AI-Powered Compensation Insights

Why It Matters

Ask the AI assistant to compare roles, find listings, or summarize a company’s advertised ranges. The tools use job listing data. By default, medians and percentiles are calculated from the lower bounds of advertised ranges, converted to monthly amounts. You can ask for the midpoint or upper bound instead.

What You Can Ask

The AI assistant understands six types of compensation queries:

Capability	What It Does
Salary benchmarks	Returns min, max, median, 25th and 75th percentile salary for one or more roles or a job title keyword, optionally filtered by region, experience level, employment type, technology, or currency
Role comparison	Compares salary data across 2–4 roles side by side so you can see how compensation stacks up
Job listing search	Finds specific listings matching your criteria, filtered by technology, experience level, location, employment type, or minimum salary
Company insights	Shows a company’s compensation profile: salary ranges across roles, top technologies, and hiring volume
Market trends	Displays salary direction over time (up, down, or stable) with regional and technology breakdowns
Available role categories	Lists all role clusters available for benchmarking

Example Prompts

Try these in the Kit AI assistant:

• “What’s the median salary for a senior backend engineer in Warsaw?” Returns salary benchmarks filtered by role, experience level, and region.

• “Compare frontend developer vs full-stack developer salaries” Side-by-side comparison of two roles across all tracked regions.

• “Show me remote DevOps positions paying over 25,000 PLN” Searches listings with location, role, and minimum salary filters.

• “What salary ranges does Google advertise for data engineers?” Returns company-specific salary ranges and hiring patterns.

• “Are machine learning engineer salaries going up or down?” Shows 6-month trend direction with monthly averages.

• “What roles can I benchmark?” Lists the currently available role clusters.

• “Compare backend engineer, DevOps engineer, and cloud engineer salaries in EUR” Multi-role comparison with currency conversion.

• “Find senior React positions in Krakow” Searches listings filtered by technology, experience level, and city.

• “Benchmark helpdesk titles in Poznań on the midpoint of the advertised range” Benchmarks listings whose title contains “helpdesk”, using each range’s midpoint.

• “Benchmark Application Support and IT Support together, B2B only” Pools two roles into one benchmark.

Choosing What to Benchmark

Lower bound, midpoint, or upper bound

Benchmarks use the lower bound of each advertised range unless you ask otherwise. Ask for the midpoint or the maximum to see the middle or top of the ranges. Some listings advertise a single figure with no upper bound. Those count at that figure, and the answer says how many there were.

Several roles or a job title

A benchmark can cover:

• One role, such as “Backend Engineer”
• Several roles pooled together, such as “Application Support and IT Support”
• A job title keyword, such as “helpdesk” or “service desk”, across every role
• A role narrowed by a keyword, such as Application Support listings with “SAP” in the title

The answer names the roles and the keyword it used, and how many listings matched.

Which role does my job map to?

Kit sorts each listing into a role by the most specific keyword in its title. Ask “What roles can I benchmark?” to see the list. Some common matches:

Your role	Role to benchmark
Helpdesk, service desk, desktop or L1/L2 IT support (wsparcie IT, Specjalista IT)	IT Support
L2/L3 application or production support, technical support engineer, ITSM	Application Support
Customer service, customer care, player support	Customer Support
In-house IT generalist (Informatyk, Administrator IT), server administrator	System Administrator
Master data specialist	ERP Specialist

If no role fits, benchmark by title keyword instead.

Reading the Results

Every benchmark says how many listings it is based on and how many were left out, and why.

• Unknown employment type. Some boards don’t state whether a job is B2B or permanent. When you filter by employment type, those listings are left out, and the answer says how many.
• Unknown seniority. When you filter by experience level, listings with no known level are left out, and the answer says how many.
• Seniority coverage. Each benchmark lists how many listings each level (junior, mid, senior, lead, or unknown) has, how many state a salary, and whether the level came from the job board, the job title, or an AI guess. A level with only a handful of salaried listings is too thin to quote on its own.
• Small samples. Fewer than 10 listings are flagged as indicative only.

Tips for Better Results

• Be specific: include role name, region, and experience level when you have a specific question
• Use natural language: “How much do QA engineers make?” works as well as formal queries
• Compare up to 4 roles at once for side-by-side analysis
• Specify currency if you want results in something other than the default

MCP Access

The same compensation tools are available to external AI clients (Claude Desktop, Cursor, and others) via Kit’s MCP server. This means you can query your compensation data from any MCP-compatible tool without opening Kit.

See MCP Tools Reference for setup instructions and the full list of available tools.

Quick Checklist

• Open the AI assistant and try a salary benchmark question
• Compare two or more roles side by side
• Search for specific listings by technology or location
• Look up a competitor’s compensation profile
• Check market trends for your most important roles

Account admins can choose the provider, add account-owned keys, set spend caps, and inspect usage. See AI Providers & Spend Controls.