What is Artificial Intelligence?

Artificial Intelligence (AI) refers to technology that exhibits a level of intelligence normally displayed by humans, such as the ability to process text-based conversations or speech. An AI agent is one that acts autonomously, taking input from its environment and executing different actions. Closely related to AI is machine learning, a field of computer science that focuses on giving machines—as computing pioneer Arthur Samuel described it in 1959—”the ability to learn without being explicitly programmed.” It is a specific type of AI that characterizes how machines “learn” from existing data patterns to make inferences or new predictions.

What are common recruiting applications for AI?

An increasing number of organizations across multiple industries are deeply involved in automation projects, with many using AI and robotics to perform routine tasks, augment human skills, or restructure work entirely. Common uses for intelligent tools in talent acquisition include:

• Applicant Screening: analyzing candidate profiles and determining their fit for a job opening
• Profile Enrichment: automatically sourcing additional information about a candidate from publicly available data across the web
• Candidate Sourcing: surfacing qualified candidates from internal and external talent pools for current vacancies
• Personalized Assessments/Training: tailored solutions that adapt to each candidates’ skills and abilities
• Candidate Matching: identifying and evaluating the strongest active or passive candidates for an open position based on their relevant skills, experience, or other defined criteria
• Chatbots: conversational UI for candidates or prospects for pre-screening, Q&A, scheduling, and more
• Programmatic Job Advertising: automated job distribution and budget optimization based on historical data and real-time campaign performance

What is SmartAssistant?

SmartAssistant is the industry’s first and only native AI-powered recruiting service. Designed with a focus on candidate screening, sourcing, and matching, SmartAssistant allows recruiters and hiring managers to automatically evaluate job applicants with a Match Score, surface candidates from internal databases for open positions, and automate job alerts to candidates for opportunities that align with their skills and experience. Learn more here.

Is SmartAssistant available to all SmartRecruiters customers?

SmartAssistant is an add-on purchased on top of the platform, but as an existing customer, it can be activated in a few simple steps, as the necessary data already exists within the base platform.

How is SmartAssistant activated for my organization?

Activating SmartAssistant is a fast and smooth process that lasts between 3-6 business days, with no IT or any other support required from customers. The activation steps include indexing and processing of customer data by the SmartRecruiters engineering team, enabling the add-on for the customer account by a designated CSM, and assigning seats for individual users in the Admin section.

What differentiates SmartAssistant’s design from other AI models?

SmartAssistant operates completely within the SmartRecruiters Talent Acquisition system. While other AI models require third-party integrations, SmartAssistant’s native functionality means data never leaves the ATS platform for processing by a separate vendor. All recruiting actions (new hires, rejections, etc.) happening within the SmartRecruiters environment generate data points which are used to improve the accuracy of SmartAssistant’s matching algorithms.

Additionally, SmartAssistant is continuously learning through regular updates and audits as new jobs, titles, and skills emerge, ensuring the model maintains a high level of accuracy.

What makes SmartAssistant’s matching technology unique?

Some AI models are built on simple keyword matching technology. For example, in these systems, if a resume or CV lacks a certain vocabulary or “keywords”, the system fails to identify a match. Alternatively, SmartAssistant relies on multiple algorithms that cluster concepts to infer skills that are not always explicitly mentioned in a candidate’s resume or CV. SmartAssistant automatically enriches the candidate’s profile with this additional meta information as part of the Match Score and recommendations process.

Is the SmartAssistant AI technology maintained in-house?

Yes, SmartAssistant is developed and updated by a dedicated SmartRecruiters product and engineering team.

What datasets were used to build/train the SmartAssistant AI?

SmartAssistant was trained on 5 years of job and resume data collected in 12 countries. The model processes a wide range of datasets pulled from various sources, including:

• Employee profile data (current and historical), together with skills, education, project assignments, performance reviews, etc.
• Applicant Tracking System (ATS) candidate data: interview ratings and other feedback from hiring teams

Employment databases: information on skills, abilities, knowledges, work activities, and interests associated with certain occupations

• Graph databases containing all work-related entities combined with weighted interconnections to create clusters of skills and abilities
• Company databases: industry data, company type, size, location, etc.
• Educational institution databases: degrees, diplomas, certifications, licenses, etc.
• Collections of sample job descriptions
• Collections of anonymized resume data

SmartAssistant is also continually updated with new data and maintained for accuracy by a dedicated SmartRecruiters product and engineering team.

Is SmartAssistant GDPR compliant?

Yes, due to a large number of SmartRecruiters customers in Europe, GDPR compliance was a top priority in SmartAssistant’s development. SmartRecruiters appended its privacy policy in April 2018 with specific language clarifying how candidate data is used.

If AI learns from past behavior, isn’t it likely to learn bias from users?
Say an organization only employs people between the ages of 25 and 40, won’t the algorithm assume that these must be consequently more successful in that company and therefore better hires?

The SmartAssistant algorithms do not analyze nor use any personal candidate data (name, age, gender, email address, phone number, or any answers to required diversity questions) when calculating a match score, mitigating potential bias on any of these factors.

The SmartAssistant matching algorithms learn through feedback loops provided by recruiters, and by continuous analysis of hiring decisions (e.g. applicant conversion/rejection, interview score, etc.) to improve accuracy and matching.

SmartAssistant also leverages a massive collection of general market data to compare against customer data to offset any existing biases within the customer data.

Additionally, algorithmic learnings and improvements are applied across the entire SmartRecruiters platform to enforce data-driven hiring decisions. Because SmartAssistant operates within the SmartRecruiters ecosystem, existing bias is easily discovered, isolated, and removed from further training data sets.

Can I customize the Match Score parameters?

No, customizing the matching algorithms is not possible. This prevents the introduction of user bias as well as any breach of compliance. Matching algorithms are under continuous development and adjustment by the SmartRecruiters Data Science team. Centralized control and QA ensure high accuracy and scalable solution engineering.

Who on the hiring team can see Match Scores?

Only recruiters with an assigned seat are able to see candidates’ Match Score and other SmartAssistant features.

Are candidates able to see their Match Score against a job?

No, as SmartRecruiters does not want candidates to be able to cheat the system. If the process were too transparent, some candidates could find ways to unfairly exploit algorithms in order to put themselves in higher standing.

SmartAssistant is, by definition, an assistant. Recruiting teams ultimately make the final hiring decision, and it would be unfair to mislead candidates with false hopes over a potential job offer based solely on a high matching score.

Why are some/all of my applicants not scored?

In order to generate a Match Score or Recommendation, SmartAssistant requires a job description and a parsed resume. Applicants with an attached PDF or .doc resume, or those with no profile data, will not be scored.

Currently, only English content is being considered, though SmartRecruiters plans to expand to additional languages in future updates. This applies to both the job description and candidate profile.

In rare cases, the SmartRecruiters system can be under heavy load, which leads to increased time for applicant processing.

How does SmartAssistant evaluate candidates with poor resumes or those with little or missing data?

In such cases, SmartAssistant attempts to enrich the profile with the model’s proprietary Skills Graph to give these candidates a fair chance to be discovered.

Do screening questions and/or knockout questions influence the Match Score?

Answers to screening questions are not evaluated for scoring. Candidates that are rejected by a knockout question will be scored similarly to all other applicants.

Which candidates are part of the Discovery and Recommendations?

Included:

• Candidates who have applied to jobs in the past or have been added through other sources
• Internal employees with a profile

Excluded:

• Candidates outside of a user’s system access level
• Past candidates rejected for principal reasons (e.g. bad culture fit and similar)
• External candidates (e.g. third-party resume databases)

Are internal candidates included in SmartAssistant?

Yes, as long as there a profile exists for those candidates.

Which jobs will show up in the “Recommended Jobs” widget?

Only jobs within a user’s access group will be eligible.

Are candidates notified when they’re added to a job as part of Discovery/Recommendation?

Candidates are not automatically notified when they are added to a job, but when the recruiter reaches out to them. These candidates appear as “Lead”, with the source “SmartAssistant”.

Are past interview feedback and ratings taken into account when recommending candidates?

Yes, SmartAssistant uses additional information pulled from multiple internal and external data sources to determine candidates’ fit for a job.

Is the list of recommendations ever updated?

Yes, recommendations are updated on a regular basis in order to include new, relevant candidates that have been added to your database.