For Legal & General (L&G), a nearly 200‑year history of handling life’s biggest financial moments is meeting a very modern approach to customer service: a multi‑year collaboration with Microsoft to deploy an AI‑first contact‑center platform built on Microsoft Dynamics 365 Contact Center and Copilot. The program promises a single, real‑time “pane of glass” for retail colleagues, embedded generative‑AI tools to remove repetitive admin, and new conversational intelligence — all designed to scale across the insurer’s 12‑plus million customers and thousands of service interactions every day.
L&G’s retail business has consolidated several long‑standing operational units into a single retail organisation, an organisational change that created both the need and the opportunity to rationalise customer‑service systems. Craig Brown, Chief Operating Officer for Retail, oversees roughly 2,800 colleagues who manage more than 20 million customer interactions a year — the very workload the new platform is designed to simplify. The initial rollout targets workplace savings, retail protection policies and annuities, with subsequent phases planned to expand coverage.
This work sits squarely on Microsoft’s Copilot‑first contact‑center stack: Dynamics 365 Contact Center as the omnichannel front end and case backbone; Dataverse as the enterprise data plane; and Copilot plus Copilot Studio as the generative‑AI and agent‑behaviour layer. Microsoft positions Dynamics 365 Contact Center as a cloud CCaaS (Contact Center as a Service) solution that natively supports real‑time transcription, sentiment analysis, conversation summarization and “agent assist” workflows — capabilities L&G is explicitly adopting.
Watch for these red flags:
That said, the outcomes L&G and Microsoft describe remain partly vendor‑reported. Realising durable gains requires rigorous baseline measurement, ongoing auditability, tight data governance and commercial visibility into Copilot consumption. If L&G maintains that discipline while extending the platform, the initiative could become a defining case study in how large insurers modernise customer care with AI — not by replacing human judgement but by amplifying it, one enhanced conversation at a time.
Source: Microsoft UK Stories How L&G is reinventing customer care with AI
Background / Overview
L&G’s retail business has consolidated several long‑standing operational units into a single retail organisation, an organisational change that created both the need and the opportunity to rationalise customer‑service systems. Craig Brown, Chief Operating Officer for Retail, oversees roughly 2,800 colleagues who manage more than 20 million customer interactions a year — the very workload the new platform is designed to simplify. The initial rollout targets workplace savings, retail protection policies and annuities, with subsequent phases planned to expand coverage. This work sits squarely on Microsoft’s Copilot‑first contact‑center stack: Dynamics 365 Contact Center as the omnichannel front end and case backbone; Dataverse as the enterprise data plane; and Copilot plus Copilot Studio as the generative‑AI and agent‑behaviour layer. Microsoft positions Dynamics 365 Contact Center as a cloud CCaaS (Contact Center as a Service) solution that natively supports real‑time transcription, sentiment analysis, conversation summarization and “agent assist” workflows — capabilities L&G is explicitly adopting.
What L&G is building: a practical architecture
A single pane of glass for agents
At the heart of the initiative is a unified agent workspace that consolidates voice, chat and email into a single view so colleagues can see the full customer context without switching between multiple legacy systems. That single pane of glass reduces call transfers, eliminates repetitive re‑explanations, and gives advisors the time and context to focus on empathic conversation rather than data hunting.Core technical components
- Dynamics 365 Contact Center — the omnichannel agent UI, routing engine, and transcription/sentiment surface.
- Microsoft Dataverse — canonical storage for transcripts, case state and telemetry to enable auditable records and downstream analytics.
- Microsoft Copilot / Copilot Studio — conversation summarization, suggested next steps, knowledge retrieval, and the low/no‑code environment for authoring constrained “Copilot agents.”
How the AI features integrate into daily workflows
- Real‑time transcription and sentiment cues appear in the agent UI while a call or chat is in progress. Agents receive contextual prompts and recommended next steps based on the conversation.
- Copilot automates routine wrap‑up work: creating case summaries, drafting follow‑up emails and indexing conversation highlights into Dataverse so downstream systems and analytics can use them. Administrators can enable auto‑summarization so summaries appear when an agent joins or ends a conversation.
- Supervisors and operations teams get live dashboards and telemetry for KPIs such as average handling time, transfers, first contact resolution and agent assist usage — all feeding a continuous improvement loop.
Why this matters: business and customer benefits
For colleagues: less toil, more time for empathy
The most immediate operational win claimed by L&G is the removal of mechanical tasks that consume agent time. With transcription, auto‑summaries and suggested replies, frontline staff spend less time on after‑call work and more time on high‑value, human interactions — particularly important in emotionally sensitive moments such as bereavement or retirement planning. This human‑centred automation approach is central to L&G’s stated objectives.For customers: faster, more consistent service
Consolidating customer history and presenting it before the agent answers a call reduces transfers and repetition — both common drivers of poor customer experience. The platform also enables proactive contact and personalised outreach based on signals surfaced across thousands of conversations. Early L&G messaging stresses improved first‑contact success and fewer hand‑offs.Strategic outcomes: product and process refinement
Because Copilot and the Contact Center store structured summaries and sentiment data, L&G can spot systemic friction points across product lines and communications. Summarised insights across thousands of interactions become an input to product design, customer communications and process redesign — a continuous learning loop that shifts the business from reactive fixes to proactive improvement.What the technology actually does — technical anatomy
Transcription, summarization and sentiment
Dynamics 365 Contact Center and Copilot deliver real‑time transcription and post‑call conversation summaries. These features are generally available products within the Microsoft stack and can be enabled to auto‑generate summaries at key triggers (for example, when an agent joins a call or when the conversation ends). Sentiment analysis can be surfaced to agents and supervisors to help identify vulnerability or escalation needs.Copilot agents and Copilot Studio
Copilot Studio provides a no/low‑code environment to author constrained agents — pre‑configured AI behaviours that retrieve answers from curated knowledge bases, execute safe automations, and escalate to humans when necessary. That "agent network" model allows organisations to create specialised behaviour for tasks such as intent detection, knowledge base updates, or guided case creation while retaining governance and audit trails.Dataverse as the audit and analytics plane
Dataverse stores transcripts, Copilot interactions and agent feedback. Because this data is the canonical record, it enables auditable trails for compliance, training datasets for constrained agents, and the telemetry that powers supervisor dashboards. Microsoft documentation makes explicit the value of this separation for regulated industries.Early wins and corroborated achievements
L&G points to recent digital advances that the new contact‑center program will build on. Two items are notable and independently verifiable:- A fully digitised claims submission process for retail protection policies that L&G says has reduced average claim times by nearly two weeks; industry reporting confirms reduced timelines and lower follow‑up evidence requests following the online claims submission launch.
- A market‑leading workplace pension app and the Guided Retirement Planner — products L&G uses as proof points for the broader digital transformation the contact‑center effort is meant to extend. L&G’s corporate communications and annual reporting document the retail footprint and digital investments underpinning these claims.
Strengths: why the Microsoft‑first approach is sensible for L&G
- First‑party alignment reduces integration risk. Choosing Dynamics 365 + Dataverse + Copilot shortens integration work, leverages pre‑built connectors, and simplifies identity and compliance integration with Azure and Entra controls. That makes the architecture easier to secure and support at scale.
- Human‑centred automation. L&G’s explicit stance — use AI to augment, not replace, human agents — is operationally pragmatic for regulated financial services where empathy and human judgment matter. Embedding Copilot as an assist rather than an autonomous advisor helps manage risk.
- Observability and continuous improvement. Built‑in telemetry and supervisor dashboards make it possible to measure outcomes (AHT, FCR, CSAT) and iterate. This observability is essential to validating ROI and tuning AI behaviour over time.
- Repeatable delivery pattern. Microsoft and its partners have documented factory‑style implementation blueprints for Dynamics‑centric contact centers that accelerate deployments across markets and languages — a practical advantage for enterprise rollouts.
Risks, governance and the caveats every enterprise must manage
Vendor‑reported metrics need independent validation
Several operational claims — faster handling times, reduced transfers and improved CSAT — are meaningful but so far are vendor‑reported or company statements. These figures should be treated as promising indicators that require independent baseline measurement and ongoing, auditable tracking to validate long‑term impact. L&G’s own press materials and Microsoft stories present these as outcomes to be realised rather than independently audited facts. Treat such metrics as provisional until third‑party validation is available.Generative AI risks: hallucination and inappropriate guidance
Generative models can produce plausible but incorrect outputs (hallucinations). In a financial‑services context — where customers may act on advice about pensions, annuities or claims — uncontrolled generative responses are a material risk. Best practice requires that Copilot agents be constrained to curated knowledge bases and that confidence thresholds trigger automatic escalation to human advisers when uncertainty exceeds policy limits. Microsoft’s guidance and product controls provide mechanisms for this, but organisational discipline is the real guardrail.Data governance, residency and privacy
Storing conversation transcripts and AI‑derived metadata in Dataverse is convenient, but multijurisdictional regulations require clear policies for data residency, retention, access control and deletion. Firms must document cross‑border data flows, apply least‑privilege access, and embed audit trails suitable for regulators and internal compliance teams. Financial services regulators expect traceable decision paths for any automation that affects customer outcomes.Commercial exposure: licensing and consumption costs
Copilot consumption can be usage‑sensitive. As automation grows, token‑ or action‑based consumption may materially increase costs if not actively monitored and capped. Organisations should negotiate clear pricing terms, monitoring dashboards, and governance controls to prevent runaway costs as Copilot usage scales.Voice automation acceptance and authentication
Voice bots and conversational IVRs can improve containment but introduce authentication and accessibility challenges. L&G will need robust voice‑auth flows and fallback paths for customers who prefer human engagement or where ID verification is critical. Early pilots should measure voice completion and escalation rates before broadening deployment.A practical 6‑step operational playbook for regulated firms
- Baseline measurement: capture pre‑rollout metrics (AHT, CSAT, FCR, transfers, agent occupancy) and define success thresholds.
- Start small: deploy Copilot for agent assist (summaries, knowledge retrieval, draft replies) before moving to constrained automation for self‑service.
- Harden governance: map data flows, set retention policies, and implement least‑privilege access across Dataverse and Copilot agents.
- Curate knowledge: maintain versioned, validated knowledge bases for agent retrieval; ban unsanctioned web ingestion for regulated guidance.
- Monitor and iterate: use live dashboards and operational telemetry to tune routing, agent prompts and escalation thresholds.
- Stage voice automation carefully: pilot voice IVR automation in low‑risk use cases and validate authentication, completion and NPS before scaling.
What success will look like for L&G (and what to watch for)
Success will be visible in both quantitative and qualitative signals. Quantitatively, expect to see measurable reductions in average handling time, after‑call work, and transfers; improvements in first‑contact resolution and operational throughput; and, crucially, improved CSAT scores in target product lines. Qualitatively, success will be reflected in agent sentiment (less burnout, higher engagement), faster product changes informed by conversational insight, and more personalised customer interactions. These outcomes are achievable, but they depend on disciplined measurement, strong governance and rigorous change management.Watch for these red flags:
- Vendor‑only metrics without audited baselines.
- Unconstrained generative outputs used for decisioning without human review.
- Spikes in Copilot consumption or third‑party service usage that drive unexpected costs.
The broader industry context
Microsoft’s Copilot‑first contact center push is part of a wider industry shift: CRM vendors, cloud providers and specialist CCaaS players are embedding generative AI into every customer‑service workflow. Reuters and multiple industry reports have chronicled Microsoft’s efforts to bring Copilot into contact centers, and Microsoft’s product documentation and customer stories demonstrate both the feature set and the expected operational outcomes. There is growing evidence from early enterprise deployments that generative AI can materially reduce repetitive work and free agents for higher‑value tasks — provided governance and measurement are in place.Conclusion
L&G’s Microsoft collaboration is a textbook example of how a large, regulated financial services organisation is applying a measured, human‑centred approach to AI in customer service. By consolidating channels and data into Dynamics 365 Contact Center and using Copilot to automate time‑consuming administrative tasks, L&G aims to give its 2,800 retail colleagues the context and tools to deliver faster, more empathetic service to more than 12 million customers. The architecture aligns with Microsoft’s recommended patterns — Dataverse as the enterprise data plane, Copilot Studio for governed agent behaviour, and Dynamics 365 for omnichannel orchestration — and the company’s early digital wins (a digitised claims process and a top‑rated pensions app) create a credible runway for impact.That said, the outcomes L&G and Microsoft describe remain partly vendor‑reported. Realising durable gains requires rigorous baseline measurement, ongoing auditability, tight data governance and commercial visibility into Copilot consumption. If L&G maintains that discipline while extending the platform, the initiative could become a defining case study in how large insurers modernise customer care with AI — not by replacing human judgement but by amplifying it, one enhanced conversation at a time.
Source: Microsoft UK Stories How L&G is reinventing customer care with AI
