AI Competitive Intelligence Engine

A NASDAQ-listed staffing and professional-services firm engaged Taller to build an AI-driven competitive intelligence system that could identify patterns, clients, and skill demand hidden inside public job-posting data.

The client wanted to understand what its competitors were doing in the market before those activities showed up through traditional sales channels. Competitors were posting job descriptions across staffing-agency sites, often under generic or anonymized client names, which made it hard to tell which companies they were serving, what roles they were filling, which skills were trending, and where demand was shifting across technology and finance. The real challenge was turning unstructured, public job-posting data into competitive intelligence: identifying the hidden client behind a posting and building a usable market map for the client’s commercial and recruiting teams.

Taller designed and built a pre-LLM AI system combining web automation, data engineering, unsupervised learning, supervised classification, neural-network modeling, and a custom taxonomy. The system collected public job-description data from competitor and market sources, then ran it through a set of AI pipelines built specifically for the recruiting domain.

The first two layers handled the data foundation: NLP pipelines extracted and grouped skills from job descriptions, and unsupervised clustering organized related roles into a consistent taxonomy the system could reason over.

The third layer classified jobs by area: technology, finance, and other commercial categories, so the client could analyze demand by business line, role family, skill set, and hiring trend rather than reading postings one by one.

The fourth and most distinctive layer was the job-unmasking engine. Taller trained a neural-network model on a labeled dataset built from public job descriptions and company information, so it could learn the writing patterns, terminology, and role structures associated with specific companies. In parallel, Taller built bots that searched the public web for matching wording and patterns, comparing anonymized competitor postings against open-market listings and company-specific language. By combining the neural-network prediction with web-pattern matching, the system could infer which company was most likely behind an anonymized posting. The result was a platform that could reveal competitors' clients, identify what they were hiring for, detect emerging skill demand, and surface market signals that would otherwise stay hidden. It was built from custom machine-learning research and domain-specific data engineering, before large language models were widely available.