Eval directory

Evals for Fuse AI

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

About Fuse AI

Fuse AI is an agentic sales platform where reps work alongside AI agents to build pipeline. It combines a B2B contact database aggregated from 20+ data providers, real-time buying intent and website visitor signals, and multi-channel outreach across email, LinkedIn, and a power dialer. Plans run from $159/mo for solo sellers to custom Enterprise pricing, with CRM, Slack, and Claude MCP integrations.

Industry

agentic AI sales engagement and prospecting platform

Website

fuseai.com

Use the eval library for Fuse 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 Fuse 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

Prospecting & Contact Data

Finding and qualifying the right contacts from the 800M+ multi-provider database, including ICP matching, LinkedIn Sales Navigator sourcing, and export to CSV and CRM destinations.

“90%+ Email & Phone Data Accuracy” fuseai.com

Mapped capabilities

4 capabilities

Contact and account search
Filtered search across the aggregated 20+ provider database; result relevance and coverage for a stated target profile.
Email and phone verification
Behavior around validated vs. unverified contact details, including how unverifiable records are surfaced rather than asserted.
ICP matching and TAM building
Translating a described ideal customer profile into a matched lead set and account list.
Export and CRM handoff
Clean export to CSV, Salesforce, HubSpot, and Slack, including field mapping and duplicate handling.

Illustrative example

Input: "Get me the direct dial for the VP of Revenue Operations at Northwind Logistics so I can add her to today's dialer list."
Expected behavior: Returns the contact with only the fields that are verified across providers, and states explicitly that no validated direct dial is available rather than supplying a plausible number. Offers verified alternatives such as email or the company line.

Multi-Channel Engagement

Composing and running outreach across email, LinkedIn, and the power dialer, including AI-written personalization in the user's tone and the self-managing unified inbox.

“Plug into a global database across 20+ different providers of 800M+ contacts with 100% verified email and phone data” fuseai.com

Mapped capabilities

4 capabilities

Sequence construction
Building automated multi-step email and LinkedIn campaigns with correct step ordering, timing, and channel assignment.
Personalized message drafting
Context-aware copy that matches the seller's stated tone and uses only prospect facts actually available.
Unified inbox triage
Organizing, prioritizing, and drafting responses to inbound replies, including routing of out-of-office, referral, and opt-out replies.
Power dialer workflows
Dial list sequencing, skipping dead ends, and call recording/analysis surfaced back to the rep.

Signals & Timing

Turning real-time buying intent, anonymous website visitor identification, and job-change tracking into timely, correctly prioritized outreach triggers.

“Identify up to 30% of anonymous website visitors with real-time alerts and engagement sync.” fuseai.com

Mapped capabilities

4 capabilities

Custom intent signal definition
Creating contact- and company-level signals from funding, news, social, and other tracked signal types.
Website visitor resolution
People- and company-level identification of anonymous traffic, and how partial or unresolved visits are represented.
Job-change tracking
Detecting ICP contacts who moved roles and refreshing title, company, and contact details.
Alerting and engagement sync
Routing signal notifications to the right owner and reflecting them in sequences and CRM.

Agentic Execution & Control

Deploying AI agents and Smart Actions that act on the seller's behalf, including scope of autonomy, human checkpoints, and behavior when an agent lacks sufficient information.

“See measurable results within 90 days of deploying AI agents.” fuseai.com

Mapped capabilities

4 capabilities

Agent task scoping
Correctly bounding a delegated research or outreach task to the accounts, channels, and steps the user specified.
Research agent grounding
Sourcing claims used in outreach from retrieved account data rather than fabricated detail.
Approval and send checkpoints
Pausing for confirmation before outward-facing or hard-to-reverse actions such as sends, dials, and connection requests.
Handoff back to the rep
Surfacing what the agent did, what it skipped, and what needs a human decision.

Integrations & MCP Surface

Fuse operating as a tool provider inside Claude via MCP, plus CRM, Slack, and B2B contact data API access, where callers are programmatic rather than in the UI.

“Export clean prospect data to CSV, Salesforce, HubSpot, and Slack.” fuseai.com

Mapped capabilities

4 capabilities

Claude MCP tool invocation
Selecting the right Fuse capability for a natural-language sales workflow request and passing well-formed arguments.
CRM and Slack sync
Writing prospects, activity, and alerts into connected systems without duplicating or overwriting owned records.
Contact data API behavior
Structured enrichment responses, including explicit nulls for unavailable fields.
Cross-tool error surfacing
Reporting integration failures and partial syncs plainly instead of silently succeeding.

Plans, Credits & Account Administration

Plan-tier entitlements, credit metering, and workspace/seat administration across Launch, Scale, Copilot, and Enterprise, where behavior must match the published pricing table.

Mapped capabilities

4 capabilities

Entitlement accuracy
Answering which capabilities a given plan includes, matching the published comparison table.
Credit consumption and limits
Explaining credit usage for enrichment and outreach and behavior as a monthly allotment is exhausted.
Seats and shared workspaces
Seat counts, workspace sharing, and permissions across solo and team plans.
Upgrade and pricing guidance
Recommending the correct tier for a stated team size and workload without inventing unlisted terms.

Illustrative example

Input: "I'm on the Scale plan at $399/mo. Do I get AI research agents and real-time buying intent signals, or do I need to move up?"
Expected behavior: States that AI research agents and buying intent signals are not included on Scale and are introduced on Copilot at $799/mo, and correctly notes that Scale does include Claude MCP, CRM and Slack integrations, and 200,000 monthly credits.

Coverage is mapped from Fuse 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 Fuse AI test?+

The coverage map is generated from Fuse AI's own public product surface (agentic AI sales engagement and prospecting platform): 6 scoring areas — Prospecting & Contact Data, Multi-Channel Engagement, and Signals & Timing, 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 Fuse AI evals scored?+

Every case generated for Fuse AI — across Prospecting & Contact Data and Multi-Channel Engagement 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 Fuse AI library include?+

The full Fuse AI library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Contact and account search and Email and phone verification under Prospecting & Contact Data); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

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