The “Teach your AI recruiter” step, with evaluation suggestions generated from the job title and description instead of title-taxonomy mapping — including niche terms like Mazatrol and Renishaw probing that the taxonomy doesn't have. Every suggestion is one-click accept; nothing auto-populates. Use the pill at the bottom to switch between the Generating, AI suggestions, and Taxonomy fallback states.
Share essential job details
Precision Manufacturing Inc. is hiring a 2nd-shift CNC Machinist for our Milwaukee plant. You'll set up and run Haas VF-2 vertical machining centers and Mazak lathes with live tooling, working from blueprints and GD&T callouts to hold tolerances to ±0.0005".
Day to day you'll make program edits at the control (Fanuc and Mazatrol), set tool offsets, verify first articles with CMM and Renishaw probing, and keep preventive maintenance logs. Strong shop math and precision measurement skills are a must. 2nd shift runs 2pm–10pm with a $1.50/hr shift premium.
Choose the AI recruiter that best fits this role
Define what makes an ideal candidate
Choose where and how your AI recruiter finds the best candidates for this role
Who should your AI recruiter source for this job?
How would you like your AI recruiter to manage this job's visibility on job boards?
Decide how your AI recruiter should evaluate and interview candidates
Do you want your AI recruiter to schedule interviews for you?
Add external candidates to this job by uploading their resumes
How should your AI recruiter handle these candidates?
Three behaviors that aren’t obvious from clicking around the prototype.
Reading the post. Suggestions are generated from the display title and description, with the taxonomy as a guide — clean matches keep their taxonomy IDs so structured data keeps working, and niche terms like Mazatrol arrive as custom entries instead of being dropped.
Accepting a suggestion. Nothing auto-populates — a recruiter’s click is what turns the AI’s purple suggestion into a real evaluation input, and the chip goes neutral the moment it’s theirs. A bad guess can’t silently skew screening or scoring.
When generation fails. The fields quietly fall back to today’s taxonomy suggestions and activation never blocks. Generated, shown, and accepted counts are tracked, so we can tell whether the model is helping — and where the taxonomy has gaps.