How to Prompt

How to Prompt

How to Write Effective Natural Language Prompts for Candidate Search

The quality of your search results in Beesla depends largely on how you describe the role. This page explains how to write effective prompts that return strong candidates while keeping result volume and token usage under control.

You do not need special syntax. You just need to be clear.

Start With the Role, Not the Resume

Write prompts the way you would explain the role to a hiring manager or another recruiter.

Focus on:

• What the person actually does
• What experience matters most
• What would disqualify someone

Avoid trying to “optimize” language. Beesla is not matching keywords.

Include the Signals That Matter Most

Strong prompts usually include:

• Role title and seniority
• Core skills or technologies
• Relevant environment or company type
• Clear exclusions or dealbreakers

Example:

Senior backend engineer. Strong Python experience. Has worked at early-stage startups. Avoid candidates with frequent job changes.

This gives Beesla enough context to rank candidates accurately.

Use Job Descriptions Intentionally

Pasting a job description works well, but only if it’s used correctly.

Best practice:

• Paste the full job description
• Add a short plain-English note with priorities or exclusions

Example:

Use the job description below. Prioritize hands-on Python experience. Avoid candidates focused primarily on management.

This helps Beesla separate must-haves from nice-to-haves.

Be Specific Without Over-Constraining

Overly broad prompts return too many results.

Overly strict prompts return too few.

If results are too broad:

• Add seniority
• Add exclusions
• Narrow the role scope

If results are too narrow:

• Remove one constraint
• Broaden experience slightly
• Remove nonessential requirements

Small changes make a big difference.

Control Result Volume (and Token Usage)

Each candidate returned by a search costs tokens. You control that cost through prompt quality.

To reduce unnecessary results:

• Avoid vague role descriptions
• Avoid generic titles without context
• Explicitly exclude mismatches

Good prompts save time and tokens.

Things You Don’t Need to Do

You do not need to:

• Use Boolean operators
• List every possible skill
• Match resume language
• Optimize phrasing for search engines

Clarity beats complexity.

Common Prompting Mistakes

Writing like a job board post

Focus on signal, not formality.

Listing everything as required

This filters out strong candidates unnecessarily.

Rerunning the same prompt unchanged

Adjust based on what you learn from results.

What’s Next

Once you’re comfortable prompting:

• See Results and Ranking to understand how candidates are evaluated
• See Run History to track past searches
• See Tokens and Billing to manage usage