Read in samples.
Under time pressure, teams read a share of responses and estimate the rest.
What are the responses actually saying?
Every response read, not a sample, so the summary reflects what people actually said and every point can be traced back to the words that made it.
See it worked throughThousands of responses.
Weeks to read them.
A public consultation can bring in thousands of responses: letters, emails, forms and long reports. Each one deserves to be read.
Campaigns count twice.
Hundreds of near-identical letters can drown out one detailed reply with new evidence.
Hard to show your working.
When the summary is challenged, finding the responses behind each point takes days.
What the AI reads.
Every response read against the consultation questions, whatever form it came in.
In
- Online form responses
- Emails and letters
- Reports from organisations
- Scanned handwritten replies
- The consultation questions
- Earlier consultations
At work
- Sense Reads forms, emails, scanned letters and long reports.
- Understand Finds the arguments in each response and groups them by question.
- Check Flags organised campaigns and near-duplicate responses, and keeps them countable.
- Operate Prepares a theme list with quotes and counts for officials to write from.
Out
Every argument grouped, counted and quoted. Officials write the summary.
Every response read.
Every theme quoted.
The policy team sees what respondents argued, how many said it, and the words they used, before writing a line.
1,842 responses · 5 themes · 1 organised campaign
| Theme | Argument | Finding | Detail | Responses |
|---|---|---|---|---|
| T1 | Charges will hurt small shops | Widely raised | Quoted from 14 local businesses, with sales figures in 3 | 612 |
| T2 | Support for more cycle parking | Widely raised | Many cite safety near the school | 418 |
| T3 | Blue badge holders need closer spaces | Specific evidence | Detailed reply from a disability charity, with a survey | 96 |
| T4 | Pre-written letter opposing all changes | Campaign | Same wording in 540 replies; from one template | 540 |
| T5 | Evening charges unclear in the proposal | Question raised | Respondents read the times differently | 176 |
Campaign
Same wording in 540 replies; from one template.
T4 · Pre-written letter opposing all changesSpecific evidence
Detailed reply from a disability charity, with a survey.
T3 · Blue badge holders need closer spaces
Illustrative example. Council, consultation, respondents and figures are fictional.
Built for open policy-making.
The AI reads and groups. Officials weigh the arguments and write the summary.
Every response counts.
Nothing is dropped. Campaigns are flagged and counted, not removed.
How Check works ↗Every theme shows its source.
Quotes link back to the response they came from, so the summary can be checked.
How Understand works ↗Runs in your environment.
Responses, including respondents’ personal details, stay inside your boundary.
Deployment options ↗Same records.
Other questions.
The documents and data behind this work answer other questions too.
What must be redacted before release?
Records release
Find personal and sensitive details across a release, with every redaction shown for review.
See how →What does this new rule change for us?
Regulatory changes
Map new rules to what the business must change.
See how →Which cases can be decided today?
Casework backlog
Clear the routine cases fast, and send the hard ones to a caseworker.
See how →What personal data is in these files?
Sensitive data
Find it and remove it before anything is shared.
See how →Bring us the responses to a past consultation.
We’ll show you the themes, the quotes behind them and any campaigns, and you can compare them with your published summary.
Talk to us →LondonDiscovery AI Limited · United Kingdom
HyderabadReDscvry Technology Private Limited · India
ChennaiReDscvry Technology Private Limited · India