6 Questions to Ask Vendors Selling AI-Powered HR Tech Solutions
Thanks to the appearance of ChatGPT/generative AI, being a CHRO now feels like fighting your way through Istanbul’s famous Grand Bazaar – swarms of eager application vendors pulling at your sleeve, promising you wondrous solutions that will change your company’s HR practices forever.
Top HRTech Insights: From Tactical to Strategic: HR Solutions That Drive Business Value
Here’s the tricky part: many of these vendors are shills, but some of them are right. Generative AI truly is a game-changer in HR Tech. The central challenge for CHROs is to figure out which generative-AI-powered solutions will truly transform their world and which are useless trend-surfing hypeware.
As someone developing generative-AI-powered HR Tech solutions, let me suggest six fundamental questions/challenges all CHROs must put to solution vendors as CHROs plot a new course for their organizations’ people strategies.
How will generative AI address my HR problems better than the solutions I use now?
- Ask vendor to clearly articulate the business problem AI is helping solve and how they have ensured that the problem is solved better with AI.
- Identify the promised outcome such as workflow efficiency, cost-reduction, new KPIs, and valuable new content such as targeted interview and assessment questions.
- Understand what kind of application this is – is it about coaching, analytics, workflow enhancement or new content creation. Each requires a different technique of bringing generative AI into the solution. Ensure vendor is able to clearly define their approach.
How has generative AI been fine-tuned for this specific application?
- Generative AI must be trained for specific applications. Vendor must be able to clearly define the specific training model used for the solution.
- Generative AI can deliver inconsistent and false results. Validation is crucial.
- Get a walk through of the validation process.
- Build your own confidence by looking into the vendor’s bench of subject matter experts used to measure the accuracy and applicability of the solution’s output.
- Many companies are offering generative-AI-based interview and assessment questions in the hiring process. These are high risk applications. Ask for clarity on how they are vetted.
How does generative AI in the solution result in better outcomes?
- Despite the shiny-object trendiness of having generative AI under the hood, spend the time to understand how this solution materially improves KPI metrics such as better hiring outcomes, workflow satisfaction.
- Check whether the AI in this solution is pre-curated or real time, and why the vendor chose the approach used. Given the risky nature of generative AI’s potential for hallucinations, pre-curated, validated content is much safer than real-time AI summoning output from heaven knows where.
- Ask for the process the prompt engineers followed to ensure the solution’s output is accurate.
- Ensure vendor is clearly defining how the various decision choices are offered in any talent process. In my view, AI-powered HR Tech solutions should never make decisions. This is not just an ethics issue. AI may be biased and requires human oversight. Good AI-powered solutions analyze the situation and offer human users the widest variety of problem-solving options. Humans must make the final decision.
Read More: SmartRecruiters Appoints Rebecca Carr as Chief Product Officer
The more AI figures in such important decisions as hiring or placing employees, the more important it is to eliminate bias. One example is what happened to early AI solutions trained to boost executive diversity: The AI “learned” the proper profile for executives featured an Ivy League education. Why? Because that’s who historically got executive positions, and the data “learned” from history. The AI needed to be adjusted to become equitable.
- Ask the vendor to prove the solution includes bias reduction.
- Get vendor specifications and documentation (AI explainability) on which models of accuracy were used in fine-tuning the solution and how any adverse impact was checked for.
- Confirm that given the same inputs and asked the same questions, the solution consistently delivers the same or nearly the same outputs and answers.
- Confirm the solution is trained in compliance with all existing anti-discrimination and other applicable regulations.
Who owns the risk if something goes wrong?
Laws regarding legal liability for actionable mistakes differ across U.S. states and across countries on a global basis. Customers must be comfortable with a vendor’s position on legal liability. Some vendors may shoulder all the risk, while others try to push the risk onto customers.
- Get the explanation on vendor’s position on legal risk, anywhere in the world.
What is the plan to keep this solution fresh as AI capabilities evolve?
As we’ve seen in the last year, AI evolves at a dizzying pace.
- Understand the plan to ensure this solution will remain state of the art.
- Ensure there is a clear commitment to AI related roadmap for the next several years.
- Get evidence of investment commitment to stay abreast of AI’s evolution.
Normally, organizations embrace change at their comfort levels. But, the advent of generative AI is not a normal event. CHROs don’t have the luxury of adopting at their accustomed pace. We must all be innovators/earliest adopters now.
As someone developing AI-powered HR Tech solutions, I put myself in the customers’ place and test my team by posing these six fundamental challenges/questions. I’m sharing them here to help guide CHROs in some of the most important buying decisions they’ll ever make.