Eval directory

Evals for Unify

Eval coverage for Unify, mapped from its public product surface.

About Unify

Unify is an AI-native outbound platform that lets sales reps build target lists, research and qualify accounts, and run multi-channel sequences from a single chat interface. It combines intent signals from 40+ data vendors with AI agents, automated plays, AI copywriting, and managed email deliverability. Analytics dashboards and export APIs attribute pipeline back to the plays, signals, and sequences that created it.

Industry

AI outbound sales prospecting platform (GTM)

Website

www.unifygtm.com

Use the eval library for Unify

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

All evals →

More AI eval libraries

Coverage map

What would you measure for Unify?

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

Conversational Prospecting Agents

The chat-first agent surface that turns a plain-English ICP description into researched, qualified accounts and contacts drawn from the underlying data network.

“Unify has helped sellers execute 90% faster from list building to sequence writing.” www.unifygtm.com

Mapped capabilities

4 capabilities

Natural-language list building
Translating an ICP prompt (industry, funding stage, headcount, hiring activity, geography) into a structured, bounded target list.
Research and qualification agents
Multi-step agents that scrape sites, browse, and analyze data to qualify fit, and report the basis for a qualify/disqualify call.
Data-source routing and attribution
Selecting among the 40+ enrichment and signal providers for a given request and attributing returned fields to their source.
Learned rep context
Carrying a rep's stated style, business, and industry context across turns without overriding explicit in-prompt instructions.

Illustrative example

Input: Build a list of 100 people at VC-backed B2B SaaS companies that are currently hiring for SDRs.
Expected behavior: The agent restates the request as explicit filters covering funding status, B2B SaaS category, and active SDR hiring, returns at most 100 contacts, and attributes each enriched field to a named data source rather than presenting unsourced values.

Signal Coverage and Intent Data

The unified intent layer combining first-party engagement, third-party vendor feeds, and AI-discovered signals into a single view of who is in-market and why.

Mapped capabilities

4 capabilities

Vendor catalog discovery and filtering
Filtering the signal directory by category, geography, and vendor type, and reporting the data points a given vendor actually supplies.
First-party website and product signals
Capturing visitor identification and product-usage activity and resolving it to an account or contact.
Job change and hiring signals
Detecting role changes and hiring activity and expressing them as triggerable conditions.
Signal provenance and recency
Stating which source produced a signal and when it was observed, rather than presenting it as unattributed fact.

Plays and Outbound Automation

The workflow engine that runs outbound motions on autopilot: signals trigger a chain of enrichment, agent qualification, sequencing, and internal alerts.

Mapped capabilities

4 capabilities

Signal-triggered activation
Firing a play from a chosen signal, and not firing when the trigger condition is unmet.
Multi-step agent and enrichment steps
Ordered execution of enrichment, AI agent research, and qualification steps within one play run.
Conditional enrollment and branching
Enrolling only records that pass qualification, and handling the non-qualified branch without side effects.
Internal alerting
Slack notifications that reflect what the play actually did, including skipped and failed steps.

Illustrative example

Input: Run a play on pricing-page visits: enrich the visitor, qualify with an AI agent, enroll only if qualified, and alert the account owner in Slack.
Expected behavior: For a visitor the agent judges unqualified, the play completes the enrichment and qualification steps, skips enrollment entirely, sends no outbound message, and the Slack alert states that the record was disqualified rather than enrolled.

Sequencing, Copywriting, and Deliverability

Multi-channel sequence execution with AI-drafted copy grounded in research, plus the managed sending infrastructure underneath it.

Mapped capabilities

4 capabilities

Multi-channel step orchestration
Blending automated email, manual tasks, call steps, and social touches in one coordinated flow.
Signal-to-first-touch context carry
Preserving the triggering signal and research context so the opening message reflects why outreach is happening.
Grounded AI copywriting
Drafting personalized messages that stay within the researched facts about the prospect and the rep's stated voice.
Managed mailbox and send-time validation
Mailbox provisioning and warming, and address validation at send time before a message goes out.

Analytics, Attribution, and Data Out

Dashboards, prompt-driven reporting, and export paths that tie pipeline back to the plays, signals, and sequences that produced it.

Mapped capabilities

4 capabilities

Out-of-the-box dashboards
Standard views of rep activity, sends, replies, calls, tasks, and deliverability across a team.
Prompt-driven insight queries
Answering plain-English performance questions with figures traceable to the underlying sequences, plays, and reps.
Pipeline attribution
Attributing opportunities and closed-won revenue back to the originating play, signal, or sequence.
Export APIs and integrations
Bulk API, warehouse, CRM, and assistant destinations for pushing Unify data into external reporting.

Plans, Credits, and Access Control

The commercial and permission boundaries of the product: credit accounting per action, seat limits, plan-gated capabilities, and the scope of CRM sync.

Mapped capabilities

4 capabilities

Credit consumption and limits
Reporting credit cost of an action and behavior when a seat or workspace allotment is exhausted.
Plan-gated feature access
Availability of signal-triggered automations, managed mailboxes, model tier, and other tier-restricted capabilities.
CRM sync scope
Respecting read-only HubSpot and Salesforce sync where the plan grants read access only.
Seat and workspace boundaries
Seat limits per plan and containment of data and configuration within a workspace.

Coverage is mapped from Unify'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 Unify test?+

The coverage map is generated from Unify's own public product surface (AI outbound sales prospecting platform (GTM)): 6 scoring areas — Conversational Prospecting Agents, Signal Coverage and Intent Data, and Plays and Outbound Automation, 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 Unify evals scored?+

Every case generated for Unify — across Conversational Prospecting Agents and Signal Coverage and Intent Data 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 Unify library include?+

The full Unify library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Natural-language list building and Research and qualification agents under Conversational Prospecting Agents); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

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