Eval directory

Evals for Avina

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

About Avina

Avina is an AI agent platform that detects buying signals across the open web, LinkedIn, job posts, first-party site visitors, and CRM events to find in-market B2B accounts. It scores accounts for ICP fit, enriches contacts with verified details, and automates outreach through AI sequences, ad audiences, ABM segments, and existing sales tools. Results sync back to the customer's CRM, Slack, and webhooks, and it is sold on tiered credit-based monthly plans plus a fully managed service.

Industry

B2B AI sales intelligence & signal-based outbound automation

Website

avina.io

Use the eval library for Avina

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 Avina?

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

Signal creation and detection

The natural-language signal builder and the agent that scans the open web, LinkedIn, job posts, first-party site visitors, and CRM events to detect buying intent across the customer's TAM. Covers how a plain-language signal description becomes a monitored, daily-refreshing live audience.

“Avina is an AI agent platform that uses buying signals to help B2B sales teams” www.avina.io

Mapped capabilities

4 capabilities

Plain-language signal specification
Turning a described buying signal ("they just hired their first DevOps engineer") into a correctly scoped, monitored signal rather than a title-match list.
Source coverage and attribution
Detecting across open web, LinkedIn, job posts, site visitors, and CRM events, and attributing each hit to the source and evidence that triggered it.
Live audience refresh
Keeping audiences current as new signals fire daily, including adding newly qualifying accounts and aging out stale ones.
Signals library and custom AI signals
Out-of-the-box signal templates versus custom AI signals, including where a custom signal is the right answer over a prebuilt one.

Illustrative example

Input: Create a signal for cybersecurity buyers: companies that just brought in a new CISO. We sell breach-response tooling.
Expected behavior: Builds a leadership-change signal keyed to a recent CISO appointment event detected from the open web, LinkedIn, and job posts, with a recency window. Explicitly distinguishes this from a static filter on the CISO title, which would return every seated CISO rather than accounts entering a buying window.

ICP scoring and contact enrichment

AI account scoring for ICP fit, target-contact identification at scored accounts, and waterfall enrichment to verified contact details. Covers whether ranking is defensible and whether enrichment claims of verification hold up.

“Avina scores every account for ICP fit, then finds target contacts at those companies” www.avina.io

Mapped capabilities

4 capabilities

ICP fit scoring
Scoring accounts for fit and filtering out accounts that fired a signal but do not match the customer's ICP.
Target contact selection
Identifying the right persona at a scored account rather than any reachable contact.
Waterfall enrichment behavior
Enriching contacts with verified details across providers, including behavior when the first enrichment attempt fails or returns low confidence.
Third-party data integration
Use of built-in RB2B, Vector, and Clearbit inputs and how they combine with first-party visitor data.

Outreach activation and channel routing

How a qualified, enriched account becomes action: AI-drafted sequences, hyper-targeted ad audiences, synced ABM segments, and feeds into traditional sequences in existing sales tools. Covers message grounding and correct channel choice.

“Push enriched contacts, live signals, and account scores into your CRM and Slack in real time” www.avina.io

Mapped capabilities

4 capabilities

AI sequence drafting
Drafting outreach that references the specific signal and evidence, with the right persona, timing, and message per the customer's positioning.
Channel selection
Choosing among AI sequence, ad audience, ABM segment, or existing-tool sequence for a given signal and account state.
Trigger timing and buying windows
Acting on a signal within its useful window and adjusting the pitch as the window shifts by event type and stage.
Audience and segment construction
Building ad audiences and ABM segments that stay consistent with the underlying signal definition.

CRM sync, alerting, and integrations

Pushing enriched contacts, live signals, and account scores into the customer's CRM and Slack in real time, plus webhooks to any tool that supports them. Covers write correctness, signal history, and no-code automation setup.

“Live audiences refresh daily as new signals fire.” www.avina.io

Mapped capabilities

4 capabilities

Real-time CRM writeback
Pushing contacts, signals, and scores into the CRM with full signal history and without creating duplicate records.
Slack feed and signals inbox
Surfacing priority signals to reps with the context and evidence needed to act.
Webhook and automation triggers
Firing webhooks and no-code automations to downstream tools on the intended events.
Workspace, roles, and access
Workspace creation, administrator designation, and authorized-user access as defined in the terms of service.

Credits, plans, and entitlements

The tiered credit-based monthly plans (Starter, Team, Growth, Agency) and the metering rules published on the pricing page. Covers whether the product explains and enforces cost and feature boundaries accurately.

“1 credit per signal. Tech Stack and Custom AI signals cost 2 credits each.” www.avina.io

Mapped capabilities

4 capabilities

Credit accounting
1 credit per signal, 2 credits for Tech Stack and Custom AI signals, first enrichment attempt included, additional enrichment 1 credit.
Plan feature boundaries
Which capabilities belong to which tier, including AI Outbound and Tech Stack signals on Team and AI account scoring on Growth.
Trial and upgrade guidance
The 7-day free trial and recommending the tier that matches a stated team size and monthly signal volume.
Managed service scope
What the Agency plan's done-for-you signal-to-pipeline motion includes versus what self-serve tiers require the customer to run.

Illustrative example

Input: We're on the Team plan at 3,000 credits. This month we ran 1,200 standard signals and 400 Custom AI signals, and 300 contacts needed a second enrichment attempt. Where do we land?
Expected behavior: Computes 1,200 standard signals at 1 credit, 400 Custom AI signals at 2 credits, and 300 additional enrichment attempts at 1 credit each, noting the first enrichment attempt is included. Reports the 2,300 total against the 3,000 allowance and states remaining headroom.

Vertical signal reasoning and comparative positioning

Applying signal-based selling to the specific verticals Avina publishes support for, and answering evaluation questions about how Avina differs from database-first and competing tools without overclaiming.

Mapped capabilities

4 capabilities

Vertical-specific signal mapping
Naming the signals that matter for a stated vertical, such as new CISOs for cybersecurity or first DevOps hires for devtools.
Off-database account discovery
Explaining and demonstrating agentic prospecting for accounts that snapshot databases like ZoomInfo and Apollo miss.
Competitive framing accuracy
Comparing against tools like Clay and Unify on motion fit and cost without asserting capabilities the published material does not support.
Claim discipline
Declining to invent metrics, customer names, or guarantees beyond what Avina publishes.

Coverage is mapped from Avina'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 Avina test?+

The coverage map is generated from Avina's own public product surface (B2B AI sales intelligence & signal-based outbound automation): 6 scoring areas — Signal creation and detection, ICP scoring and contact enrichment, and Outreach activation and channel routing, 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 Avina evals scored?+

Every case generated for Avina — across Signal creation and detection and ICP scoring and contact enrichment 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 Avina library include?+

The full Avina library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Plain-language signal specification and Source coverage and attribution under Signal creation and detection); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

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