How AI Candidate Scoring Works (And Why It Saves 10+ Hours Per Hire)

The screening bottleneck every recruiter knows

Every open role generates a stack of applications. A mid-level marketing position might attract 120 CVs. A software engineering role at a well-known company can pull in 300 or more. The recruiter's job is to find the five to ten candidates worth interviewing from that stack, and the traditional approach is manual review: open each CV, read it, decide whether the candidate is plausible, move on to the next one.

Manual CV review takes between three and five minutes per application when done properly. For a role with 150 applicants, that is seven and a half to twelve and a half hours of reading before a single interview is scheduled. Multiply that by the number of open roles a recruiter handles simultaneously, and the arithmetic becomes punishing. Corners get cut. CVs get skimmed rather than read. Strong candidates get missed because their CV happened to land in the middle of an exhausting review session.

AI candidate scoring was designed to solve this specific problem: how do you identify the strongest candidates in a large applicant pool without spending days reading every CV in detail?

How AI candidate scoring actually works

AI candidate scoring analyses each applicant's CV against the requirements of a specific job and produces a numerical score from 0 to 100. The score represents how closely the candidate's profile matches what the role requires. A score of 85 means the candidate's skills, experience, education and background align strongly with the job. A score of 35 means there are significant gaps between the candidate's profile and the role requirements.

The scoring process examines multiple dimensions of each candidate's profile:

Skills match evaluates whether the candidate possesses the technical and soft skills listed in the job requirements. The AI reads the candidate's CV for evidence of these skills, looking beyond exact keyword matches to identify relevant capabilities described in different terms.

Experience relevance assesses the depth, recency and relevance of the candidate's work history relative to what the role demands. A candidate with eight years of directly relevant experience in the same industry will score differently from a candidate with two years of tangentially related experience.

Education alignment checks whether the candidate's educational background matches the stated requirements. This includes degree level, field of study, and any specific certifications or qualifications the role demands.

Keyword relevance examines the overall alignment between the language, terminology and domain-specific vocabulary in the candidate's CV and the job description. This dimension captures industry knowledge and familiarity with the specific tools, methodologies and frameworks the role involves.

Configurable weights: not all dimensions matter equally

Different roles have fundamentally different requirements, and the scoring system needs to reflect that. A graduate trainee programme cares primarily about education and potential. A senior infrastructure engineer role cares primarily about specific technical skills and depth of experience. A sales director position might weight experience and industry knowledge above formal education entirely.

Treegarden's AI scoring allows recruiters to configure the weight assigned to each scoring dimension on a per-job basis. The four dimensions — skills, experience, education and keywords — can each be assigned a weight that reflects their importance for the specific role. A technical lead position might use weights of 40% skills, 35% experience, 15% education and 10% keywords. An entry-level analyst role might use 20% skills, 10% experience, 50% education and 20% keywords.

This configurability is not a minor feature. It is the difference between a scoring system that applies the same formula to every role — producing scores that are meaningless for half of them — and one that genuinely reflects what matters for each specific hire. The recruiter understands what a given role needs. The weight configuration lets them encode that understanding into the scoring process.

What the score badges mean

Raw numbers need interpretation at a glance, especially when a recruiter is scanning a Kanban board with dozens of candidate cards. Treegarden uses a colour-coded badge system that makes score interpretation instant:

Green badge (70-100%) indicates a strong match. The candidate's profile aligns well with the role requirements across the weighted dimensions. These candidates should be prioritised for detailed review and interview scheduling.

Blue badge (40-69%) indicates a moderate match. The candidate has relevant elements in their profile but also has gaps or areas where their background diverges from the ideal. These candidates are worth reviewing, particularly if the applicant pool is competitive or the role is hard to fill.

Yellow badge (20-39%) indicates a weak match. The candidate's profile shows limited alignment with the role requirements. They may have some transferable skills or relevant education, but significant gaps exist across most scoring dimensions.

Red badge (below 20%) indicates a poor match. The candidate's profile has very little overlap with what the role requires. In most cases, these candidates are not suitable for the position as described.

The real time savings: where 10+ hours per hire comes from

The time savings from AI scoring are straightforward to calculate. Consider a typical role that receives 120 applications:

Without AI scoring: The recruiter reviews all 120 CVs manually. At 3-5 minutes per CV, this takes 6 to 10 hours. The quality of review degrades after the first hour as fatigue sets in. Strong candidates in the later portions of the pile receive less attention than those reviewed first.

With AI scoring: The AI scores all 120 candidates in minutes. The recruiter focuses detailed review on the top 20-30 candidates (those with green and blue badges), spending 5-7 minutes on each for a thorough assessment. Total time: roughly 2-3 hours. The remaining candidates are not discarded — they remain accessible and visible — but the recruiter's time is concentrated where it has the highest return.

The difference is 4 to 7 hours per role. For a recruiter managing 5-10 open roles simultaneously, that is 20 to 70 hours recovered per hiring cycle. More importantly, the quality of the shortlist improves because every candidate receives the same objective initial assessment, regardless of when their application arrived or how tired the reviewer was when they reached it.

GDPR compliance: advance-only, never auto-reject

One of the most important aspects of Treegarden's AI scoring is what it does not do: it never automatically rejects a candidate. Every candidate who applies for a role remains visible on the recruiter's board regardless of their score. The AI identifies strong candidates and surfaces them; it does not eliminate weak ones.

This is not just a design choice — it is a legal requirement under GDPR Article 22, which gives individuals the right not to be subject to decisions based solely on automated processing that produce legal effects or significantly affect them. An automated rejection from a job application clearly qualifies as a decision that significantly affects the individual. By operating on an advance-only model, Treegarden's AI scoring remains firmly within GDPR compliance.

The practical benefit of this approach extends beyond legal compliance. Recruiters retain full visibility of their entire candidate pool. A candidate who scores 45% for one role might be an excellent fit for a different position that opens next month. An advance-only system preserves that optionality; an auto-rejection system destroys it.

Privacy-first AI: no data leaves your control

A common concern with AI-powered recruitment tools is data privacy. Where does the candidate data go when it is processed by AI? Which third-party services have access to applicant information?

Treegarden uses Ollama, an open-source AI framework, for all candidate scoring. This means candidate data is not sent to OpenAI, Google, or any other external AI provider. The scoring runs within Treegarden's own infrastructure, and candidate information stays within the system the recruiter already trusts with that data.

For organisations subject to strict data governance requirements — financial services, healthcare, government contractors — this architecture matters significantly. There is no additional data processing agreement to negotiate, no third-party sub-processor to add to the GDPR register, and no risk of candidate data being used to train external AI models.

How to start using AI scoring effectively

Implementing AI scoring is not complex, but getting the most out of it requires some thought about configuration. Here is a practical approach:

Step 1: Define your weight priorities. Before scoring candidates for a role, spend two minutes thinking about what actually matters most. Is this a skills-heavy role or an experience-heavy role? Does education matter, or is it a nice-to-have? Set your weights accordingly.

Step 2: Run the scoring. Initiate the AI scoring for your candidate pool. The process takes seconds to minutes depending on the number of applications.

Step 3: Start with the green badges. Review the high-scoring candidates first for detailed assessment. These are your most likely interviews.

Step 4: Check the blue badges for hidden gems. Moderate-scoring candidates often include strong prospects who did not use the exact language the job description used, or who have transferable skills from adjacent domains.

Step 5: Refine over time. After a few hires, review whether your weight configurations are producing shortlists that align with who you actually want to interview. Adjust weights for future roles based on what you learn.

Frequently asked questions

How does AI candidate scoring work in an ATS?

AI candidate scoring analyses each applicant's CV against the job requirements and assigns a numerical score from 0 to 100. The score is calculated based on configurable weights across multiple dimensions — typically skills match, years of experience, education level, and keyword relevance. The recruiter initiates the scoring process and can adjust the weight of each dimension per job.

Does AI scoring automatically reject candidates?

Not in a GDPR-compliant system. Treegarden's AI scoring is advance-only — it identifies and surfaces strong candidates but never automatically rejects anyone. This approach complies with GDPR Article 22, which restricts fully automated decisions that significantly affect individuals. Every candidate remains visible to the recruiter regardless of their score.

How much time does AI candidate scoring actually save?

For a typical role receiving 80-150 applications, manual CV review takes 3-5 minutes per candidate. AI scoring reduces this to seconds per candidate for the initial assessment, with recruiters spending detailed review time only on the top-scoring candidates. Most teams report saving 10-15 hours per hire on initial screening alone.

Can I configure what the AI scoring prioritises for each job?

Yes. Treegarden allows per-job weight configuration across four dimensions: skills, experience, education and keywords. For a senior engineering role, you might weight technical skills at 40% and experience at 35%. For a graduate programme, you might weight education at 50% and reduce experience weight to 10%.