Recruiter Magazine 27th of July 2005 - By David Adams

Most recruiters would argue that nothing beats face-to-face contact for finding the best fit for a role. Yet software systems are becoming ever more sophisticated at identifying good candidates. David Adams asks: are we approaching the age of robot recruitment?

The scary consequences of the machines getting out of hand are staple of science fiction. Who could forget the polite but psychotic computer HAL in 2001: A Space Odyssey, the military defence system hellbent on nuclear armageddon in War Games and the killer robots of the Terminator movies? Back in the real world, high-profile system malfunctions such as that which crippled the UK Passport Service in 1999 are a stark reminder that computers are not infallible. Despite their flaws, most modern business rely on computer systems to perform mundane administrative tasks.

They are fast, always concentrate, don't get bored and cost a lot less than their human equivalents. Since the late 1990's, the development of software that can understand natural language has enabled recruiters and corporate HR departments to introduce automated processing of candidate resumes, whether these are received by email, through job boards or corporate websites.

These systems sit between a company's e-mail or web systems and their existing candidate databases, from where they extract resumes from e-mails or websites automatically and push some or all of the data they contains into databases. For recruiters, this transforms a mountain of unstructured information into searchable data and raises their chances of finding the perfect candidate. The danger is that recruiters come to rely too heavily on these systems when they could be meeting candidates face to face.

Speed demons
Not even the most excitable promoters of these technologies suggest they could manage the whole candidate screening process, but they are keen to highlight the time and cost savings they offer.

"It doesn't mean the consultant stops assessing the candidate in all the usual 'touchy-feely' ways, but it does save you having a bank of administrators typing in resumes," says Adele Poole, managing director of Hot Lizard. Her company sells a resume-parsing system - which searches resumes for keywords, concepts and word structures according to a recruiter's criteria - developed by technology specialist Infogistics as part of its web-based recruitment management system.

"If consultants have to spend more time searching for data, that represents a loss to the company," explains Sergei Makhmodov, commercial director at Daxtra Technologies, which sells a resume-parsing system developed by a team at Edinburgh University, "and if you do this job manually you usually end up with a resume backlog. If a competitor processes the resume faster than you, you may lose a good candidate."

When it comes to finding the best fit for a job, however, speed isn't everything. Even the most efficient IT system will be blind to some of a candidate's qualities.

"Software makes life easier, but that's not always better," warns Claire Owen, managing director of marketing recruitment specialist Stopgap. She recalls a telecommunications company she dealt with that relied to heavily on keyword searching technology for recruitment. She felt a number of good candidates had been passed over because they hadn't included the 'right' word in their resume.

"I'm all in favour of efficiency, but not if it damages the business," she says, "a lot of agencies are very sales-driven, so if they think this means they can handle more business they will go for it, but the client may end up missing out on some good people.

Hard to handle
Modern resume-handling software does much more than match keywords to queries, however. The Daxtra system, for example, can identify and rectify its own mistakes, spot duplicate or replacement resumes for those already stored on the system and update records accordingly, and guess its implied information.

"It can suggest what industry sector a candidate is working in by looking at their skills," claims Makhmodov, "and because it looks at context it can tell if someone who uses words 'Oracle database' in their resume is a user of the technology, rather than a developer." The system can also extract information from job vacancy advertisements to run structured queries through candidate databases.

Banking firm UBS uses the Daxtra system to sort through 120,000 resumes each year. It enables candidates who apply online to populate application forms much more quickly by extracting information straight from their resumes and pushing it into a form. Sergei Makhmodov claims customers can save between 70% and 90% of the sum they would spend on administrative staff performing the same task. IT recruitment agency MSB International has also made cost savings by using a resume-parsing system - an estimated £50,000 over the first year by using it - but IT director Alan Ovenden ensures that consultants don't rely on the software to do everything.

"We use it primarily for the extraction of personal details," he explains, "we don't use it to gauge candidate skills, because our tests have taught us that expert resourcers will always do it slightly better, and we want to do it the best that we possibly can, not just the quickest."
MSB has been using the system to streamline the candidate registration process and generate an automatic thank-you letter to each candidate. Candidates can also upload their resumes onto MSB's website without having to fill out multiple forms.

Fuzzy logic
Uploading or sorting resumes is one aspect of the recruitment process that software systems are more than capable of handling. But what about the more complex aspects, such as compiling an interview shortlist or matching 'soft' skills to clients' requirements?

Where the systems struggle most, according to Adele Poole of Hot Lizard, is when dealing with the least structured information, such as previous work experience. "Our system is extremely accurate on the stuff that's easy to pick up, like names, addresses, date of birth, but it's more difficult to pick up the more 'touchy-feely' stuff," she admits.

Accuracy is another issue. Alan Ovenden, for example, has experienced instances when MSB's system has failed to pick up a candidate's name, or pulled out the wrong name of a referee, but he estimates that this only happens every couple of weeks, and that it was a problem even when resumes were being processed manually.

Sergei Makhmodov claims accuracy levels for the Daxtra system are between 92% and 93%, compared with an estimated 95% accuracy rate for human data entry clerks.

"Things like previous work experience are often written in different ways," he says, "but much of the time consultants are searching using things like postcodes and salary ranges, areas where we're very accurate."

These reassurances are not enough for all recruiters, however. Claire Owen at Stopgap agrees that the systems could be helpful for processing large numbers of applications for jobs "where the personality of the individual is irrelevant to the continued success of the organisation", or to automate some aspects of candidate assessment. Her worry is that too much technology could counteract a recruiter's ability to use his or her imagination or spot hidden qualities.

"I think the danger comes if people think this is the answer to all their prayers," she says, "sometimes it's good to turn recruitment on its head and say: 'Let's not worry so much about technical skills, you can teach those, let's look at personalities'. And those sorts of things are not easily determined by a machine."

Succeeding in recruitment is frequently about being the first off the starting grid, so while no one claims to have invented a robot that can spot raw potential, resume systems can provide a massive return on investment and save time, salaries and overheads. That's certainly not something the average recruiter would find frightening.

Article by David Adams - Recruiter 27 july 2005 - http://www.recruitermagazine.co.uk