ScopeAT

In the world of sourcing and recruiting, the rumors of the death of Boolean search have been greatly exaggerated over the years. Despite many advancements in AI-powered automated matching between jobs and potential candidates, Boolean search has remained a required skill. This is largely because job descriptions themselves have historically not been accurate representations of actual job requirements, so matching from job descriptions has often had a high probability of returning results that do not match what the user is ultimately looking for.

AI has often been used in products to take basic keyword searches entered by users and expand them semantically. This has undoubtedly been helpful in sourcing and recruiting, primarily to basic users and those who are not particularly comfortable writing more detailed Boolean search strings. Now, the advent of generative AI and large language models (LLMs), which understand meaning at the word and sentence level, effectively allows users to employ natural language to write queries to return results of potential candidates. So, what does that mean for Boolean search?

When it comes to solutions that effectively implement LLM-powered natural language search and match, I don’t believe there will be much need for Boolean search.

However, some sources may never offer LLM-powered search, and the underlying skills associated with creating effective Boolean search strings will still be necessary for all sourcers and recruiters.

Instead of debating the demise of Boolean search, let’s concentrate on the challenges and fundamental skills required to effectively utilize systems for sourcing candidates, regardless of interface.

Critical thinking is the driver of effective search

It’s very important to understand that Boolean search has never really been about Boolean operators (“AND,” “OR,” and “NOT”). 

While understanding the functionality of Boolean operators is certainly important, it is probably only 5% of the process of writing effective Boolean search strings. The other 95% has to do with thinking critically about what terms to include, especially for maximum inclusivity, and what terms to strategically exclude.

As such, one of the most important core skills when it comes to sourcing talent is critical thinking, which will remain important in the age of AI.

Regardless of Boolean or LLM-powered search, a sourcer or recruiter must first have a solid understanding of what they’re looking for.

Sourcers need to keep in mind how many ways a resume can be written and interpreted

Sourcers also need to understand and appreciate the limitations and challenges associated with the text people use to represent themselves in resumes and social profiles. 

Two examples of these limitations and challenges:

There are many different ways people can express the same skills and experience, and if you don’t have a way for searching for all of the variations, you will exclude qualified candidates.Many people don’t explicitly mention all of their skills and experience, which means if you search for terms people don’t mention, you will exclude them — you’ll never even know they exist, although they are there to be found if you know how to retrieve them. I refer to these people as the “dark matter” of databases.

As I like to say, all searches and matches “work,” meaning they return results. This is true of both Boolean searches and automated AI matching. 

This makes it seem to the layperson that Boolean search and even sourcing is easy. Enter a few keywords or simply click a button and you have results to view. However, not all searches are created equal, and some searches are much more inclusive and find more qualified candidates than others while at the same time excluding the fewest.

The goal should be to find as many of the best people as possible, not just some of them

A talent strategy to consistently find as many of the best people to be found requires sourcers and recruiters who are able to think critically and better articulate their needs, regardless of interface (Boolean or natural language), in specific consideration of the data challenges associated with resumes and social profiles. This will enable the ability to consistently find more and better talent for their organizations using the exact same sources that sourcers and recruiters from other companies have access to. 

Make no mistake — this is a competitive advantage.

Sourcers need to effectively communicate their specific needs to computers

When sourcers and recruiters write Boolean search strings, what they are really doing is communicating with computer systems. This is known as human-computer interaction (HCI). 

The reason why Boolean search has been spoken and written about for the past couple of decades when it comes to sourcing and recruiting is that practically all information systems “understand” Boolean. This includes ATSs, CRMs, LinkedIn, job board databases, and internet search engines. 

That’s why Boolean search has been important for so long — because sourcers and recruiters have had to be able to effectively translate their talent needs into a “language” that computer systems could understand. When sourcing for potential candidates, a Boolean search string essentially represents a user asking the system what people they have that match what they are looking for.

The advent of generative AI and large language models (LLMs) has changed the game. 

With LLMs, we have finally reached the point where we no longer have to translate or explain our needs and wants to computers using Boolean queries. Search interfaces augmented with LLMs allow users to effectively search for potential candidates using natural human language, either written or verbal.

However, even though LLM-powered search and match solutions can understand natural human language, that doesn’t mean that people are necessarily very good at writing or speaking their sourcing/talent needs in natural human language. In fact, it is deceptively simple to use LLMs, but the quality of the prompt directly impacts the quality of the results.

This is why communication (and specifically HCI) — being able to effectively articulate talent needs to computer systems — is such an important underlying core skill required for sourcing candidates, regardless of the interface. 

Not all natural language prompts are created equal

The ability for human sourcers and recruiters to critically think and articulate their talent needs to the various information systems they have access to is directly tied to the quality, quantity, and consistency of sourcing results. This is true regardless of the search interface. 

Just as not all Boolean searches are created equal, not all natural language prompts are created equal. While there is no shortage of prompt engineering guides and advice available on the internet, it is still largely up to the individual user to figure out how to get the best results out of an LLM-powered interface — to get what they specifically need and want from the source.

If there’s one thing I know for sure, it’s that people who are and have been good at writing very effective Boolean search strings are highly likely to be good at prompt engineering and, by extension, good at extracting maximum value from LLM-powered solutions. This is because writing effective Boolean search strings requires the effective use of critical thinking and communication (HCI) skills — skills that transfer to prompt engineering, although I suspect it will take time and effort for people to adapt.

Six tips for writing effective search prompts

Knowing how to write effective prompts is a critical skill if you’re looking to extract maximum value from any LLM-powered sourcing and recruiting solution.

In addition to required job elements, here are some instructions you can consider including in your prompts to instruct an LLM-powered solution to make your searches more effective: 

Include a wide range of job titles that are equivalent to or closely related to the target role to help capture candidates who may have the necessary skills and experience but hold different titles.Use a broad set of keywords and phrases related to the required skills and experiences, including both technical and soft skills.Incorporate synonyms and alternate phrases for skills, experiences, job titles, education, etc., to account for all of the different ways candidates may describe their qualifications.Include common acronyms and abbreviations (CRM, DEI, QA, etc.) related to the skills and required elements of the job.Consider variations in spelling and formatting to ensure you don’t miss qualified candidates due to these differences.Include terms related to industries or roles that may have transferable skills to the target role to help identify candidates with relevant experience from different backgrounds.

I often use the instruction to be “maximally inclusive,” which should cover all of the above, but at this time, I still recommend providing more detailed instructions. Otherwise, you’re putting a lot of faith in the LLM to properly and exhaustively interpret and adhere to a short two-word prompt element. 

As such, you can also consider instructing the LLM to “ensure the search criteria are unbiased, inclusive of diverse backgrounds, and focus solely on skills, qualifications, and experience relevant to the job requirements.”

If you plan on using a job description as a part of your sourcing prompt, you can also instruct the LLM to avoid overly restrictive qualifiers (such as years of experience or specific certifications) which may inadvertently exclude potentially qualified candidates.

In the very near future, well-engineered solutions will likely begin to prompt their users to help them better explain their talent needs.

However, even with this advancement, it will still be important for users to be able to effectively articulate their talent needs. This will come down to critical thinking and communication skills, which have always been, and will continue to be, the core skills of sourcing top talent.

Glen Cathey is a strategic thinker and global keynote speaker with extensive experience in talent acquisition and leadership. He is passionate about making a difference, developing others, and solving problems. Glen has served as a thought leader for sourcing and recruiting strategies, technologies, and processes for firms with more than 2 million hires annually. He has played a key role in implementing and customizing ATS and CRM systems, and has hired, trained, and developed large local, national, global, and centralized sourcing and recruiting teams. Glen has spoken at numerous conferences, including LinkedIn Talent Connect, SourceCon, Talent42, and Sourcing Summit Europe.

Uncategorised