How did you choose the products?
We included two main streams of vendors. The first is popular workplace software — tools that typically handle task management or inquiry tracking. The second is governance and civic-tech tools, many of them open source and far less widely known.
We believe these are two sides of the same coin and ought to be considered together — if not bundled into the same product. Where a vendor offers a broad suite (Granicus, for example), we evaluate the suite as a whole rather than picking apart individual modules.
We excluded narrowly niche products, such as those built solely for electronic voting or participatory budgeting. We also limited our scope to vendors whose websites are available in English.
How is the analysis performed?
The analysis runs on the Ordinizer framework, built around our own Participatory Governance Framework (PGF), which maps the distinct functions governance software can perform. For each function we wrote a set of questions to put to the data.
For each vendor, we pull and save the relevant pages from their website, load them into a vector database, and then run a retrieval-augmented (RAG) analysis that answers each question against that content.
We currently use Claude Sonnet 4.6, and we're experimenting with newer (and more costly) models.
Why use AI for this work?
Our aim is to be objective: what would a typical reviewer see when evaluating this software? Plenty of people already use AI to ask questions like "How does product X support feature Y?" We've simply done that at scale — across 33 products, six main feature areas, with roughly 7–10 detailed questions each — and turned the results into comparable charts.
Can you trust AI for this?
There are real pitfalls. The analysis is only as good as its input data, which raises two questions: do we have the right pages from each product's website, and do those websites actually present everything we need? We assume they do, though some sites we've encountered are poorly designed for machine reading.
How often is the data updated?
The current dataset reflects what we scanned as of July 2026. We plan to update it at least quarterly, and we're exploring options for more frequent updates over time.
How can I update data in the scorecard?
Email us. The more durable solution is to make sure the information lives on your own site and is clearly readable — both to people and to machines.
Will you add subjective reviewers?
We're looking at ways to layer subjective analysis on top of the generated ratings. The People Powered platform ratings cover many of the same products and are subjective. One difference from our approach: their model consolidates all functionality into a single score, whereas we keep the functions separate.
What are the broader goals of the scorecard?
First and foremost, we want to help buyers. If buyers and vendors share a common language for these functions, buyers can specify their needs far more clearly.
We'd also like to inspire some purposeful competition. We hope niche players look into expanding their offerings, and that commercial vendors consider pricing models that work for volunteer organizations — whose members might use the software eight hours a month rather than eight hours a day like a typical commercial user.