BIGGER PICTURE Public & everyday services
50,000 public responses became a map analysts could inspect
Consult helped a water-sector review organize a mountain of public submissions. Its useful role was to create a first pass that people could check, correct and use for deeper analysis.

- Water review responses
- >50,000
- Automated categorization
- About 2 hours
- Expert checking
- 22 hours
How the idea works
Collect the original responses
Free-text submissions enter the analysis workflow.
Map overlapping themes
The model groups concerns without forcing one label per answer.
Inspect and interpret
Analysts check labels and develop the final analysis.
Listening can become a sorting bottleneck
Public consultations collect free-text responses, not neat database rows. People may describe the same concern in different language, or raise several issues in one answer. For the Independent Water Commission, more than 50,000 responses created a substantial organizing task before policy experts could examine the concerns and evidence in depth.
Sources: [1]
One answer can belong to several themes
The government’s transparency record explains that Consult uses language models to generate themes and assign responses to them. An answer can receive zero, one or several theme assignments. The output gives analysts an initial thematic landscape, with human review and relabelling built into the process. The aim is to organize public input, not manufacture public opinion.
Sources: [2]
Two hours was not the whole job
The October 2025 account reports around two hours of automated categorization, costing£240, followed by 22 hours of expert checking. Detailed manual review of stakeholder submissions also continued. These distinctions matter: the reported machine time covers sorting, not the entire consultation or final report. Staff used the organized material to focus on interpretation and recommendations.
Sources: [1]
Preserve the route back to people’s words
Theme assignment involves judgement, and the earlier evaluation found imperfect agreement among both humans and AI. Speed therefore cannot substitute for checking whether unusual or minority views survive the categorization. Consult’s promising contribution is reducing the initial organizational burden while preserving an analyst-led process. Projected savings across government remain projections, separate from this documented application.
The original sources
- UK GovernmentConsult supports Independent Water Commission response analysis
- UK GovernmentDSIT: Consult algorithmic transparency record