Market Salary Benchmarking

Market salary benchmarking is an AI agent that collects compensation data from available sources and produces a justified range for a specific position. The solution is triggered by a recruiter or HR manager when a new vacancy opens, replacing manual searches across aggregators and informal "favor" calls to colleagues. The result is a summary report with a range, median, and justification for each figure.

What the agent does, step by step

1. Accepts input parameters from the recruiter: job title, grade (junior/middle/senior/lead), city or work mode, industry, optional key skills.
2. Searches for matches in open sources: job aggregators, published salary reviews, industry reports, available via connected APIs or a RAG index.
3. Pulls internal rates from the HRIS for comparable positions — so the external market is compared against what the company currently pays.
4. Normalizes the data: converts salaries to a single currency, gross/net, and period (year or month), excludes outliers and duplicates.
5. Calculates the median, 25th and 75th percentile across the sample, and separately shows the spread by location and grade.
6. Generates a structured report with sources, publication dates, sample size, and a text summary for presentation to the finance department.
7. Saves the artifact to the HRIS or a shared repository — so that in six months it is possible to compare how the market has shifted for the same role.

What automation does not do

• Does not replace negotiations with the candidate. The final offer still depends on motivation, counter-offer, and negotiating position — the agent provides a reference point, not a decision.
• Does not account for the hidden portion of competitors' compensation. Options, RSUs, relocation bonuses, and non-monetary benefits are often not published — the agent works with what is available.
• Does not guarantee data accuracy. Output quality is limited by source quality: if few vacancies are publicly available for the required niche, the sample will be small, and the agent reports this honestly.

The solution is positioned as a risk-reduction tool, not a replacement for an experienced recruiter. The AI agent takes on the routine of collection and normalization, freeing the person for interpretation and negotiations with the candidate.

Technically, the solution is built on two patterns: RAG Q&A for searching unstructured salary data sources and an analytics module for normalization and statistics calculation. The agent runs on top of HRIS as a vertical SaaS layer and does not require replacing the current HR records system.

Data flow architecture

1. Trigger. The recruiter creates a job draft in HRIS or fills out a request form with role parameters. A webhook or manual trigger passes the payload to the agent.
2. External source collection. The agent queries connected sources — public APIs of job aggregators, indexed salary surveys, industry PDF reports in the RAG store.
3. Internal context. In parallel, the agent pulls from HRIS the current rates of employees in comparable positions, the history of recent offers, and approved compensation grades.
4. Normalization. The LLM step brings heterogeneous data to a unified schema: currency, periodicity, gross/net, fixed component versus bonuses.
5. Statistics. The calculation module (not LLM) computes the median, percentiles, sample size, and spread by location. This is the deterministic part — its results are reproducible and auditable.
6. Report assembly. The LLM generates a short narrative with justification of the figures, references to sources, and warnings about weak points in the sample.
7. Delivery. The report is published in HRIS as an attachment to the job, sent in Slack to the recruiter, and stored in the archive for later comparison.

Implementation steps

1. Source audit. The Grow2.ai team together with the HR manager compiles a list of salary data sources available to the company — public surveys, subscriptions, local aggregators.
2. HRIS connection. Setting up data reading for employees and jobs via API or export. Writing reports back is optional, at the second stage.
3. Grade configuration. The company's grade matrix is loaded into the agent so that external data maps to the internal position system.
4. RAG index assembly. Salary surveys and industry reports are loaded into the vector store for queries such as "median salary of a DevOps engineer in Poland".
5. Calibration. Using historical jobs, it is verified how closely the agent's responses match the company's actual offers. Source weights are adjusted.
6. Launch with one recruiter. A pilot on 5-10 jobs with feedback — what is wrong with the wording, which sources are not accounted for.
7. Rollout. Connecting the remaining recruiters, training in a one-hour format, documenting the process in the HR playbook.

Solution components

The solution uses an AI model or a comparable model for the normalization and narrative assembly steps. Deterministic statistics are moved to a separate module so that the figures in the report are reproducible — this is the key to trust from the finance department when approving the salary range.

The solution requires basic data readiness and a single process owner on the HR side. Without HRIS, implementation is possible, but will require manual export of comparable rates — this limits the frequency of use.

Access and data

• HRIS with the ability to read vacancies and employees via API or regular export (BambooHR, HiBob, Personio, 1С:ЗУП and similar).
• Company grade matrix — a description of levels and areas of responsibility for key roles. If it does not exist, the first step of the project is to compile it at least in a minimal form.
• List of market data sources, that the company is willing to use: public surveys (DOU, Habr, industry associations), subscriptions to paid reports, local job aggregators.
• Current compensation policies — fixed component vs. bonuses, gross/net, review frequency.

Team readiness

• Process owner — HR director or lead recruiter who makes decisions on the report structure and source calibration.
• IT access for HRIS integration — typically one sprint is sufficient with the internal IT team or a contractor.
• Readiness to pilot on 5-10 real vacancies before full rollout to the entire hiring team.

Implementation timeline

A typical project of "weekend" complexity level fits within 2-4 weeks provided that HRIS is already operational and the grade matrix exists. Without a grade matrix, 1-2 weeks are added for its formalization. If the data market for the company's niche is limited (narrow industry, small region), the timeline remains the same, but the sample quality will be lower — this is communicated honestly at the source audit stage.