Back

Sentiment analysis AI makes big savings for local authority housing team

With manual processes prevailing, a local authority housing team was struggling to prioritise and resolve cases effectively, leading to high risk of undesirable outcomes for tenants.

Overview

Following service design workshops with the complaints team we identified triage and prioritisation as the biggest opportunities for efficiency gains. Building an AI sentiment analysis algorithm and integrating this with other key systems, we first proved the value of the solution with a pilot project before achieving full roll out, including staff training, in 12 weeks. The project reduced annualised costs by £250k, representing a year one ROI of 200%.

Challenge

The local authority’s complaint management function was overwhelmed by high volumes meaning critical complaints were buried under low-priority issues, causing delays in addressing serious cases.

  • Manual triage processes meant staff spent excessive time categorising and routing complaints.

  • A lack of real-time tracking made it difficult to monitor complaint statuses or identify escalation risks.

  • Without standardised processes, responses varied across departments, reducing customer satisfaction.

These inefficiencies increased the risk of Ombudsman escalations, reputational damage, and litigation. This also put strain on the customer support team, leading to low morale and high attrition.   

Solution

We implemented our proprietary four step process:

Phase 1. Assessment

We conducted several on-site service design workshops with the complaints team and analysed 12 months of historical complaint resolution data. This helped us map the existing state of the system, highlighting inefficiencies in triage, routing and tracking. We were then able to isolate the most valuable applications of automation and AI.

  • AI sentiment analysis and categorisation of complaints using Natural Language Processing (NLP).

  • Automated prioritisation of complaints at high-risk of escalation to the Ombudsman.

  • Granular automation of medium to low risks for routing into a tenant centric digital self-service triage.

We defined the following success metrics: 100% tenant responses within 24 hours, reduction in Ombudsman escalated complaints, reduced overall workload for the team and improved citizen satisfaction.

Phase 2. Prototype

Our deep dive assessment enabled us to create targeted concepts for proof of value.

  • We built an NLP a 'sentiment analysis' algorithm to categorise complaints based on safety risk, threat of legal escalation and customer vulnerability.

  • We coded integrations with existing digital triage and Customer Satisfaction survey tools.

  • We trained team members on training the algorithm to further increase efficiency.

These concepts were tested in small scale pilot projects with excellent results.

Phase 3. Scale

With proof of value successfully achieved, we scaled the new complaint management system complete with AI sentiment analysis and automations, across the existing Microsoft Azure stack. Key features included:

  • Automated routing of complaints to relevant teams and digital triage systems based on urgency and category.

  • Integration with existing systems for seamless data flow and compliance with regulatory standards.

  • A centralised dashboard showing key real-time metrics and projections.

The scaling process included training sessions for staff across all departments to ensure smooth adoption of the new system.

Full implementation was completed in only 10 weeks, covering the authority’s complaints team, and external partners handling resident complaints.

Phase 4. Measure

We monitored key metrics for six months post-implementation:

  • Efficiency gains helped to achieve 100% response rate within 24 hours.

  • AI-powered categorisation achieved a 94% accuracy rate in identifying high-risk complaints.

  • Escalations to formal complaints and litigation were reduced by 56%.

Overall, the project lowered operational costs by £250,000 annually due to reduced manual effort, faster resolutions and fewer housing condition claims. 

This represents a year one ROI of 200% and savings are projected to increase year-on-year.

Conclusion

This success story demonstrates four deep’s structured approach to delivering automation solutions that combine rapid assessment, development and proof of value followed by implementation at scale to deliver measurable ROI. In this case, AI driven sentiment analysis empowered the local authority to streamline its complaint management process while improving tenant satisfaction and reducing legal risk.

The Human Perspective

This project is a great example of technology and humans in balance. The solution saved our client £250k a year, which funded a new frontline service on the ground. We helped to train a number of the Customer Liaison team as they were redeployed to help vulnerable tenants with a more proactive service.