems with the help of an experienced vendor to find out what your databases are currently facing. Once you know where you’re at, build a strategy to ensure that only the best, most up-to-date, clean data enters your systems.

Reality Checks

Be realistic about what AI and automation can and can’t do before letting your expectations run wild. It’s essential to work closely with your vendor to understand the real-life abilities of a product and build plans and goals accordingly. 

It’s also crucial to relay that information to your team. When implementing automation or AI, recruiting teams can feel concerned about being phased out. In reality, recruiters should use automation and AI to optimize operations and streamline tasks rather than replace the human hands in control.

Consultative Partnerships

When you’re working with your vendor to find the limits of the technology, work with them to set goals together. An experienced vendor will look forward to building a consultative partnership with you to ensure that you achieve the success you’ve imagined.

What KPIs are yourbeing held to? How does automation serve those goals? It’s essential to have a strategy when implementing a new tool and to understand how that tool directly impacts your goals is paramount.

As our EVP, Dave Barthel said in a recent post on LinkedIn,

Blue and white image that reads: "When relationships are built effectively, and trust is grown, they will transform into mutually beneficial consultative partnerships; the customer will trust you to serve their best interests, and you can trust them to help you guide your product to be the best it can be.”

Consultative partnerships should be equally beneficial for both you and the vendor, so don’t be afraid to set that standard for the relationship.

Ultimately, automation needs quality data, relevant expectations, and a solid partnership to bring the results that will transform your recruiting. How are you going to ensure those exist in your processes?