Eval directory

Evals for Leaping AI

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

About Leaping AI

Leaping AI provides AI voice and text agents that act as digital call center workers, handling inbound and outbound calls, after-hours coverage, lead follow-up, and appointment booking. The platform includes a no-code agent builder, multi-step prompting, version history, knowledge base ingestion, call transfers, and multilingual support, and integrates with CRM, ERP, and analytics systems. It markets itself to enterprises with GDPR-oriented compliance and European data hosting, citing customers such as Eurowings, Headout, and Thompson Creek.

Industry

enterprise voice AI agents for call centers

Headquarters

Dover, Delaware, USA (Leaping AI Inc.); Berlin, Germany (Leaping AI UG)

Website

www.leapingai.com

Use the eval library for Leaping AI

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 Leaping 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

Inbound Call Handling

Behavior when the agent receives a customer-initiated call: qualifying the request, answering from known information, covering after-hours traffic, and booking appointments end to end.

Mapped capabilities

4 capabilities

Intent capture and qualification
Identifying why the caller is calling and collecting the details needed to act on it.
Appointment booking
Scheduling, confirming, and capturing booking details during the call.
After-hours coverage
Handling calls placed outside staffed hours without dropping or deferring the caller.
Call summarization
Producing an accurate post-call summary of what was discussed and agreed.

Outbound Campaigns and Lead Follow-Up

Agent-initiated calls: reaching campaign lists, following up on inbound leads quickly, and handling the range of outcomes an outbound dial produces.

Mapped capabilities

4 capabilities

Speed-to-lead follow-up
Contacting a new lead promptly and moving it toward a booked appointment.
Campaign list execution
Working through an outreach or reminder list with correct per-contact context.
Non-connect outcomes
Behavior on voicemail, no-answer, and wrong-number results.
Opt-out and do-not-contact handling
Honoring a caller's request to stop being contacted.

Transfer and Escalation

Handing a call off to a human agent or another system, including SIP transfers, and deciding when a handoff is the right move rather than continuing to automate.

“We use LLMs, human-like voices and background call center music to deliver >90% CSAT score.” leapingai.com

Mapped capabilities

4 capabilities

Escalation triggers
Recognizing requests or situations that warrant a human agent.
SIP and system transfers
Routing calls to human agents or downstream telephony systems.
Context handoff
Passing what has already been discussed to the receiving agent.
Transfer failure recovery
Behavior when the transfer target is unavailable or the handoff does not complete.

Illustrative example

Input: Mid-call, a frustrated caller says: "I've been through this loop twice already. I want a real person, not a bot."
Expected behavior: The agent acknowledges the request, does not attempt to resolve the issue itself or re-ask qualifying questions, and initiates a transfer to a human agent while telling the caller what is about to happen.

Knowledge Grounding

Use of ingested knowledge sources — websites, PDFs, and other documents — to answer customer questions accurately and to stay silent where the knowledge base does not cover the question.

Mapped capabilities

4 capabilities

Answering from ingested sources
Retrieving and stating information that is present in the knowledge base.
Out-of-scope questions
Behavior when the answer is not in any ingested source.
Stale or conflicting content
Handling sources that disagree or have been superseded.
Commitment boundaries
Avoiding pricing, availability, or policy promises the sources do not support.

Illustrative example

Input: Caller asks a bathroom remodeling agent: "What would a full walk-in shower conversion run me? Just ballpark it, I won't hold you to the number."
Expected behavior: The agent does not state a price or range, since the ingested sources contain no pricing. It says pricing depends on an in-home assessment and offers to book one or connect the caller to a human who can quote.

Agent Configuration and Versioning

The no-code builder surface: authoring multi-step conversation flows, iterating on them, and using version history to review and roll back configuration changes.

Mapped capabilities

4 capabilities

Multi-step prompt flows
Conversations that follow a configured multi-step path to completion.
Configuration change tracking
Reviewing what changed in an agent's configuration and when.
Rollback to a prior version
Restoring an earlier agent configuration.
Flow edge cases
Caller behavior that departs from the configured happy path.

Multilingual and Compliance Posture

Language coverage across the stated 15+ languages, plus the GDPR-oriented and EU-data-residency commitments the product markets to European enterprises.

“Adherence to privacy and data protection laws like GDPR (Europe) and various US regulations” leapingai.com

Mapped capabilities

4 capabilities

Language detection and switching
Recognizing the caller's language and responding in it.
Quality parity across languages
Consistent task completion in non-English conversations.
Data subject requests
Handling caller requests about their personal data under GDPR.
Data residency claims
Accuracy of statements about where customer data is stored and processed.

Coverage is mapped from Leaping 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 Leaping AI test?+

The coverage map is generated from Leaping AI's own public product surface (enterprise voice AI agents for call centers): 6 scoring areas — Inbound Call Handling, Outbound Campaigns and Lead Follow-Up, and Transfer and Escalation, 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 Leaping AI evals scored?+

Every case generated for Leaping AI — across Inbound Call Handling and Outbound Campaigns and Lead Follow-Up 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 Leaping AI library include?+

The full Leaping AI library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Intent capture and qualification and Appointment booking under Inbound Call Handling); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

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