- Summary:
- New HR tools are cropping up in large numbers. Many are using Artificial Intelligence (AI) to parse data so that recruiting, HR and operations managers can make better hiring and promotion decisions. But, the way you implement these is different and so are the costs.
With the big HR show kicking off in Vegas this week, thousands of potential software buyers will hit the expo hall, keynotes and breakouts looking for new HR tech to solve their HR challenges. And, in today’s market, one of the biggest challenges will be finding piles of talent.
But getting (and retaining) talent will require new sources of talent, new methods of recruiting, new processes to speed up the recruiting process, and, better results than before. One type of solution that many CHRO’s and Recruiters will likely examine at the HR Technology Conference will involve AI-powered tools that use big data to improve talent sourcing and career advancement.
But readers shouldn’t assume that these new software products will or can be implemented like other HRMS solutions. No, the implementation will be quite different as will the costs.
Read on….
(The following counsel is intentionally brief and is meant to provide directional (not absolute) guidance.)
The new AI-Based HR solutions
These new tools rely on lots of data. Data that will come from your ATS, HRMS, Performance Management and other internal systems. They may also tap into new solutions (e.g., ModernHire’s Virtual Job Tryout), external databases, insights from video interviews, etc. While some of this data is structured, some of it is not.
To make sense out of all of this data, these tools require integrations to many existing systems and external databases (i.e., content) and to the new AI-based products/tools you might acquire. This takes integration technology and tools that understand the meaning of the data elements within each data source (i.e., context).
Some of these tools use completely new (to you) technologies. Where you know flat files and relational databases, the new tools use graph databases.
These tools are not transaction processing applications like old school apps such as payroll and financial software products. These tools need someone to make value judgments based on the connections they find. Also, the tool configuration at one company may have little or no value at a different firm as the base data may differ significantly, the business needs are different, etc.
And, if you have enough data to put into one of these new systems, you might get some interesting correlations to pop out. At that point, you’ll need to determine:
- What assumptions you and others in your firm have been making for years (e.g., our best new hires rowed crew at an Ivy League university) that might not be correct..
- What extra data you are collecting that is frustrating job seekers and adding no viable value.
- How you will change your talent sourcing strategies going forward..
- How you will get others in your firm to change their mindsets and accept different practices, ideas, candidates, etc.
- And more.
Implementing AI-based HR solutions
Many of us have experience implementing traditional systems like an HRMS, ERP, etc. Those projects often required people to spend time:
- Converting and mapping old data to fit the new system.
- Populating key systems tables.
- Cleaning up transaction and balance data from prior systems.
- Training users to operate the new solutions.
- Interviewing key personnel to configure processes and controls.
- Develop reports/queries/etc.
- Etc.
And the bigger the firm, more time has to be spent rationalizing data and processes across the firm’s far-flung empire. Software has to be configured so it would serve all plants, divisions, business models and users.
Those steps didn’t vary much regardless of the application being implemented. Yes, a traditional application (like payroll) has some process or function-specific needs but the basics around documenting requirements, converting data, etc. are fairly consistent and common. Well, they were until these new-generation tools came into being.
In conversations with new HR vendors (e.g., Eightfold.ai, Vettd.ai, retrain.ai, Modern Hire) and others, the implementation is now quite different and so may be the cost structure. The new/different steps include:
- A very different kind of discovery work – With these new HR tools, HR, operations, line managers and recruiters need to be extensively interviewed to uncover the factors people are using to evaluate jobseekers or the performance of existing jobseekers. If typical, you might find a substantial variety in identified factors and a lot of ambiguity, too. These factors need to be identified so that the AI/ML tools can determine if there is any correlation between these and the results the firm wants and expects.
- The existing data may be insufficient and require supplementing – The problem here is that the needed data may not currently exist (e.g., where is it documented why some people were not only promoted but did well in that new position, or, vice versa?). More information may need to be acquired via interviews, external databases, new systems, etc. Access to external databases may bring additional costs into the picture.
- The initial running of these tools may uncover some uncomfortable truths – Long-held “beliefs” may turn out to be someone’s unfounded opinion with no basis in fact. For example, one retail CEO believes that ALL hourly workers will become less productive sometime within three years after joining the firm and all of them should be terminated before then. Some firms will learn that a specific manager may be a great leader while another is guilty of running off lots of the company’s best talent. Right or wrong, the team must confront these.
- After a round of discoveries, expect to do it again – You may find out that you haven’t been using great techniques in choosing who to hire/promote. That knowledge may drive you to capture and review more/different data to gain better insights. To quote shampoo bottles, ‘rinse and repeat, as needed’.
- Expect to pay for this initial discovery work – Software firms and consultancies aren’t going to loan you their scarce, valuable information/data scientists for nothing. These initial assessment efforts can cost your firm $200,000+. Yes, some may charge only $100,000 but you might not get real value or insights from this. And, a rare firm might do it for free but watch out, they’ll get their compensation later in higher subscription or license fees.
- Unlike some software, this isn’t a configure and forget situation – Over time, some aspects of the underlying algorithms and data may need to change. Why? Changing business conditions, new business models, new employee expectations, inorganic growth, etc. can alter the goals of the tool. Also, some unintended biases may present themselves and need to be dealt with. So, where some old school software applications needed periodic patches and upgrades, these new solutions may need new data and adjustments to the underlying models.
- There may be significant change management challenges with this effort – If the initial discovery throws cold water on long-held management ideas and practices, you can expect pushback. Additionally, if the new findings trigger new, shorter hiring processes or make recommendations for promotions that surprise some managers, expect challenges. New insights trigger change and change makes people uncomfortable.
My take
For those HR executives examining new solutions, know that the cost structure for these tools is very different from traditional HR applications. Besides the upfront discovery effort, there’s the annual subscription and periodic tuning work that must also be budgeted. And, don’t forget that you might need to have one or more data scientists on your HR technical support team, too!
Smaller firms may remain priced out of this new space as the implementation costs may be unaffordable. But, small firms also have the disadvantage in that they lack enough data and experience points for statistically meaningful correlations to appear. The best candidates for these tools may be firms that have large workforces and years of historical data to look for correlations.
While these tools can be great in debunking long-held biases, they still are correlation not causation tools. In other words, just because one kind of employee has been shown to often advance quickly or stay with the company for a long time (correlation), that doesn’t mean that all employees with similar characteristics will exhibit the same behavior (causation).
This also means that unique people or people with dissimilar characteristics from the general population will not get a fair shake. I hired one of these exceptions. She was a liberal arts major when we were only hiring B-school and engineering graduates. She’s outlasted all of us, myself included. Great tools can spot correlations when the results of a large population of similar people are in the database. These tools can’t really do much with people who are unique. Use these tools correctly.
Image credit - Pixabay
0 Commentaires