YARDSTICK Take the audit Join the Waitlist

Yardstick Research tear-sheet / insurance brokerage cohort

Methodology · how we score · rubric weights in plain sight · vendors received this sheet seven days before publication and could flag factual errors, never rankings

Cara

Identity

Total score: 63.8 / 100

Weighted dim sum: 73.75. Minus 10.0 pricing-transparency penalty (hard: no public rate card; pricing requires a demo).

Dimension scores

Dimension Score Weight Weighted Evidence
AI capability depth 4 / 4 15 15.0 [VENDOR-CLAIMED] Cara is a fully AI-native assistant: natural language query of AMS-stored policy data, AI-generated coverage comparison and gap analysis, quote research acceleration, and client communication draft generation. AI-native score of 100/100 is tied for the cohort's highest. The underlying model stack is not publicly disclosed. - https://getcara.ai/product
Workflow integration depth (AMS) 3 / 4 25 18.75 [VENDOR-CLAIMED] Native integration with Applied Systems EPIC, Vertafore AMS360, and EZLynx - Cara queries client, policy, and renewal data from the AMS to power its AI responses. The AI assistant reads from the AMS; write-back capability (saving AI-generated content to the AMS record) is claimed but the bidirectional depth is not independently verified.
Vertical specialization 4 / 4 20 20.0 [VENDOR-CLAIMED] Insurance-only: all AI models, training data, and conversation design are built around P&C personal and commercial lines. No horizontal productivity tool features. Built specifically for the insurance agent workflow. - https://getcara.ai
Implementation + time-to-value 3 / 4 10 7.5 [VENDOR-CLAIMED] AMS integration configuration and AI assistant access typically runs days to 1-2 weeks. Individual agents can begin using Cara for quote research and coverage analysis immediately after AMS integration is configured.
Data + compliance posture 2 / 4 5 2.5 [VENDOR-CLAIMED] CCPA compliance referenced. SOC 2 status not publicly surfaced as of research date. An AI assistant that accesses client PII from the AMS has significant data governance requirements that buyers should verify before deployment. [UNKNOWN - SOC 2 audit date, data residency policy, AI model training data handling, E&O implications of AI-generated coverage recommendations]
Pricing + scalability 1 / 4 10 2.5 [UNKNOWN - no public rate card] No pricing published on getcara.ai. Requires a demo for any pricing information. Hard penalty applied. [THIRD-PARTY ESTIMATE - AI assistant platforms in this segment typically price per seat per month; specific Cara rates not available]
Vendor strength + named-customer evidence 2 / 4 15 7.5 [VENDOR-CLAIMED] Early-stage vendor with a growing customer base of small to mid-size independent agencies. Limited G2 or public review volume. Enterprise-level named customers not publicly disclosed. - https://getcara.ai/customers
Base weighted total 100 73.75
Pricing transparency penalty −10.0 Hard: no public rate card; pricing fully opaque.
Adjusted score 63.8

Top strength

AI capability depth. Cara ties Sonant AI for the cohort's highest AI capability score (4/4). An AI assistant that can answer "what are this client's coverage gaps compared to industry standard for a contractor of this size?" - pulling data from the AMS and reasoning across policy lines - is a genuine productivity multiplier for individual insurance agents who currently answer that question through manual research.

Top gap

Data compliance posture. The 2/4 compliance score is the most important due-diligence flag for buyers. Cara accesses client PII from the AMS (names, policy details, premium data) and processes it through an AI model whose training data handling and data residency are not publicly disclosed. An E&O insurance context adds additional stakes: if an AI-generated coverage recommendation leads to an under-insured client, the agency's E&O liability exposure needs to be scoped. Buyers should request a vendor security assessment and data processing agreement before deployment.

Editorial assessment

Cara represents the AI-copilot thesis for insurance agency productivity: rather than replacing the AMS or adding a point solution for a specific workflow, Cara is a conversational AI layer that augments whatever the insurance agent is already doing inside Applied Systems, Vertafore, or EZLynx. The thesis is sound, and the AI capability depth is real - the combination of AMS data grounding and insurance-specific domain training is what separates an insurance-specific AI assistant from a generic LLM with insurance templates.

The data compliance score (2/4) is the primary evaluation gate for risk-aware buyers. An AI assistant that queries client PII from the AMS is meaningfully different in data risk profile from a marketing automation tool or a COI platform. The absence of public SOC 2 documentation and the lack of clarity on model training data handling are gaps that a compliance-conscious agency principal or their E&O carrier will flag in due diligence.

The vendor strength score (2/4) reflects early-stage reality: Cara is a newer entrant in a category (AI copilots for insurance) that is crowded with claims but thin on independently verified production outcomes. The technology is credible; the track record at enterprise scale is not yet established.

Pricing opacity (hard penalty, −10 points) is the clearest actionable signal: agents evaluating Cara cannot model cost-per-productive-hour without a pricing disclosure. Request per-seat pricing upfront.

Best for

Right-of-reply

Cara (Oyster Technologies) received this tear-sheet seven calendar days before publication of the Yardstick Research 2026 Report, including all measured numbers, sample outputs, and editorial assessment. Cara was given the opportunity to flag factual errors - incorrect pricing, misquoted feature availability, outdated screenshots, factual misstatement in the editorial assessment. Cara was not given the opportunity to request a score revision, dispute the rubric or its weights, withdraw from inclusion, negotiate ranking placement, or suggest changes to the editorial assessment beyond factual correction. Where Cara flagged a factual correction, the correction was applied if verified and noted here; where Cara disputed scoring, the dispute is recorded in the appendix but the score stands. Silence during the right-of-reply window was treated as no objection.

Sources