Skip to content
Citizen Infrastructure

Bowling Green's What Could BG Be? (2025)

TL;DR. In early 2025, Bowling Green, Kentucky — the fastest-growing city in the state, with about 75,000 residents and a population expected to double within 25 years — ran a month-long online consultation, “What Could BG Be?”, to feed a 25-year strategic plan called the BG 2050 Project. Over 33 days, 7,890 residents posted ideas and cast 1,034,868 votes on the open-source platform Polis. Google’s Jigsaw then used its Sensemaker tool, running on Gemini, to sort the ideas into 12 themes and 72 subtopics and to surface where the city agreed and where it split. It is a concrete example of pairing mass online input with AI sensemaking at municipal scale — while keeping humans in the loop to read the results.

At a glance

Section titled “At a glance”
When14 February – 17 March 2025 (33 days); recommendations to county leadership expected late 2025
WhereBowling Green & Warren County, Kentucky (~75,000 residents)
WhoOpen to all residents — 7,890 took part. Run by Innovation Engine (a local strategy firm; co-founding partner Sam Ford) for the BG 2050 Project, sponsored by Warren County (judge-executive Doug Gorman)
What they didPosted and voted on ideas for the city’s future on Polis (open-source, by The Computational Democracy Project); 1,034,868 votes cast
The AIJigsaw’s Sensemaker (using Gemini) categorised every idea into 12 topics / 72 subtopics, generated the public report’s summaries, and flagged high-agreement and divisive ideas
OutcomeA public report (whatcouldbgbe.com); BG 2050 volunteers will turn it into recommendations to Warren County leadership for the long-range plan

What happened

Section titled “What happened”

Bowling Green is the fastest-growing city in Kentucky, and its leaders expect the population — around 75,000 — to double over the next 25 years. To plan for that, the BG 2050 Project wanted public input at a scale its roughly 100-person volunteer leadership team could not gather on its own. Working with Innovation Engine, a Bowling Green strategy firm, and Google’s Jigsaw, the project ran an open online conversation, “What Could BG Be?”, on Polis — the open-source platform built by The Computational Democracy Project, where participants submit short statements and vote to agree or disagree with others’.

Over 33 days, from 14 February to 17 March 2025, 7,890 residents weighed in 1,034,868 times. Rather than read every statement by hand, the organisers ran the input through Jigsaw’s Sensemaker, an AI tool built on Google’s Gemini, to categorise the ideas into 12 high-level topics and 72 subtopics — arts and culture, transportation, economic development, housing, education, public safety, environment, and more — and to write the summaries in the resulting report. Sensemaker also separated where residents converged from where they split: by the Jigsaw write-up’s account, 2,370 ideas, over half of them, drew agreement rates above 80 percent. The report names seven areas of strong consensus, among them more family-friendly activities, grocery access, connecting downtown to the riverfront, walkable sidewalks, easing traffic on Scottsville Road, and flood and disaster preparedness.

The organisers are explicit that the AI did not replace human judgement. Jigsaw’s account stresses that “having a human in the loop when reviewing civic input is still the best path for unlocking the most actionable insights” — local knowledge was needed to interpret the output, for instance to connect riverfront ideas to investments the county was already weighing. What the AI bought was speed: an analysis that might have taken days or weeks was done in minutes, which is what made acting on a million votes feasible at all. Innovation Engine’s post-conversation surveys reported that 70% of participants came away more confident their voice matters and 83% felt they better understood other residents’ views, though those figures are self-reported. The findings now feed the formal BG 2050 planning, with volunteers due to bring recommendations to Warren County leadership in late 2025.

Why it matters

Section titled “Why it matters”

Most of the AI-and-democracy run reports here use AI to help a recruited or sortition group deliberate. Bowling Green is a different shape: an open, mass online consultation where the hard problem is not recruiting a representative few but making sense of input from thousands. That is the gap civic listening names, and it is where AI is doing the most practical work right now — scaling the read side of deliberation rather than the talk side. The toolchain is a recognisable lineage too: Polis is the same platform behind Taiwan’s vTaiwan and its Alignment Assemblies and California’s Engaged California, now paired with an LLM to do clustering a human team once did by hand. The honest caveat the organisers volunteer — that the AI summaries still needed local people to interpret them — is the part worth carrying forward: the technology made a million votes legible, not self-explanatory.

Sources

Section titled “Sources”