In a report, Deloitte presents the evolvement of HR technology in four stages. The first stage describes the period during the 1970s and 1980s, when the main attention of software vendors was on systems that help HR managers make records. During the second stage between the 1990s and early 2000s, capabilities to support recruiting, training and performance controlling were developed. Around 2010, at the third stage, vendors started to offer cloud services and more user-friendly systems to engage with employees’ self-services.

The Deloitte report claims that today we are in the fourth stage of HR technology. In order to react to work environment microtrends, vendors have to design tools that are targeted for teams, individuals and networks and also enhance people’s productivity. At JANZZ.technology, we too think that an HR software should assist HR managers in being more productive and in focusing more on value-added tasks. For example, it should reduce the effort needed for mundane tasks such as screening thousands of candidates and should enable to focus more on value-added tasks such as interviewing candidates. We further believe that this is going to be achieved by means of Artificial Intelligence (AI).

AI for recruiting

Talent acquisition is undeniably one of the most important parts of corporation management. This makes the recruitment software market the most competitive and interesting market to observe. Governmental organizations like Public Employment Services (PES) are also actively seeking solutions to deal with this matter. According to Crunchbase, solely in 2018 recruiting software startups received over $ 600 million of VC finance. In the latest Market Map by HR Tech China, recruiting software vendors constitute the largest part of all HR technology vendors.

The offers made by vendors include the testing and assessment of candidates, background checks, video interviewing and many recruitment platforms. Since early 2017, AI has emerged in the recruiting process. The ensuing rise of recruiting software’s so-called AI capabilities could make any HR manager feel overwhelmed.

‘AI for recruiting’ is an emerging technology used in HR recruitment processes. It employs AI technology and mainly aims to reduce repetitive, time-consuming and banal tasks, which helps recruiters and hiring managers focus on value-added activities. 52% of talent acquisition leaders state that the hardest part of recruitment is the screening of candidates from a large applicant pool. [1] For instance, an AI recruitment software can screen thousands of applications and recommend the top 5 candidates in a blink. Thus, HR managers using AI for recruiting will have more resources to assess the top picks in-depth, which in turn heightens their chance to really find the best suitable candidate.

By contrast, the larger public uses the word AI vaguely and often even inaccurately. Many companies use ‘AI’ to describe their products in order to make them appear as ‘upgraded to the next level’. In most cases, this sort of product advertising promises too much. It is therefore extremely important to be able to evaluate both such products and their vendors. The process is similar to that of hiring: only if you have various ways to assess options, you can find the best one possible.

Evaluating AI for recruiting technology

How should HR avoid over-promised software when choosing from all the products? We have come up with three principles to go by when choosing AI recruitment software:

Principle I: Be aware of the bias within AI recruiting software

Last year, the story about the Amazon AI recruiting tool secretly being biased against women was a wake-up call for all of us: machine learning can be just as biased as human beings. Therefore, it is extremely important to focus on algorithmic fairness and transparency.

You must be aware of how the software processes personal information data such as date of birth, gender and nationality. Which factors does the software take into consideration when matching? And does the software ignore irrelevant factors?

Apart from algorithms you have to make sure that the software has representative training data. In the case of the Amazon AI recruiting tool, the tool turned against female applicants because, for over ten years, the company trained its computer models with resumes summited by male applicants. Hence, make sure you ask your software vendor how they deal with their data source.

Principle II: Be sure to test before you purchase

Before buying a car, you would certainly test drive it first. This rule applies to buying recruiting software as well. Since a recruitment software system is an expensive and long-term investment for your business, it is wise to complete a POC (proof of concept).  In this trial run you’ll find out if the software can really solve your prioritized problems, perform the promised functions and handle your data within the required scale and scope.

JANZZ.technology conducted a POC with an intergovernmental organization to find out if our solution can truly help them save time when finding suitable candidates. Our AI software competed against their HR team in matching candidates worldwide to their open junior intern positions. After screening over thousands of applications, they were pleasantly surprised by our results.

Most good software vendors offer free trials. It is important to prepare your test data well in order to get the most out of your practice and maximally optimize this process. One of the often-ignored aspects when choosing a recruitment software is the maintenance and support needed after the purchase. Only with constant updating, the software is able to develop in parallel with the fast-changing markets, customer requirements and ongoing digitalization. Don’t simply trust marketing lines such as “50 other leading companies from your industry are use our software” or “top 100 out of the top 500 companies are also using our software”.

Principle III: Taking the “soft skills” of an AI recruiting software into consideration

Just like assessing a job candidate by evaluating his/her skills and soft skills, an AI recruiting software has its skills and soft skills as well. We determine the hard skills of a software as functionality, accuracy, data security, speed, languages capability and similar highly valuated competences.

However, many might overlook the soft skills of an AI recruiting software. Soft skills of an AI recruiting software are the ability to deeply understand the language, education, working and social system etc. of your practice region, and the ability to localize certain regions or countries.

For instance, a country like Spain has more than one language: Castilian(Spanish), Catalan, Valenciana, Galician and Basque. A good AI recruiting software should be able to understand the different languages and know the common terms that are used in the four different languages to refer to exactly the same thing.

Conducting job matching across Europe is not an easy task, for instance because each of the 44 countries has its own education system (even with the bologna framework). It is an enormous amount of work to compare the different education levels and match people to jobs. Does your vendor have the right knowledge to solve such problems within your practice region?

Besides language and education, there are many more and equally important categories that need to be taken seriously.  As an international corporation operating in different countries, you want to have a software which understands all your markets.

Limitations of AI 

You may have heard of all the promised benefits of using an AI recruiting software. Before the AI can do its magic, we are here to give you a word of caution. Don’t expect the AI software to make the hiring decision on its own. Many of the previous use cases proved to us that the technology just is not ready yet. “How to make sure the algorithm is really interpretable and explainable – that is quite far off.” [2]

If you are expecting AI and the algorithm to do their job well, it is equally important for you to check your company’s or organization’s readiness for AI, in order to maximize its power. We all know that identifying patterns and making predictions requires a large quantity of data. With the widespread open-source algorithms, the real game changer is the data used to train the algorithms. Any business or organization should have a clear plan on how to generate the quantitative and qualitative data which will help your AI recruiting software get more accurate results, this is economically profitable for your business as well.

JANZZ.technology supplies AI solutions for your recruiting system and helps you find the right skills and talents. The ontology JANZZon! and the smart matching engine JANZZsme! make complex problems such as job and skills matching computable and completely change the way we handle skills and talent searching. The applications of JANZZ.technology are structured semantically, meaning occupations, specialization, function, skills and qualifications etc. are interlinked logically. The JANZZ.technology applications can deliver meaningful results for complex searches in real time and across multiple languages. Our applications are constantly fed with new data generated from our users, therefore they become more accurate over time. Let the tools of JANZZ.technology assist you in finding your best fit candidates.  For a demo, please write now to sales@janzz.technology

[1] ideal. 2019. AI for recruiting: A definitive guide for HR professionals. URL: https://ideal.com/ai-recruiting/ [2019.05.28]

[2] Jeffrey Dastin. 2018. Amazon scraps secret AI recruiting tool that showed bias against women. URL: https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G [2019.05.28]