How Definity Cut 40% of Call Handle Time by Flipping When AI Intervenes

Definity Financial Corporation (TSX: DFY), Canada's largest independent property and casualty insurer, deployed Google Cloud Contact Center AI in partnership with Deloitte Consulting across its live agent operations. Within one month of go-live, average call handle time dropped by 3.5 minutes per call — equal to 40% of average handle time — recovering more than 20,000 agent hours per year. Source: Google Cloud's official published case study, corroborated by Deloitte's implementation team.

The cost hidden inside each call

Definity's contact center handles more than 500 customer calls per day across its Canadian insurance portfolio. During each interaction, agents navigated multiple legacy applications simultaneously — pulling policy data, claims history, and authentication records from separate systems while keeping the customer engaged. After each call, agents spent an additional three to five minutes writing summary notes by hand. The compound result: extended call durations, slower wrap-up queues, and agent capacity absorbed by low-value administrative work. Definity tasked Deloitte with a precise question: identify the tasks consuming agent time that deliver zero additional value to the customer experience. The answer pointed directly at in-call information retrieval and post-call summarization.

The decision that reversed the standard playbook

The conventional AI assist sequence in contact centers follows a post-call logic — complete the call, then apply AI to generate a summary or transcript. Definity's team reversed this sequence entirely. The decisive choice: deploy AI as an active layer during live conversations, surfacing relevant policy and claims data to the agent in real time, before the caller had to wait or repeat context. Four CCAI capabilities went live in staged sequence — virtual agent authentication at the front of the call, real-time agent assist during the conversation, automated post-call summarization via Vertex AI, and sentiment analysis for intelligent routing. Technically, audio transcriptions flow through Google Cloud's Data Loss Prevention layer, into Vertex AI large language models for data extraction and summarization, then return to Salesforce for storage — eliminating manual note entry entirely. The prerequisite: a 10-month BigQuery and Vertex AI data migration, completed 50% faster than industry benchmarks, that gave Definity the unified data foundation from which CCAI could pull multi-system information in real time.

The result — measured at one month

Within approximately one month of going live at end of 2023, Definity measured a 3.5-minute reduction in average call handle time. Against a pre-implementation baseline where calls ran three to five minutes longer due to in-call information lag and post-call summarization, the 3.5-minute recovery represents 33 to 40% of total handle time. Across a daily volume exceeding 500 calls, these per-call savings compound: Google Cloud's published case study puts the annualized total at more than 20,000 agent hours recovered per year. A candid note on scope: the headline metric captures operational efficiency — the time removed from each transaction. Qualitative reporting from Definity describes notably improved agent satisfaction and active progress redeploying freed agents toward higher-value work including underwriting support. Hard CSAT or NPS deltas remain absent from published sources.

What other organizations can learn

Definity's result rests on three replicable conditions. First, a unified data foundation: Definity's 200TB migration into BigQuery made it possible for CCAI to pull multi-system data in real time. Organizations that operate on fragmented data infrastructure will encounter latency that erodes any in-call benefit. Second, sequenced deployment: Definity staged four capabilities — authentication, agent assist, summarization, sentiment analysis — validating each phase before the next went live, rather than launching everything at once. Third, immediate measurement: results emerged in one month precisely because Definity established pre-go-live baselines and measured the delta at speed. The 40% reduction would have remained invisible absent that measurement discipline from day one. The transferable question for any enterprise running high-volume contact centers: at which point in the call does your AI assist arrive? Shifting that intervention from post-call to real-time — the architectural decision Definity made — is the change that produced 40% handle time reduction and more than 20,000 hours recovered annually.