Corsac
Use Corsac
All evals
CR

Eval directory

Evals for Common Room

Eval coverage for Common Room, mapped from its public product surface.

About Common Room

Common Room is an AI GTM platform that unifies enrichment, buying signals, identity resolution, and AI agents into a single buyer intelligence layer. Its RoomieAI agents capture signals across first-, second-, and third-party channels, research accounts, and run 1:1 outbound and pipeline plays. It is sold to RevOps, SDR, AE, demand gen, and ABM teams, with access via integrations, a Chrome extension, an MCP server, and a CLI.

Industry

AI go-to-market / buyer intelligence platform

Website

www.commonroom.io

Use the eval library for Common Room

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

Generate your own →

Related

All evals →

More AI eval libraries

Coverage map

What would you measure for Common Room?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

  • Adversarial robustness
  • Workflow quality
  • Safety gates
  • Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

01

Signal capture and coverage

Always-on capture of buying signals across first-, second-, and third-party channels, and their correct classification and freshness once ingested.

“Never miss a buying signal with agentic, always-on signal capture across hundreds of 1st, 2nd, and 3rd-party channels.” www.commonroom.io

Mapped capabilities

4 capabilities

  • Job change detection

    Champion moves are detected, attributed to the new account, and surfaced as an actionable signal.

  • Website visit and dark funnel signals

    De-anonymized visits and off-platform activity are captured and tied to a known person or account.

  • Product-led sales signals

    In-product usage events are ingested as buying signals rather than isolated telemetry.

  • Signal stacking and recency

    Multiple signals on one buyer combine into a single prioritized view with correct timestamps.

02

Identity resolution and enrichment (Person360)

Waterfall enrichment and identity resolution that turn fragmented handles, emails, and visits into one person and one account record.

“Up to 750k contacts 10K RoomieAI research credits 15K Prospector credits” www.commonroom.io

Mapped capabilities

4 capabilities

  • Cross-channel identity merge

    Separate identifiers for the same human resolve to one Person360 profile without over-merging distinct people.

  • Waterfall enrichment ordering

    Enrichment providers are consulted in configured order and the winning field value is traceable to its source.

  • Account association

    People are attached to the correct company, including after a job change or acquisition.

  • Prospector coverage and credits

    Prospector lookups return contacts within the workspace's credit balance and report consumption accurately.

Illustrative example

Input
A tracked champion's LinkedIn title changes from Director of RevOps at Acme to VP of RevOps at Northwind. Show their Person360 profile.
Expected behavior
The profile stays a single merged identity rather than splitting into two people, and the current employer resolves to Northwind with Acme retained as prior history. The job change is surfaced as a dated signal on the new account.

03

RoomieAI agent research and grounding

Agent-generated research, summaries, and answers that must be grounded in captured buyer intelligence rather than generic public data.

“Revenue agents built on complete buyer intelligence.” www.commonroom.io

Mapped capabilities

4 capabilities

  • Account research (Brief)

    Generated account briefs cite the underlying signals and records they were built from.

  • Ask CR Anything

    Natural-language questions are answered from workspace data, with abstention when the data is absent.

  • Spark signal triage

    Surfaced sparks reflect genuine in-market activity and explain why the buyer was flagged.

  • Grounding and hallucination control

    Agents do not assert firmographic or intent facts that are not present in the resolved record.

Illustrative example

Input
Ask CR Anything: "What is Northwind's current headcount and which of their engineers visited our pricing page last week?" No website visit signals exist for Northwind.
Expected behavior
The answer states that no pricing page visits from Northwind were captured rather than naming any individuals, and any headcount figure is attributed to an enrichment source. It does not fabricate visitor names or intent.

04

Outbound generation and pipeline plays

Actions and the Activate agent that turn a signal into a prioritized play and a 1:1 outbound message at scale.

“Hunt for signals, orchestrate pipeline plays, and run 1:1 outbound at scale with RoomieAI™ agents” www.commonroom.io

Mapped capabilities

4 capabilities

  • Play triggering and timing

    Plays fire on the intended signal condition and suppress when the trigger is stale or already worked.

  • 1:1 message personalization

    Generated outbound references the specific signal and account context behind the play.

  • Prioritization and lead scoring

    Scored leads and surfaced plays rank in an order a rep would recognize as highest-intent first.

  • Send guardrails

    Volume, suppression, and brand-safety constraints hold rather than defaulting to spray-and-pray sends.

05

CRM sync and data integrity (DataAgent)

Keeping the CRM aligned with reality — write-backs, field mapping, and conflict handling across connected integrations.

Mapped capabilities

4 capabilities

  • Write-back correctness

    Updates land on the right CRM object and field without clobbering higher-confidence existing values.

  • Conflict and duplicate handling

    Conflicting or duplicate records are reconciled or flagged rather than silently merged.

  • Integration sync reliability

    Sync failures are detected and reported instead of leaving stale data presented as current.

  • Segment and workflow consistency

    Segments, alerts, and workflows reflect the same resolved data the UI shows.

06

Headless and embedded access surfaces

Parity and access control across the execution surfaces: MCP server, CLI, Chrome extension, and plan-tier entitlements.

“Connect AI assistants to your buyer intelligence through the MCP Server or access it headlessly through our CLI” www.commonroom.io

Mapped capabilities

4 capabilities

  • MCP server responses

    Connected AI assistants receive the same resolved intelligence available in the app.

  • CLI automation

    Scheduled jobs and pipelines get stable, scriptable output with clear failure signals.

  • Chrome extension context

    In-page buyer context matches the underlying Person360 record.

  • Plan entitlements and limits

    Seats, contact caps, credit pools, and integration availability enforce per the purchased tier.

Coverage is mapped from Common Room's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Common Room test?+

The coverage map is generated from Common Room's own public product surface (AI go-to-market / buyer intelligence platform): 6 scoring areas — Signal capture and coverage, Identity resolution and enrichment (Person360), and RoomieAI agent research and grounding, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Common Room evals scored?+

Every case generated for Common Room — across Signal capture and coverage and Identity resolution and enrichment (Person360) and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Common Room library include?+

The full Common Room library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Job change detection and Website visit and dark funnel signals under Signal capture and coverage); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Common Room or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Common Room areas and set them up in a Corsac workspace, where you can run every test case against Common Room or your own agent with your own data.