Northern IrelandAIPublic Relations

Summarising ABC Council Consultation Responses With AI, Carefully

A careful method for summarising ABC Council consultation responses with AI: preparing data, checking themes, representing every town fairly and keeping people accountable.

Geoffrey Moffett 4 Min Read

Quick Answer

Summarising ABC Council consultation responses with AI can save hours, but only with a careful method: remove personal data first, ask the AI to group themes, check those themes against a sample of real responses, look for differences between towns and groups, and have a person write and sign off the final summary.

Key Takeaways

  • AI can speed up theme-finding in large consultations, but people must check it.
  • Remove personal data before using any AI tool.
  • In a borough of many towns, check whether views differ by place.
  • Be open about how AI was used, and review the method as tools change.
In This Article

Big consultations can produce hundreds or thousands of free-text responses. Reading them all properly takes days. Summarising ABC Council consultation responses with AI can shorten that work dramatically, but a consultation is a promise to listen, so the method matters as much as the speed.

This guide sets out a careful approach that a council like Armagh City, Banbridge and Craigavon could use in 2027. It's general, not legal advice, and it doesn't describe any council's current practice. AI tools are changing quickly, so expect to revisit the method at least once a year.

Why Summarising ABC Council Consultation Responses Needs Extra Care

A borough with three main hubs and many villages raises a specific risk. A summary that says "most respondents supported the proposal" could hide the fact that people in one town felt very differently. AI tools, left alone, tend to flatten responses into a single overall picture.

There are other risks too:

  • AI may over-weight frequently repeated phrases, such as campaign template responses;
  • it can miss sarcasm, local references or strong feeling;
  • it occasionally invents a theme or misquotes;
  • personal data in responses needs protecting.

A Seven-Step Method

Step 1: Plan Before Responses Arrive

Decide in advance how you'll analyse responses. If you want to compare places, ask respondents for a general area (for example, which town or district they live nearest) rather than full addresses. Note in your privacy information how responses will be processed.

Step 2: Prepare The Data

Export responses into a spreadsheet. Remove names, email addresses, phone numbers and any personal details people have typed into free-text boxes. Keep a reference number for each response so you can trace back.

Step 3: Read A Sample Yourself First

Before using AI, two officers read a random sample of perhaps fifty responses and note the themes they see. This gives you a human baseline to compare against.

Step 4: Ask AI To Propose Themes

Using an approved tool, ask the AI to propose themes, with a short description and the reference numbers of responses that fit each one. Ask it to flag responses that don't fit any theme.

Step 5: Check The Themes

Compare the AI's themes with your human baseline. Then spot-check: pick a few reference numbers for each theme and read them. Do they really say what the AI claims? Adjust themes as needed.

Step 6: Look For Differences

Ask for the themes broken down by area or group where you have that information. Do views in Armagh differ from Portadown or Banbridge? Do younger respondents raise different issues? Check any differences by reading responses directly.

Step 7: People Write The Summary

Officers write the final summary, using the AI's grouping as a tool, not a script. Include direct anonymised quotes that represent each theme fairly, including minority views.

Handling Campaign Responses

Sometimes many people submit the same or similar wording. That's legitimate participation. AI can help identify near-identical responses so you can report them clearly ("around a third of responses used similar wording supporting…") without letting them drown out other views.

Paper And Event Responses

Not every response arrives online. Paper forms, letters, notes from drop-in sessions in village halls and comments made at events all count. Type them up consistently, mark their source and include them in the same dataset, so the AI's theme-finding doesn't quietly favour people who replied online. Check afterwards that themes from paper and event responses appear in the summary. In a borough with a large rural population, this step matters.

Keeping An Audit Trail

Keep a short record of what you did: which tool, which prompts, which version of the data, who checked what and what changed after checking. If anyone later asks how a theme was identified, you can show your working. It also makes the next consultation faster, because the method is written down.

Being Open About AI

Add a short methodology note to the published report: AI was used to help group responses into themes; officers checked the themes against the original responses; officers wrote the summary. That's honest and reassuring.

A Prompt To Start With

"Below are anonymised consultation responses, each with a reference number. Propose up to ten themes. For each theme, give a one-sentence description and list the reference numbers that fit. List any responses that don't fit a theme separately. Don't invent quotes or themes that aren't clearly supported."

A Made-Up Example

Imagine a consultation on a new play park design that gets 900 responses. AI proposes eight themes in minutes. Officers' spot checks show one theme, "parking concerns", is mainly from residents near one site, which matters for that location. A second theme is merged because it overlapped. The final report is written by officers in two days rather than two weeks, with a clear note on method.

A Quick Checklist

  • Analysis method decided before launch
  • Personal data removed
  • Human sample read first
  • Themes proposed by AI and checked
  • Differences by area and group examined
  • Summary written by officers with representative quotes
  • Methodology note published
  • Method reviewed as tools change through 2027

Pitfalls To Avoid

  • Copying the AI summary straight into the report. Always check and rewrite.
  • Ignoring minority views. They can be the most important.
  • Losing the place. In this borough, location often explains disagreement.
  • Skipping the privacy step. Remove personal data first.

How Starien Can Help

Our AI Coaching helps council teams build practical, careful AI workflows, including how to check AI output. Our Public Relations support helps turn findings into clear, honest communication. We've worked with public sector teams, and our six years leading Coleraine BID's digital delivery included plenty of listening to businesses and residents.

We also help local businesses across Armagh, Banbridge and Craigavon understand their own customer feedback and turn it into better marketing. Get in touch if that would be useful.

Frequently Asked Questions

Can AI read and summarise thousands of consultation responses?

It can help group and summarise large volumes of text quickly, but it can miss nuance, over-weight common phrases and occasionally misrepresent views. Treat its output as a first pass for people to check, not a final report.

Is it safe to put consultation responses into AI tools?

Only with tools the council has assessed as suitable for that data, and ideally after removing names, addresses and other personal details. This isn't legal advice; check with your data protection lead.

Should we tell people AI was used?

Yes. A short note explaining that AI helped group themes, and that officers checked and wrote the final summary, supports trust.

Written By

Geoffrey Moffett

More Articles

Back To Top