AI Implementation · SimpsonScarborough · 2024–2025
AI Implementation in Action
The three case studies below are drawn from my direct professional experience. Each one started with an inefficiency I identified, a solution I built, and a result I measured.
01
Replacing a Specialized Research Tool with an AI Solution
Inefficiency
The research team at SimpsonScarborough depended on a $50k third-party platform to analyze open-ended survey responses. It created an ongoing vendor dependency and added cost without adding flexibility.
Solution
Built a LLM-based replacement and validated it against pre-coded datasets spanning multiple client projects, comparing thematic groupings and sentiment classifications against manually reviewed outputs to establish a measurable accuracy benchmark.
Result
LLM-based analysis came within 5–10% of the contracted platform across both dimensions. Presented findings to a decision-maker who approved the switch, eliminating the vendor contract and retaining full capability in-house at effectively zero marginal cost.
02
Automating Repetitive Data Work Through Vibe Coding
Inefficiency
Research projects at SimpsonScarborough involved recurring data preparation tasks such as cleaning survey exports, standardizing formats, and prepping files for analysis. These steps were time-consuming and prone to human errors.
Solution
Used AI tools to write and refine Python and R scripts that automated the most repetitive parts of the data cleaning process. Rather than building each script from scratch manually, used AI as a coding partner to generate, test, and adjust the logic.
Result
Reduced data preparation time by two to eight hours per project cycle depending on scope, and made the process more repeatable across projects. Demonstrated a practical model for how AI can extend a team's technical capacity without requiring everyone to be a programmer.
03
Eliminating Manual Data Pulls with an AI-Assisted Pipeline
Inefficiency
The Data Operations team at SimpsonScarborough manually pulled survey data on a recurring basis to feed reporting, creating a bottleneck that slowed turnaround and introduced room for human error.
Solution
Used AI tools to navigate the architecture of the Qualtrics API, analyze third-party vendor documentation, and work through advanced configuration problems in Tableau. Built a pipeline connecting the survey platform to Google BigQuery and surfaced the data through a live dashboard, automating the full data flow.
Result
Eliminated daily manual data pulls for the Data Operations team entirely, creating a real-time self-updating reporting system that required no ongoing analyst intervention.