Meet Amy.
Amy is the future of human resources.
I met Amy because she works for Abby.
I met Abby at an HRUndergroundX meetup hosted by Nick Larche. She told me about a new technology she was using to recruit candidates for her start-up clients.
To find out more, I sent her this email:
“Abby, it was nice to meet you the other night. I’d love to get together and have coffee.”
Abby responded,
“… Amy CC’d has my schedule and will reach out to coordinate with you. She’ll also send you a calendar invite once confirmed. Amy, please block this as a 30 minute coffee meeting. If Crema is not a good location for Greg, please let him select our meeting spot.
Thanks!
Abby M…”
Twenty-one minutes later, I received an email from Amy Ingram:
“Hi Greg,
Happy to get something on Abby’s calendar.
Does Monday, Jul 18 at 4:00 PM work? Alternatively, Abby is available Tuesday, Jul 19 at 11:00 AM or 4:00 PM.
Abby likes Crema Coffee House, 2862 Larimer St, Denver, CO 80205, USA, for coffee.
Amy
Amy Ingram | Personal Assistant to Abby M.”
I responded by saying:
“Amy,
Unfortunately those times don’t work for me. I am open at 11am on July 22nd.
Let me know if that would work.”
Twenty-eight minutes later, I got a response from Amy:
“Hi Greg,
Abby is available Friday, Jul 22 at 11:00 AM.
I’ll send out an invite.
Amy”
I received and accepted the invite. The day before my meeting with Abby, I sent a confirmation email:
“Amy,
Just want to confirm that Abby is still available for our appointment tomorrow at 11AM?”
To which, Amy replied:
“Hi Greg,
This meeting is scheduled for Friday, Jul 22 at 11:00 AM and will be at Crema Coffee House, 2862 Larimer St, Denver, CO 80205, USA. I haven’t heard anything otherwise from Abby.
Amy”
From this exchange, you’d conclude Amy is the model personal assistant. She provides timely, accurate responses, blended with a human touch.
And you would be correct, except for the part about being human.
Amy Ingram, as her initials imply, is not a person.
She‘s AI.
Artificial Intelligence.
She‘s a virtual personal assistant created by a company called x.ai.
As of this writing, these assistants are in beta, but there is a waiting list to sign up for them.
Amy is an example of how close AI is to replicating human interaction.
Amy takes the place of a recruiting coordinator scheduling candidates for interviews and managing Abby’s calendar.
When I heard about Amy, I started thinking:
“What other HR functions could AI replace?”
What does AI look like?
Hal-9000, The Terminator, The Matrix.
Hollywood has shown us the ways AI can go bad.
It makes for great cinema: Human beings create an intelligent machine that turns on its human creators who have to fight for survival.
When we think of AI, the idea of the computer achieving consciousness is what we envision.
Stephen Hawking, Elon Musk, Bill Gates and other scientists and technologists aren’t just debating when this consciousness will happen, but whether or not it means the end of humanity.
I am not going to get into what will happen when AI achieves this consciousness. Way smarter people have written way more about this already.
If you are interested in the future of AI and the ethical conversations about it, read this article at Wait But Why.
Set aside some time and mind space because it goes deep and gives a comprehensive summary of AI and its implications for the future.
It will blow your mind.
When I started contemplating AI and its impact on the future of human resources, I was thinking about Artificial Narrow Intelligence (ANI).
This is the lowest caliber of AI and is created for one specific purpose.
Instead of being the scary kind of AI that could possibly turn on its human creators, this kind of AI is available today in your phone, your Amazon account, and your Netflix recommendations.
These are the programs that determine what you like based on what you have read or watched before.
In fact, when I searched “AI in HR” to find out what was already being done in this field, most of the articles I read were about algorithms that could identify talented candidates or find trends in turnover to help mitigate it.
While these programs are useful, I had something else in mind for AI in HR.
When we ask Siri to help us find something online or get recommendations for our next purchases from Amazon, it feels like we are talking to a machine.
The interface doesn’t feel human.
When you communicate with an AI entity doing work previously reserved for a person, it feel less like an interface and more like an interaction.
It feels human.
Like Amy.
And as AI becomes more human, my unscientific estimate is it will take over 80% of the work that HR professionals are doing today.
Should HR Fear AI?
When I think about HR, I break it into two parts:
- Strategic ideas that optimize the people in the organization, and
- Administration and compliance to reduce risk in the organization.
The first part is where HR professionals wish they could spend their time.
The second part is where HR professionals actually spend their time.
This second part, which represents 80% of HR tasks, is where AI is going to take over.
To demonstrate, I picked four HR functions which could be replaced using AI technology similar to the kind which makes Amy such a great assistant.
1. Recruiting
Amy showed what can be done today for recruiting support.
This is just the start for AI in recruiting.
In the future, a hiring manager can send the following email to Phil, the AI recruiter:
“Phil, I need to open a req for a new business analyst.”
Phil answers almost immediately:
“Of course, let’s start by finding out what you are looking for. Could you tell me what skills or experience this person would need to have?”
The exchange goes on between Phil and the hiring manager until Phil has the information needed to build a candidate profile.
Once Phil has the profile, he begins searching all online sources available to find someone who most closely matches this profile. These sources could be databases or online social media profiles.
In a short period of time, he’s generated a list of candidates with resumes and profiles. He sends them to the hiring manager for review.
Each time the hiring manager responds to the candidates he received, Phil uses the feedback to refine the search and find candidates that more closely match the hiring manager’s perception of the ideal candidate.
At the same time Phil is looking for candidates, he is contacting the ones the hiring manager wants to talk to and finding out which ones would be interested in the position.
He does this through the same mechanism he used to build the profile, which is an email exchange with each candidate.
Phil works 24 hours a day, 7 days a week to match the candidates with the role. Like a non-threatening Terminator, he doesn’t stop until his mission is complete.
And Phil can do this for all hiring managers in the organization at the same time.
Why wouldn’t hiring managers just enter the candidate profile in an ATS system and let the system search databases to find candidates?
If this was the solution, one of the myriad ATS vendors would have figured it out by now and taken over the industry.
The issue isn’t the systems or the databases. It’s that hiring managers don’t go through the hiring process often enough to remember how to login to the ATS system, let alone create a new requisition.
It’s the interface via a channel they use every day, in this case email, which makes the interaction between Phil and the hiring manager work.
It’s the Socratic method of inquiry an AI can replicate that leads to actionable information.
The same back and forth recruiters participate in with their hiring managers can be done by Phil.
Right now, recruiters are reading this and saying, “There’s more to recruiting than that,” and I would agree.
If the hiring manager and Phil can’t get to a point where they agree on the candidate profile or the hiring manager is frustrated because his needs aren’t being met, the recruiting mentor/coach/leader can step in and help sort it out.
If there are candidates that need to talk to someone on the phone, the recruiting mentor/coach/leader can get on the line and give them the voice of a person on the other end.
At the same time, recruiters need to be honest about how much of their time is spent finding candidates and getting basic details from hiring managers and then rejoice in the prospect of having to do less of both of these things.
Instead of being a recruiter that does all of these repetitive processes, recruiters become talent coaches who spend time with hiring managers talking about who to hire and how to keep the best talent in the organization.
2. Compensation
Compensation analysts in most organizations get this request daily:
“Can you look at Sarah’s salary and tell me if we are paying her in the range?”
In a technologically-enabled, open organization, the manager that sent this email could login to a compensation system and find a graphic depiction of a range with Sarah’s placement in it.
I bet more than 97% of the people reading this don’t work in such an organization.
Even if you had the system to enable this, your manager can’t remember how to get into the system because she asks this question only a few times per year.
If she can remember how to login, she can’t remember how to get the data.
Compensation data is not her job.
She runs sales, or operations, or IT, or finance, and can’t be expected to be proficient in the internal HR systems as well.
What happens next?
The comp analyst looks Sarah up in the HRIS system, finds her current salary, and calculates her position in the range.
He may even pull in some external survey data to show where Sarah would be in the “market.”
If the comp analyst is a high performer, he will do an internal analysis of where Sarah is in relation to other people in similar departments in the company.
With experience, he learns about what certain managers like, what questions they will ask next, and tries to provide the answers before the manager even asks the question.
Now imagine that our comp analyst is named Buck.
Buck is like Amy or Phil and is an AI assistant that takes the original question (“Can you look at Sarah’s salary and tell me where we are paying her in the range?”) and runs through the steps outlined above.
The difference is Buck can do all of this in seconds.
Also, Buck can do this for 10, 20, 100 managers at the same time.
Buck sends the information, in an email, to the managers.
Just like the high-performing, human comp analyst, Buck learns what questions managers will ask next or what they might do with the information.
After providing all the salary detail, Buck asks,
“Are you thinking about giving Sarah an increase or promoting her to a new role? Let me know and I can give you information for either scenario.”
When Buck does all of this, what does the compensation team do?
The compensation analysts/coaches/leaders work across the organization to ensure managers are making the best decisions about pay practices.
They employ data science and and behavioral psychology skills to develop incentive programs which will drive business results.
The repetitive tasks in the compensation department get done by Buck and the people in the compensation department get to do the deeper, higher level compensation work they have always aspired to do.
3. Onboarding and Knowledge Transfer
When someone starts with your company, what does the first day look like?
At the very least, the person has some sort of laptop or computer that enables them to connect to the company intranet so they can begin learning about the organization.
What do they do next? They likely spend hours on eLearning modules or other slides, decks, or exasperating documents reading about why the company exists, how the company works, and what is expected.
This doesn’t include any sort of compliance training that is required. Usually people are left on their own with a checklist of materials to review.
Companies like Tasytt are trying to change this.
They have created Obie which is like having a personal onboarding coach to help you find all of the information you need to get up to speed on the new organization.
According to Tasytt, “Obie offers a familiar, conversational user-experience you’ll actually enjoy. He can answer questions and send bite-sized knowledge to the team.
Obie is a quick-study — the more you use him, the more he delivers relevant and accurate content.”
Where an HR generalist or onboarding specialist had to provide the documents and check in on the new hire, Obie will become the coach who makes sure everyone has the answers they need to get up to speed quickly.
4. Benefits
The world of employee benefits mirrors the world of HR — strategic and administrative.
The strategic part involves looking at plan design and claims trends to determine how to build a benefits offering that is competitive, yet manages the financial risk of the organization.
The administrative part happens if an employee asks a question about when their benefits are effective, or if they have a problem signing up for their benefits, or if a claim isn’t processed as expected, or when their COBRA coverage begins.
Most of these administrative issues are resolved by applying a set of policies.
Experienced benefits specialists have accumulated the knowledge to answer these questions and just about every other question that comes up. They answer these questions over and over.
This knowledge takes time to build and is difficult to replace when one of your benefits specialists is promoted or leaves the company.
Imagine something similar to Obie, but with a knowledge base in benefits.
This AI benefits specialist, let’s call him Benny, will answer employees’ questions about any benefits in the organization.
As Benny answers questions, he’ll get better at it.
At first, he’ll need help from the people who are in the benefits specialist roles, but as time passes and he learns how to apply the policies, he’ll need less help from his human counterparts.
What happens to the benefits specialists?
They can stop spending their time answering the same question over and over.
They can identify the root causes of the problems that lead to the questions in the first place.
They can help educate employees about how to pick the right medical plans.
They can help explain what increasing the deferral percentage in an employee’s 401k would mean for future retirement savings.
They could do all the things they talk about doing, but never quite get around to because they are answering the questions today that Benny will answer in the future.
Are these the only roles AI can fill in HR?
If you are an HR professional who specializes in employee relations, you might think AI has no place in your world.
ELLIE is an AI psychologists being developed by the U.S Military to diagnose psychological problems in soldiers.
She can ask questions and visually scan the faces of the soldiers as they respond to pick up on expressions that help her evaluate the truthfulness of the responses.
If an AI psychologist can be used to talk about the war experiences of a soldier, the technology can certainly be used to talk to employees about issues they are having with their co-workers or managers and determine whether a human investigation is needed or not.
Why will AI succeed where other technology has failed?
Each one of the roles I described uses technology today which was sold on the promise it would free up resources or require fewer people to deliver information to the people who need it.
The sales pitch was technology would enable managers to open their own requisitions, get their own comp data, and provide onboarding and benefit information in ways employees could easily understand.
This promise hasn’t been realized.
The reason is the interface isn’t human enough. Human managers need a human link between themselves and the data in the systems.
Today, that human link is the recruiter, comp analyst, HR coordinator, or benefits specialist.
Tomorrow, AI will become this link. Even though it is not human, it will feel human enough to bridge the gap between the manager and the data.
What does this mean for the people in these HR jobs?
If you are reading this and you currently have one or more of these roles in your organization, I understand your concern.
I expect your first reaction will be to disagree and say, “there are things that machines cannot do.”
You are thinking there is something special about the humanness you deliver to your business partners and you can’t be replaced by a machine.
Of course there are times when you pick up the phone and explain something to an employee or to a manager, but if you’re being honest with yourself, most of your day happens through email exchanges that could easily be managed by an AI program that feels no less human than your response.
My advice would be to embrace this change and start learning how to use your newfound freedom from mundane tasks to coach the managers in your organization how to be more effective leaders.
In their book, Only Humans Need Apply: Winners and Losers in the Age of Smart Machines, Tom Davenport and Julia Kirby explain 5 things employees can do in response to the automation of their jobs. These action “steps” have been nicely summarized in this article by Bernard Marr:
“Stepping Aside means leaving the machines to do what they do best, and picking a career requiring skills such as creativity or empathy.
Stepping Narrowly means developing a specialty, in a field where there is little demand or no business case for implementing automation (a local tour guide, or a wine expert specializing in a particular region, being possibilities here).
Stepping Up means taking oversight of and responsibility for the work carried out by computers and AI — essentially becoming their boss, and considering the big picture strategy of implementing technology across an organization.
Stepping In means to become involved with the work being carried out by machines, to fine-tune and provide human oversight in areas where it is still needed. Real world examples here could be an accountant trained to spot errors caused by an automated system, or an ad buyer who can spot when a brand could be damaged by a particular placement, for reasons a robot might not comprehend.
Lastly, Stepping Forward is to work on developing the next generation of robotic and AI-driven technology. Robots can solve problems for us, but we still need to tell them what problems need solving. It still takes a human to understand that automation will be of benefit to a particular area of business, and a human to put together a strategy for automating that section.”
Remember how I said there are two parts to HR?
The good news is that AI is going to do the second, so that HR pros can spend their time on the first.
Thinking in these terms, and along the lines of the 5 “steps” in Davenport and Kirby’s book, should give HR professionals hope, rather than fear, in the face of the impending AI automation for HR.
What should the HR teams of today do to prepare?
Two things will make HR pros successful in the future: speaking the language of data and focusing on the “human” in human resources.
Recruiters should hone their coaching skills.
Learn how to spot and explain hiring biases to mangers.
Understand how candidates become high-performing employees and how that understanding can help managers craft candidate profiles and select the best mix of skills and cultural fit.
Compensation teams and benefits specialists should focus on data science and behavioral psychology.
I’m not suggesting everyone in HR needs a Ph.D in data science. There are free or cheap resources available to get started. Start at DataScienceCentral and explore the tutorials and resources.
You can download two of the more common data science languages, R and Python, for free and find tons of free materials to get started.
Check out DataCamp.
Use these skills to find trends and opportunities within your compensation structure or your claims trends.
Learn about behavioral psychology, which is the study of why people behave the way they do, by reading about it.
Start with Freakonomics, then read anything by Dan and Chip Heath, then read Nudge, and, if you are really ready for something deep, read Thinking Fast and Slow.
As you learn more about human behavior, you’ll realize most of the compensation and incentive programs in use today run contrary to what motivates human behavior.
Use this knowledge to design more effective compensation programs or to identify which benefit programs add the most value to the employee and the organization.
The AI End Game
Photo by Dino Reichmuth — Courtesy of Unsplash
Think about all of those times you said, “We could improve x by doing y if only we didn’t have all these administrative tasks to complete.”
Figure out what it would look like if you had time to work on all those y’s to deliver the value of x to your organization.
When the non-human machines take on all of the administrative HR tasks, those of us in HR can finally focus on the ideas that will make our organizations, well, human.
Unlike the Terminator or The Matrix, AI in the future won’t be eliminating or enslaving HR.