ChatGPT is helping the state of Minnesota’s Enterprise Translations Office (ETO) deliver high‑quality translations across state government on time, stretching a six‑person team serving the state’s 25 agencies that reach Minnesota’s diverse communities, including sizable populations speaking Spanish, Somali, and Hmong.
The story starts in 2022, when the Minnesota Legislature authorized a new, centralized Enterprise Translations Office (ETO) and funded six linguists to provide translations for state agencies representing 50,000 total employees. Adam Taha, director of ETO and fresh from a long federal career as a linguist and language officer, saw the mandate and did the math. ChatGPT was needed to multiply linguistic forces and provide an efficient service. The team began experimenting with ChatGPT Enterprise.
Initially, Taha heard some pushback about using AI. So to test quality, he ran blind comparisons. Without knowing the source of a translation, native speakers rated ChatGPT’s Spanish translations higher than human‑only translations. With that signal, ETO began operations mid‑2024. One year in, they’ve built out a polished workflow: agencies submit documents through a simple portal, ChatGPT does a first pass, and human linguists review and perfect the output. Corrections are then fed back weekly so the system keeps learning. In Somali, Taha says baseline accuracy has climbed from about 90% to 94–95% purely through this learning loop.
Taha’s performance metric is streamlining government through “cost avoidance”: how much the state doesn’t spend on outside vendors because ETO delivers in‑house. He estimates the office now saves Minnesota between $120,000 and $140,000 per month while handling 400–600 translation requests monthly, with a focus on Spanish, Hmong, and Somali. The gains aren’t just financial. “More and better translations are reaching people on time,” Taha says, pointing to urgent public‑health announcements, voter information from the Secretary of State, and Department of Public Safety materials like the driver’s manual and Class D exam.
These efficiency and productivity gains in Minnesota are similar to what other states have experienced after introducing ChatGPT to their workforces. Results from a pilot program in Pennsylvania found state employees saving an average of 95 minutes a day using ChatGPT for writing, research, summarization, and IT support. Another pilot program with North Carolina’s Department of State Treasurer found employees saving 30 to 60 minutes a day, especially on drafting and editing tasks.
In Minnesota, what began as a budget constraint has become a statewide capability: faster turnarounds, consistent voice, lower costs, and, most importantly, information people can actually use.
