Calibrate Tezi
Train Tezi on your ideal candidate profile through traditional filters and nuanced AI evaluation
Calibration trains Tezi’s system to understand your ideal candidate profile, enabling it to automate sourcing and/or inbound screening. Typically, calibration takes 15 minutes to a few hours across multiple sessions. This upfront investment enables Tezi to run recruiting at scale, giving you time back and accelerating hires.
How to calibrate
On the Calibration criteria page in role setup:
1️⃣ Set filters
Filters are objective “resume facts”: job titles, companies worked at, years of experience, schools, industries, locations, and keywords. This information is explicit and unambiguous.
Our system intelligently configures a starting set of filters based on your JD. Edit the pre-populated filters as needed to get “in the ballpark” of your target profile.
You can see available filters and add new ones from the “Add criteria” dropdown:
Aim to filter down to a pool of <5,000 candidates through filters. This helps speed up our AI Max’s processing time in the next step of calibration below.
2️⃣ Set AI reasoning criteria
AI reasoning criteria are complex, natural language criteria that Max evaluates through higher-level reasoning and interpretation of a resume, similar to how a recruiter might read between the lines. For example:
- Sales: consistently hits or exceeds quota
- Engineering: is an experienced data infrastructure engineer based on their profile or project summaries, even if their current role is a generic “Software Engineer” title
- Marketing: has experience marketing to highly technical audiences like CIOs, CISOs, and CTOs
- Any function: has been promoted multiple times in the last 5 years
Add AI reasoning criteria to capture the nuances of your search. After filters narrow the pool, Max will read and interpret each profile and find the candidates that match your AI criteria.
3️⃣ Retrieve matching profiles
Click “Retrieve matching profiles” to begin Max’s analysis of the candidate pool. Max will return a sample of candidates who match all criteria, plus an estimate of the total number of matches.
4️⃣ Iterate on criteria
Iterate on your criteria until you are satisfied with the candidates returned. We recommend reviewing at least 10 profiles before finalizing criteria.
When iterating on criteria, to evaluate whether a requirement is improving results, try implementing the opposite to see who is being excluded and whether that exclusion makes sense. For example, if you have a requirement like Has a degree in Computer Science, consider testing Does not have a degree in Computer Science.