Delegance Brokerage -- Beating Human-Level Memory in Commercial Insurance

Leading AI Platform Demonstrates Superior Recall and Reasoning Over Human Brokers

NEW YORK, June 2, 2026 /PRNewswire/ -- Delegance Brokerage, an AI powered commercial insurance platform, scores 88% on the LoCoMo long-term memory benchmark, exceeding the published human performance ceiling of 87.9%.

Commercial insurance is a relationship business. Brokers manage hundreds of clients, each with unique coverage needs, renewal timelines, and operational risks that shift over months and years. The difference between a good broker and a great one often comes down to memory: who remembers that a client mentioned expanding to a new state three conversations ago, or that their fleet size changed last quarter.

We built Delegance to be the broker who never forgets. To validate this, we tested our production system against LoCoMo, the industry-standard benchmark for long-term conversational memory.

The Benchmark

LoCoMo (Long-term Conversational Memory), published at ACL 2024 by Snap Research, is the gold standard for evaluating whether AI systems can remember facts, resolve dates, and reason across extended multi-session conversations. It tests 1,542 questions spanning 10 conversations with 300+ dialogue turns each, covering weeks to months of interaction.

The benchmark includes five question categories: single-hop fact recall, temporal reasoning (dates and durations), multi-hop inference (connecting facts across conversations), adversarial questions (false premises and unanswerable queries), and open-domain questions. It is used by every major AI memory vendor to benchmark their systems.

Results

We evaluated our production memory system against the full LoCoMo benchmark. This is the same system that serves our customers every day, not a research prototype or fine-tuned model.

What this means in practice: When a client mentions expanding to a new state in January, asks about fleet coverage in March, and calls about a claim in June, Delegance connects those dots. The system recalls expansion, knows the fleet changed, and surfaces the relevant policy context. We deploy an AI broker with a photographic memory hyper-personalized to each insured.

Why This Matters for Insurance

Most AI memory benchmarks are built for consumer chatbots. LoCoMo is different: it tests the kind of long-term, multi-session recall that mirrors what brokers do every day.

These are not hypothetical requirements. They are the daily reality of managing a commercial insurance book. We tested against LoCoMo because it measures exactly the kind of memory that makes the difference between adequate service and exceptional service.

CONTACT
Press Contact
[email protected]

SOURCE Delegance Brokerage