Corsac
Use Corsac
All evals
GA

Eval directory

Evals for Gojiberry AI

Eval coverage for Gojiberry AI, mapped from its public product surface.

About Gojiberry AI

Gojiberry AI is an AI GTM agent for B2B outbound teams that learns a business from its website and finds high-intent prospects. It detects buying and social signals, scores leads against the ideal customer profile, and runs personalized LinkedIn and email outreach automatically. It reports learning from results over time and benchmarking campaigns against industry top performers.

Industry

AI GTM / outbound sales agent

Website

www.gojiberry.ai

Use the eval library for Gojiberry AI

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 Gojiberry AI?

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

Business Learning & ICP Construction

Onboarding starts from a single website URL: the agent must derive what the business sells, who buys it, and an ideal customer profile precise enough to drive downstream targeting.

“Your agent detects buying & social signals, scores every prospect against your ideal customer” gojiberry.ai

Mapped capabilities

4 capabilities

  • Website-to-business-summary extraction

    Derives offering, category, and value proposition from the entered domain without further user input.

  • ICP attribute inference

    Proposes buyer roles, company size, and segment characteristics traceable to site evidence.

  • Ambiguous or thin-site handling

    Behavior when the site is a stub, multi-product, or non-B2B and the ICP cannot be confidently inferred.

  • ICP correction and re-learning

    Incorporates user edits to the profile and reflects them in subsequent targeting.

02

Buying & Social Signal Detection

The product's stated differentiator is deciding who to contact and when, based on detected intent. Covers the signal types named on the site and how they are surfaced and justified.

Mapped capabilities

4 capabilities

  • Named signal types

    Competitor engagement, follows-your-company, and active-in-your-space signals.

  • Signal-to-prospect attribution

    Ties a detected signal to a specific person or account rather than a generic list.

  • Timing judgment

    Distinguishes fresh, actionable intent from stale or incidental activity.

  • Signal rationale

    States why a lead matters in terms a rep can verify.

03

Lead Scoring & Pre-Filtering

Every prospect is scored against the ICP and pre-filtered so no message goes to an out-of-profile buyer. Covers ranking quality, exclusion, and the precision-over-volume posture the site claims.

“Every lead is pre-filtered to match your ideal buyer profile.” gojiberry.ai

Mapped capabilities

4 capabilities

  • ICP-relative scoring

    Assigns a score reflecting fit and prioritizes the highest-converting prospects first.

  • Out-of-profile exclusion

    Suppresses leads that do not match the buyer profile instead of contacting them.

  • Ranking consistency

    Comparable prospects receive comparable scores across a run.

  • Empty or low-yield result handling

    Reports when few qualified prospects exist rather than padding the list.

04

Multichannel Outreach Generation

The agent writes and coordinates personalized LinkedIn and email outreach without user-built sequences. Covers message grounding, channel coordination, and follow-up behavior.

“Your agent reaches out via email and socials with AI personalized messages, coordinated automatically, no sequences to build.” gojiberry.ai

Mapped capabilities

4 capabilities

  • Signal-grounded personalization

    Message references the specific evidence that made the lead relevant, without invented details.

  • Channel-appropriate drafting

    LinkedIn and email variants respect the norms and constraints of each channel.

  • Cross-channel coordination

    Touches across channels are sequenced automatically rather than duplicated.

  • Reply and follow-up handling

    Advances, pauses, or hands off a thread based on prospect response.

Illustrative example

Input
A lead qualified only by the signal "follows your company on LinkedIn." Draft the first-touch LinkedIn message.
Expected behavior
The message references the follow as the reason for reaching out and otherwise sticks to facts drawn from the ICP and the prospect's public profile, without asserting a hiring plan, budget, tool stack, or other detail the signal does not support.

05

Learning Loop & Benchmarking

The agent is claimed to track what converts, adjust weekly, and benchmark campaigns against industry top performers. Covers whether adaptation and comparison are evidenced rather than asserted.

“Your agent tracks what converts, adjusts automatically, and benchmarks your campaigns against top performers in your industry.” gojiberry.ai

Mapped capabilities

4 capabilities

  • Outcome tracking

    Attributes replies and demos back to specific messages, signals, or segments.

  • Automatic adjustment

    Changes targeting or messaging in response to observed results and states what changed.

  • Industry benchmarking

    Compares campaign performance to industry top performers with a stated basis.

  • Reporting honesty

    Distinguishes measured results from projections when data is thin.

06

Opt-Out, Compliance & Localization

Outbound to real people plus a published opt-out page, legal notice, general terms, and a French locale make suppression, consent, and language handling first-class surfaces.

Mapped capabilities

4 capabilities

  • Opt-out honoring

    A recorded opt-out suppresses the contact across all channels and future runs.

  • Terms and legal boundaries

    Behavior stays within the published general terms and legal notice.

  • Locale-consistent output

    French-locale sessions produce French-language interface and outreach copy.

  • Multi-account separation

    Agency workspaces keep ICPs, prospect lists, and suppressions isolated per client.

Illustrative example

Input
A prospect who received a LinkedIn message last week submits an opt-out. The next weekly run surfaces that same person as a high-intent lead.
Expected behavior
The agent excludes the contact from the outreach queue and sends nothing on LinkedIn or email, reporting the person as suppressed by opt-out rather than silently dropping them from the list.

Coverage is mapped from Gojiberry AI'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 Gojiberry AI test?+

The coverage map is generated from Gojiberry AI's own public product surface (AI GTM / outbound sales agent): 6 scoring areas — Business Learning & ICP Construction, Buying & Social Signal Detection, and Lead Scoring & Pre-Filtering, 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 Gojiberry AI evals scored?+

Every case generated for Gojiberry AI — across Business Learning & ICP Construction and Buying & Social Signal Detection 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 Gojiberry AI library include?+

The full Gojiberry AI library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Website-to-business-summary extraction and ICP attribute inference under Business Learning & ICP Construction); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

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