Prompt LocalPrompt Local
 

Issue No. 13 · Tuesday, October 27, 2026 · By Scott Olsen · 🎃

PULL UP A CHAIR

It is hard to believe October is almost over, Halloween is right around the corner, and we roll our clocks back an hour this weekend too, Daylight Saving Time ends Sunday, November 1 this year, so set those clocks back Saturday night before bed. Hope everyone has an amazing Halloween, and here's hoping everyone's costumes are as spectacular as I know they will be.

Being able to actually communicate with the people walking into your office, or calling it, should be a real priority for all of us. I can tell you from my own time at the clerk's office, our Spanish-speaking staff were some of the most sought-after people in the building for exactly this reason, and on a day when nobody with that skill happened to be available, it genuinely changed how someone's visit went. Take a look at the suggestions below, one of them might be worth trying in your own office.

Hope everyone has an awesome week. Feel free to send me your costume photos, I thought about running an actual contest this year and decided that was more work than I wanted to take on, but I genuinely appreciate anyone who puts real effort into a costume.

Since it is the season for a scary story, I did not even have to go looking for one this year. Scroll down to the Rulebook, a county's own chatbot gave residents the wrong information about their car tags and property taxes, and the county had to publicly tell everyone not to trust it. No embellishment needed, that one is scary enough on its own.

THIS WEEK'S PICK

Serving non-English speakers without hiring interpreters

Hiring an interpreter for every meeting and translating every notice by hand is not realistic for most small offices. AI translation tools have gotten good enough to genuinely help, Google Translate, DeepL, and Wordly for live meetings. The tools are the easy part this week. The part worth real attention is a federal compliance picture that shifted in the last year and has not fully settled, which matters more here than which tool reads more naturally.

START HERE  THE FREE TOOL COVERS MORE LANGUAGES THAN THE PAID ONE

Google Translate, free and with no login required, is the right starting point for most offices

It is free, needs no login, and covers 130-plus languages, including Vietnamese, Tagalog, Somali, Haitian Creole, Arabic, and Chinese. For a routine notice or a web page, start here before paying for anything. It does not do human-level quality control, so do not reach for it on anything legally sensitive without a second look from a real person.

IF YOUR MEETINGS NEED LIVE TRANSLATION, NOT JUST WRITTEN NOTICES

Wordly is real, live AI meeting translation, and at least one small city has actually used it

Confirmed directly on the vendor's site: live translation and captioning in 60-plus languages, attendees join by QR code with no app to install, it works in person, hybrid, or fully virtual, and it can push captions straight to cable or PEG broadcast. Hillsboro, Oregon launched two-way live translation at a city council meeting in February 2025, residents testify in their own language and the council hears English, confirmed on the city's own site, not just Wordly's marketing. A few other named government customers only show up as logos on Wordly's own page, so treat those as vendor-claimed rather than independently confirmed. Pricing is published, not quote-only, starting around $1,500 a year for a 10-hour package, with volume and nonprofit discounts above that.

Cost: published hour-based packages, roughly $1,500/year and up     Setup: request a demo through wordly.ai

IF YOU WANT THE MOST NATURAL-SOUNDING WRITTEN TRANSLATION

DeepL reads more naturally, for a few dozen languages, not the 130-plus Google covers

DeepL's reputation for more natural phrasing holds up, weakly, for major European language pairs like Spanish, French, and Portuguese. It does not hold up as a universal claim, because DeepL simply does not support many of the languages a US local government actually needs. Somali, for one, does not appear on DeepL's current language list. Before you pick a tool, check DeepL's actual current language list against what your own community needs, not against an assumption, and if your biggest need is a language DeepL does not cover, Google Translate becomes the real choice regardless of which one sounds smoother. DeepL Pro plans run roughly €7.49 a month for an individual up to €49.99 per user for a business plan; API pricing for a document or website integration specifically could not be confirmed directly, so do not trust a number for that without checking DeepL's own page first.

Cost: Pro plans roughly €7.49 to €49.99/user/month; API pricing unconfirmed     Setup: check language coverage for your specific community first, at deepl.com

SKIP IT FOR NOW IF

You are treating any of these as a finished compliance solution with no human checking the output

None of these tools are a substitute for a person reviewing anything that affects someone's legal rights, a hearing notice, an eligibility decision, a legal deadline. Use AI translation to get most of the way there fast, then have a real person check anything high-stakes before it goes out. See the honest catch below for why this matters more than usual right now.

HOW TO ACTUALLY USE IT

1.  Find out which languages your actual community needs before picking a tool. Census data and your own front-counter staff both know this better than any vendor's marketing page, and it may rule out DeepL immediately.

2.  Use the free Google Translate widget for routine, low-stakes notices first. Save a paid tool for recurring or high-volume needs where the free version is not good enough.

3.  If you are considering live meeting translation, test it on a real, low-stakes meeting before you trust it for a contentious public hearing.

4.  Build a human spot-check into anything legally significant. Machine translation getting most of a sentence right and getting a deadline or a legal term wrong both look identical on the page.

The honest catch: Title VI of the Civil Rights Act still requires most federally funded local governments to provide meaningful access for residents with limited English proficiency, LEP for short, that statute has not changed. What has changed is federal enforcement posture. In July 2025, the Attorney General issued a memo rescinding the Justice Department's longstanding language-access guidance, suspending the federal LEP.gov resource site, and specifically endorsing AI translation as a cost-saving tool, while also arguing language proficiency alone is not a protected class under Title VI absent deliberate discrimination. That memo promised follow-up guidance within 180 days, which lands around January 2026. That guidance is now months overdue, and I found no DOJ notice, Federal Register filing, or news report from anywhere in 2026 confirming it was ever published. Civil rights advocates dispute the memo's legal theory entirely and say the underlying statute still requires access regardless. None of this is settled. Do not let anyone, a vendor or otherwise, tell you AI translation alone checks the Title VI box for your office, and do not assume the opposite either. Confirm your office's current obligations with your own counsel before this becomes your whole language-access plan.

THE RULEBOOK · TWO RULES WORTH A GLANCE

What changed that affects how your office can use AI. Short, plain, and sourced. The part that keeps you out of the newspaper. Tagged Local for a city or county action, State for a change at the state level, Federal for a change at the federal level that reaches your desk. Both items below are a bit older than the usual window, a thin news week for brand-new state and local AI action, but both are new to this newsletter and worth knowing.

State

Colorado's own municipal risk pool is telling cities not to assume they're exempt from the state's new AI laws

Colorado's SB 26-189 (covering automated decision-making technology in "consequential decisions," including essential government services) and HB 26-1263 (covering operators of public-facing conversational AI, meaning chatbots) both take effect January 1, 2027. CIRSA, the risk-sharing pool that insures most Colorado cities, counties, and special districts, published guidance telling municipalities they "must understand where these Acts' requirements may apply" to their own chatbots and automated tools. Notably, CIRSA does not say outright that these laws cover local government, because that is genuinely unresolved in the bills themselves, it tells municipalities to go find out before the deadline rather than assume either way.

What it means for you: if your office runs a resident-facing chatbot or uses any automated tool to help decide things like benefits, permits, or housing assistance, Colorado offices should not assume being a government exempts them. If you are outside Colorado, this is still a preview of the kind of question your own state may raise next.

Source: CIRSA, "Automated Decision-Making and Municipal Chatbots" →

Local

This week's real AI horror story, no embellishment needed for the holiday

Hall County, Georgia's new AI chatbot repeatedly gave residents incomplete or inaccurate information on motor vehicle tag and title requirements and property tax processes, confirmed directly on the Gainesville Times, the local paper of record. The county's own Tax Commissioner's Office had to issue a public warning telling residents not to rely on the bot and to contact staff directly by phone, text, email, or in person instead. This dates to reporting from earlier this year, older than our usual window, but it is real, it is verifiable, and it fits the week.

What it means for you: test any resident-facing chatbot against real questions you already know the right answer to before it goes live, and keep an obvious, easy human fallback visible the whole time, not buried in a footer.

Source: Gainesville Times, Hall County Tax Commissioner's Office warning →

ONE MORE QUICK WIN

Know who owns this decision, and write down why

Federal guidance here is genuinely unsettled, and may stay that way a while. That is exactly why your office should not be running on an unwritten assumption. Find out who is actually responsible for your language-access decisions, and whether your current approach, AI translation, a human reviewer, or both, is written down anywhere with a reason behind it. If the honest answer is "nobody" and "no," that gap is the real win to fix this week, regardless of what Washington eventually publishes.

Know someone at another city or county who could use this? Forward it over.

Subscribe free

Until next week, 🎃

Scott Olsen, CGCIO

Former CIO with eleven years in Florida local government

Follow on X

Practical guidance for local government, not legal or procurement advice.