Westminster City Council has slashed the time it takes to handle resident queries with a new Microsoft-based AI contact centre that effectively remembers every interaction, eliminating the need for citizens to repeat information across calls. The system, which went live in September 2025 after an accelerated discovery phase earlier that year, now processes thousands of queries monthly—and early results show dramatic reductions in average call lengths.
Residents contacting the council about anything from housing repairs to council tax no longer face frustrating, fragmented conversations. Instead, the AI—built on Microsoft’s cloud and AI stack—unifies each resident’s history, notes, and past interactions into a single view, ensuring agents and automated flows pick up exactly where the last exchange left off. The outcome is a contact centre that feels less like a series of disconnected calls and more like a continuous conversation.
The Problem: Repetition as a Resident Headache
Any resident who has ever phoned a local council knows the drill: after navigating a phone tree, you explain your issue, verify your identity, and recount any previous discussions. When the call transfers or you need to call back, you start from zero. These repetitive steps pad call times, raise frustration, and drain council resources. For a borough as busy as Westminster—which fields hundreds of thousands of contacts each year across housing, social care, waste, and more—that repetition adds up to thousands of wasted agent hours.
Before the AI overhaul, Westminster’s contact centre relied on a traditional model where each call existed in isolation. Even with a CRM, agents often had to manually piece together clues from different systems. The insight from discovery work in mid-2025 was blunt: if the council could stitch together resident context across every touchpoint, it could fundamentally reshape the experience and reclaim staff capacity.
The Solution: An AI That Never Forgets
Westminster’s answer is a Microsoft-powered contact centre that actively maintains what the council calls “resident context.” Whenever a citizen gets in touch—by voice, chat, or web form—the AI surfaces a complete picture of that person’s interactions with the council. This includes recent cases, outstanding requests, preferences, and even the specific language used in prior conversations.
Microsoft’s technology enables the system to understand intent and extract key details from unstructured chat or speech, then link those details to a persistent profile. If a resident called last week about a missed bin collection and today about a council tax discount, the system immediately connects both threads and presents them to the agent or the automated assistant without the resident uttering a word of recap. In many cases, the AI resolves straightforward queries end-to-end, and when an agent does step in, they are already briefed on the full situation.
How It Works Under the Hood
While Westminster hasn’t disclosed every technical component, the architecture leans on several well-known Microsoft offerings. The likely backbone is Dynamics 365 Contact Center, a modern platform that embeds Copilot AI directly into agent workflows. Copilot can listen in on calls, transcribe and summarize conversations in real time, suggest knowledge articles, and draft responses—all drawing on the resident’s unified profile.
Crucially, the platform integrates with Azure AI services for natural language understanding and with Power Platform to orchestrate automated processes. When a resident says “I still haven’t received my housing benefit payment,” the AI recognises the phrase, correlates it with an open case, and either triggers an update to the resident or routes the call to the right team with full context pre-loaded. By preserving context, the system eliminates the jarring moment of “Who did you speak to last time?” because the agent already knows.
By the Numbers: Thousands of Queries, Shorter Calls
Since going live in September 2025, the AI contact centre has handled thousands of resident queries every month. Although the council hasn’t released precise metrics, insiders familiar with similar deployments report that average handle times can drop by 20–40% when context preservation is implemented well. For a council the size of Westminster, that equates to hundreds of additional hours saved per month—hours that can be redirected to more complex, human-led support.
The implementation timeline was notably rapid. Discovery work occurred in mid-2025, and the system was built and switched on by September—a sign that the council leveraged pre-built Microsoft components and low-code tools rather than constructing a bespoke AI from scratch. That speed-to-value is likely to attract attention from other local authorities watching Westminster’s experiment.
What Residents Actually Experience
For residents, the change is tangible. Calling the council no longer means steeling oneself for a gauntlet of verification questions. The system recognises the caller’s phone number, matches it to an existing record, and surfaces recent activity. If the caller chooses a web chat, the AI assistant knows who they are from their authenticated session and can reference the same history.
Take a typical scenario: a resident has an open repair ticket for a leaky window. They first reported it by phone, then followed up via email, and now they’re calling to check progress. In the old world, the agent might spend three minutes searching different systems and asking, “Can you describe the problem again?” With the new AI, the agent sees the full timeline instantly and can say, “I can see we sent a contractor booking on Tuesday—have you received the appointment text?” That direct relevance not only cuts call time but also builds trust that the council is on top of things.
Even automated interactions feel more cohesive. A virtual agent can greet a returning resident with a message like, “I see you contacted us last week about your council tax. Is this about the same issue?” The resident can simply say yes, and the AI jumps straight into the relevant solution flow.
The Tech Stack: Microsoft’s Ecosystem in Local Government
Westminster’s choice of Microsoft is part of a wider trend among UK local authorities that are already embedded in the Microsoft 365 ecosystem. Many councils already use Teams for internal collaboration, Outlook for email, and SharePoint for document management. Plugging an AI contact centre into that same environment greatly simplifies integration with back-office systems.
Key Microsoft products likely in play:
- Dynamics 365 Contact Center: The core contact centre platform that unifies voice, chat, and digital channels with AI.
- Copilot Studio: Allows the council to build and customise conversational AI assistants without deep coding.
- Azure AI Language: Powers intent recognition and entity extraction from resident messages.
- Azure Cognitive Search: Indexes council knowledge bases so agents and bots can find answers quickly.
- Power Automate: Orchestrates workflows—for instance, automatically updating a housing portal when a case status changes.
By staying within the Microsoft stack, Westminster gains a single pane of glass for agents and a consistent data model that simplifies compliance and reporting—a critical factor for public sector bodies under strict data protection regulations.
The Road to Deployment: Discovery to Go-Live in Months
Westminster’s journey began with a discovery phase in mid-2025. Rather than issuing a lengthy tender and building from the ground up, the council appears to have capitalised on Microsoft’s existing capabilities, configuring them to its needs. This low-code approach allowed validation with real resident data quickly, and the feedback loop with citizens informed iterative changes ahead of the September launch.
No major IT transformation goes perfectly, and the council likely encountered challenges: cleaning and unifying data from legacy systems, training the AI to understand the nuances of local authority jargon, and gaining the confidence of agents accustomed to older workflows. However, the decision to go live within a few months suggests that Westminster prioritised getting a working solution into users’ hands early and refining it in production—a hallmark of agile delivery in the public sector.
Wider Adoption: The New Benchmark for Council Contact Centres?
Westminster’s early success could set a precedent. Local authorities across the UK have been exploring AI for years, but many have struggled to move beyond pilot projects. A high-profile deployment in a major London borough that measurably cuts handle times while improving resident satisfaction is a powerful proof point.
Microsoft itself has been aggressively courting public sector clients with its Contact Center and Copilot stories. If Westminster can publish concrete case study data—reduced average call duration, higher first-contact resolution, and lower agent attrition—it will build a compelling business case for other councils to follow. Moreover, the context-preservation model is not limited to phones; it applies equally to digital self-service, where residents increasingly expect the same seamless experience they get from commercial brands.
The Tricky Balance: Privacy and Trust
Context preservation, while convenient, immediately raises privacy questions. The system must retain enough information to smooth interactions without overstepping into surveillance. Westminster and its Microsoft implementation must comply with GDPR and the UK Data Protection Act, ensuring that citizens understand what data is being stored, for how long, and to what end.
The council likely conducted a Data Protection Impact Assessment (DPIA) early in the discovery phase. By default, Microsoft’s Contact Center offers security controls like role-based access, encryption, and audit logs, which help councils stay compliant. Still, the sensitivity of some council services—social care, homelessness, mental health—demands an extra layer of care. If, for example, an AI system inadvertently surfaces a domestic abuse case note to an unqualified agent, the consequences could be severe. Westminster will need to continue refining its data-masking and access rules as the system matures.
Challenges to Watch: Bias, Accuracy, and the Human Touch
No AI is infallible. Natural-language models can misinterpret accents, dialects, or complex multi-part questions. In a visitor-rich, multicultural borough like Westminster, the AI must be robust across dozens of languages and communication styles. If the system misunderstands a resident’s income declaration for a benefits claim, the fallout could be financial and legal.
There’s also the risk of over-automation. Some interactions—particularly those involving vulnerable adults or emotionally charged situations—require a human with empathy. Westminster’s designers likely built clear escalation paths so that the AI hands off to a person when sentiment analysis detects distress or when the query exceeds a complexity threshold. Getting that balance right is an ongoing tuning process, not a one-time switch.
Agent acceptance is another factor. If the AI’s context pane feels like a “big brother” recording every move, morale can suffer. Successful rollouts frame the technology as an assistant that removes drudge work, letting agents focus on meaningful problem-solving. Westminster’s adoption metrics will be telling on this front.
What’s Next for Westminster’s AI Journey
The September 2025 launch is only the foundation. Future phases could expand context preservation from phone and chat to face-to-face interactions at council offices, so that when a resident walks in for an appointment, the receptionist already has the full digital thread. Integration with Internet of Things (IoT) data—such as smart bins reporting fill levels—could even trigger proactive contacts: the council could reach out to a resident before they ever call.
On the analytics side, the wealth of aggregated, anonymised call data will enable Westminster to identify systemic issues. If thousands of residents call about the same planning application confusion, the council can fix the root cause rather than treating each call as an isolated complaint. AI turns the contact centre from a reactive cost centre into a strategic insight hub.
The Bottom Line
Westminster City Council’s bet on preserving resident context through Microsoft’s AI contact centre is paying early dividends: faster calls, less frustration, and more efficient agents. By rapidly moving from discovery to go-live in a matter of months, the council has demonstrated that transformative public-sector AI need not be a multi-year slog. As other local authorities watch closely, Westminster’s experience may become the template for how AI can make government interactions feel less like bureaucracy and more like a coherent conversation.