Local data, planning and privacy

Why data matters for local planning

Every local office holds more data than it uses: complaint registers, water and sanitation connection records, property and tax rolls, birth and death registrations, market licences, inspection reports and development scheme files. Used well, this information can show where services are failing, which areas are growing and where money will do most good.

AI can make that data easier to work with. It can help you clean a messy spreadsheet, count and group entries, describe trends in plain words and draft a planning note. But planning on bad data produces confident, wrong plans and careless handling of residents' records can harm the people the plan is meant to serve.

What AI can help with

• Cleaning and combining. Turning handwritten or inconsistent records into a consistent table: standard ward names, dates in one format, duplicate entries marked. The how-to guide on AI for spreadsheets and data covers the mechanics.
• Counting and grouping. Complaints per ward per month, connections by area, applications by type and status.
• Describing patterns. Asking the tool to describe what a table shows, in plain language, as a starting point for your own reading.
• Drafting. A first version of a planning note or a briefing for a council meeting, built from figures you have already checked.

Notice what is not on the list: deciding which area deserves a new tube well or which scheme gets funded. Those are judgements about fairness and priorities. They belong to officers and elected representatives, not to a model.

Is the data good enough?

Before you plan with any dataset, ask:

1. Is it complete? Records from some wards or some months may be missing. A ward with few complaints may have good services or may have residents who have given up complaining.
2. Is it current? Old property rolls or connection lists may not reflect new settlements, especially informal ones.
3. Is it consistent? The same place may be spelt three ways. The same problem may be recorded under different categories by different staff.
4. Who is missing? Data reflects who can reach your office. Women, older residents, people with disabilities and those in informal settlements are often under-counted. Plans built only on recorded demand can send more resources to areas that are already better served.

AI does not fix these problems and can hide them. A neat chart drawn from incomplete data looks just as convincing as one drawn from complete data.

Checking AI's figures

When AI counts, groups or summarises data, it can drop rows, misread numbers or invent totals. Before any figure goes into a plan or a brief:

• add up columns yourself and compare with the source
• spot-check a few rows against the original registers
• ask the tool to show its working and check the steps
• confirm units and time periods

The check figures in a table prompt helps, but your own spot-check is the real safeguard.

Privacy and residents' data

Many local records contain personal data: names, CNIC numbers, addresses, phone numbers, property details and family information. Treat it with care.

• Use the least data you need. For planning, you usually need counts by area, not names. This is data minimisation.
• Remove identifiers before analysis. Strip names, numbers and exact addresses. Be aware that small areas or unusual details can still point to a person; this is re-identification.
• Use approved tools only. Do not upload registers containing personal data to a public AI tool. Follow your department's instructions on data protection and ask your IT focal person if unsure.
• Keep a record. Note which data was used, how it was cleaned and what was removed, so others can check your work.

Try it

Take a public table relevant to local services, such as figures from a published district or municipal report. Ask an AI assistant to describe the three most notable patterns. Then check each pattern against the table yourself and write down one question about who might be missing from the data.

Check yourself

1. Give two reasons a ward with few recorded complaints might not have good services.
2. What does data minimisation mean for planning work?
3. Name two checks to make before an AI-produced figure goes into a plan.

Answers: (1) Records may be missing for that ward or residents may have stopped complaining or may be unable to reach the office. (2) Using only the data you need, usually counts by area rather than names or personal details. (3) Any two of: adding up columns, spot-checking rows against the original, asking the tool to show its working, confirming units and time periods.