AI use case

AI damp and mould risk prediction for social housing

AI damp and mould risk prediction is a machine learning model that scores homes by the likelihood of damp and mould, using repair history, property data and sometimes weather or sensor data. A social landlord inspects and fixes the highest risk homes before a tenant reports a problem or is harmed.

Last verified 18 November 2026 · 2 public deployments

98%
Reported accuracy
Wolverhampton Homes, organization claim.
GBP 32,000 to GBP 180,000
Indicative value per year
A housing association with 20,000 homes. Worked example, see how it is calculated.

What problem does it solve?

Damp and mould in social housing is mostly found when a tenant reports it, and reports can be missed or slow. Wolverhampton Homes describes the usual pattern: issues "are typically dealt with after a report is received from the tenant". Greatwell Homes, a provider with just over 5,000 homes, says that with many homes to look after "things can slip through the net".

In England the Housing Ombudsman has pushed landlords toward a more proactive approach, and Awaab's Law adds legal time limits for social landlords to investigate and fix hazards. A landlord that can rank its homes by risk can direct inspections and retrofit works to where they are needed first.

How does it work?

  1. Assemble the data. Wolverhampton Homes used the condition of properties, maintenance history, whether damp and mould complaints had been received before, and environmental factors such as local flooding and weather patterns.
  2. Score each home. A model estimates the risk of damp and mould for each property. Greatwell Homes' software screens the severity and geographic risk of every repair raised, and flags other flats in a block when several have had an issue.
  3. Update continuously. When housing officers record an update about a home or resident, Wolverhampton Homes' model inside its housing system adjusts the risk level and moves the household up the repair priority list.
  4. Act. The landlord inspects or repairs high risk homes, adds ventilation or prioritizes retrofit works, instead of waiting for a complaint.
  5. Keep the data clean. Wolverhampton Homes found that accuracy fell from 98 percent to about 70 percent when the data was out of date or incomplete, so record quality is part of the system.
Audience
Back office
Autonomy
Copilot
Adoption
Early adopters
Channels
Internal tools

What is it worth?

Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.

Value benchmarks for AI damp and mould risk prediction for social housing
KPIMedianReported rangeData pointsClaimed by
AccuracyToo few to pool
98%
11 organization

Value drivers: Risk and loss reduction, Compliance quality, Customer experience.

Indicative value

A housing association with 20,000 homes

GBP 32,000 to GBP 180,000

Annual cost of proactive inspections, the budget the model directs per year

How this is calculated

Formula: homes * shareFlagged * costPerVisit. The low scenario uses every low input, the high scenario every high input.

InputLowHighBasis
Homes in the portfolio homes, homes20,00020,000Reference size, close to Wolverhampton Homes' 21,000 homes. Replace with your own stock.
Share of homes flagged for a proactive visit each year shareFlagged, share0.020.06Editorial assumption, replace with your own. No source on this page gives it.
Cost of one inspection visit costPerVisit, GBP per visit80150Editorial assumption, replace with your own inspection cost.

What it leaves out: This shows the inspection spend the model would direct, not a saving. Avoided emergency repairs, complaints and health harm are not valued, and no organization on this page has published a cost outcome.

Which organizations already do this?

2 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.

This library is not a list of Blits.ai customers. It documents public AI deployments from any vendor, each with its source.

Wolverhampton Homes

United Kingdom · Real estate · 2023

PilotGrade B

Wolverhampton Homes, the arm's length management organization for Wolverhampton's council homes, piloted an AI model in 2023 that predicts which of its 21,000 homes are at risk of damp and mould. The pilot analyzed data from thousands of homes: property condition, maintenance history, earlier damp and mould complaints and environmental factors such as local flooding and weather.

A systems manager at the landlord reports 98 percent accuracy, falling to about 70 percent when the data was out of date or incomplete. The model sits inside NEC Housing and moves a household up the repair priority list when its risk level changes. The landlord says the AI has helped it identify damp and mould risk across about two thirds of its homes, and that it uses the output to fix problems earlier, prioritize retrofit works and add ventilation.

  • Accuracy: 98%, 2023 pilot, predicting the presence of damp and mould in high risk properties; about 70 percent when the data was out of date or incomplete
    "We found that the AI technology was able to successfully predict the presence of damp and mould in high-risk properties with 98% accuracy."
    Claimed by: organization

Greatwell Homes

United Kingdom · Real estate · 2024

ProductionGrade C

Greatwell Homes, a housing provider with just over 5,000 homes in Wellingborough and Northampton, added Mobysoft's RepairSense in June 2024. The software uses machine learning on the existing repairs and property data to screen the severity and geographic risk of damp and mould cases, so that when several flats in a block report an issue it flags the others as possibly at risk. It also looks back through a block's history and follows up cancelled or unanswered visits.

The provider says the share of households with a damp and mould case fell from 16 percent in June 2024 to 2 per cent in 2025 without changing its stock or hiring. The figure is self reported, and the article does not say how much of the fall the software caused.

No outcome disclosed.

How do you implement it?

A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.

Data you need

  • Repair history and earlier damp and mould cases per property
  • Property attributes such as age, construction, heating and ventilation
  • Tenant and vulnerability records handled under data protection law

Systems to integrate

  • Housing management system, where the risk score is written back
  • Repairs and inspection scheduling
  • Optional environmental or sensor feeds

Complexity: Medium

The build needs a housing management system with several years of repair and complaint records, a property attribute table, a data quality routine and an inspection process able to take the extra visits.

  1. 1

    Audit the data first

    Check the completeness and age of repair and property records. Wolverhampton Homes found accuracy fell to about 70 percent on incomplete data, so fix the records before relying on scores.

  2. 2

    Pilot on a part of the stock and compare with known cases

    Test the model against homes where damp and mould is already confirmed and report the false positive rate, which the sources on this page do not give.

  3. 3

    Write the score into the repair process

    Make the risk level a field the repairs team sees, with a rule that high risk homes get a visit within a set time.

  4. 4

    Keep a person in the decision

    An officer reviews each flagged home before a tenant is contacted or works are ordered.

  5. 5

    Retrain and recheck on a schedule

    Recalculate accuracy against inspection results each quarter and after any change to the data feed.

Guardrails

  • A model score never replaces a tenant report; every report is still investigated
  • Vulnerability data is used only for prioritization and is access controlled
  • Accuracy is checked against inspection outcomes and reported to the board
  • Tenants can still reach a person about their home at any time

KPIs to instrument

  • Share of confirmed damp and mould cases that the model had flagged beforehand
  • Share of flagged homes that inspection confirms
  • Days from flag to inspection and to repair
  • Share of households with an open damp and mould case

Human in the loop

Housing officers decide what to do about a flagged home. Wolverhampton Homes uses the score to prioritize homes for repair and retrofit, and tenant reports keep their own route.

Common failure modes

Stale data produces confident wrong scores
Wolverhampton Homes saw accuracy drop to about 70 percent on incomplete data. Monitor record freshness per property.
Reported accuracy hides a high false positive rate
A model that flags most homes gets high recall and wastes inspection budget. Track precision as well as accuracy.
Score treated as the legal safeguard
A landlord still has duties on every report received. Keep the report route independent of the model.

What are the risks and rules?

EU AI Act

Depends on design

Annex III point 5 lists AI used by public authorities to evaluate the eligibility of people for essential public services and benefits, and AI that dispatches or prioritizes the dispatch of emergency first response services. A repair prioritization model for a council owned or housing association stock is usually minimal risk when it ranks homes for inspection. It moves toward high risk if the score is used to decide who receives a service, or if tenant characteristics drive the ranking in a way that affects access. The design decides.

Rules that apply

Guidance

  • Awaab's Law, guidance for social landlords (UK Government, Europe). Sets the timeframes in which social landlords in England must investigate and address damp and mould hazards.
  • Spotlight on damp and mould (Housing Ombudsman Service, Europe). The Ombudsman's report on how landlords handle damp and mould, which sets the expectation of a proactive approach.

Controls to put in place

  • Data quality checks on every input feed
  • Role based access to vulnerability and health information
  • Quarterly comparison of scores with inspection results

Frequently asked questions

How accurate are damp and mould prediction models?

Wolverhampton Homes reports 98 percent accuracy in a pilot, falling to about 70 percent when the data was out of date or incomplete. That is the landlord's own claim, and it does not say how accuracy was defined or what the false positive rate was.

What data does the model need?

Wolverhampton Homes used property condition, maintenance history, earlier complaints and environmental factors such as flooding and weather. Greatwell Homes uses its repair data from the last years. Sensor data is optional.

Has it reduced damp and mould cases?

Greatwell Homes says the share of households with a damp and mould case fell from 16 percent in June 2024 to 2 percent in 2025, after adding RepairSense. The figure is self reported and the article does not separate the software's effect from other work.

How to cite this page

Blits.ai AI Use Case Library, "AI damp and mould risk prediction for social housing", last verified 18 November 2026, https://www.blits.ai/ai-use-cases/damp-and-mould-risk-prediction-for-social-housing. Licensed under CC BY 4.0. Method: how we verify use cases.

Changelog
  • 18 November 2026: First published

Related use cases

Real estate

AI agent for apartment leasing inquiries and resident service

An AI leasing and resident service agent answers rental prospects and residents by chat, text, email and phone for a property manager. It answers questions about apartments and policies, books tours, takes maintenance requests, sends renewal and payment reminders, and hands anything that needs judgment to leasing or service staff.

Deployments
15 public, best grade B
Median automation rate, vendor reported
90%
3 deployments
Real estateRetail and ecommerce

AI for building HVAC and energy optimization

Building energy optimization is AI that continuously predicts a building's heating, cooling and ventilation needs and adjusts its systems in real time to use less energy. It changes setpoints, equipment sequencing and start times through the existing building management system, instead of following fixed schedules, so no person has to retune it by hand.

Deployments
9 public, best grade C
Median energy savings, vendor reported
15.8%
5 deployments
Government and public sectorLogistics and transportation

AI road surface defect detection from vehicle cameras

AI road surface defect detection is computer vision on cameras mounted on council vehicles that finds, locates and rates potholes, cracks and damaged signs during normal rounds. Maintenance is then planned from a regular condition survey of the whole network instead of from resident reports and occasional manual inspections.

Deployments
3 public, best grade B
Autonomy
Copilot
Telecommunications

AI copilot for field technicians and dispatch optimization

A field technician copilot is AI for a telecom operator's field service: it decides whether a network or customer fault needs a site visit at all, predicts the work and parts the job will need, helps plan and update appointments, and gives technicians on site guided diagnosis and answers from manuals and past jobs, so more jobs are fixed on the first visit. Assistants for the people who run and repair machines inside plants, workshops and service centers are plant operator and maintenance copilots.

Deployments
5 public, best grade B
Reported error reduction
28%
Vodafone, organization claim