Buyer decision pack

Buyer pack

A compact pack for CX, BPO, speech analytics, and AI product teams deciding whether South African language and accent evaluation is worth a 15-minute problem interview or a small paid pilot.

The first conversation should test pain, budget, evidence needs, and safety boundaries. Voice Bench SA is not selling raw data as the first ask; any paid pilot needs consent and review lineage, purpose, rights, and delivery format agreed up front.

Discuss a pilot See sample report

Inspect now

Proof links

Demo spine: the fake-fixture consent, transcript, review, and export path.
Benchmark Snapshot: a fixture-only sample buyer report.
Task preview: the no-submit contributor task loop.
XPRIZE readiness: the redacted public claim-gate dashboard.

Pilot menu

What can be scoped

Benchmark Snapshot for a quick model or workflow evaluation.
Custom Evaluation Set for a named language, accent, or channel.
Sponsored Contributor Sprint for impact-sourcing and benchmark coverage.
Buyer inputs needed: target scenario, success criteria, rights, and privacy constraints.

Buyer call

Questions to answer

Which South African voice or text scenario fails most painfully today?
Is the need QA, model evaluation, sales proof, compliance, or customer experience?
What result would make a pilot credible enough to buy?
Who signs off on budget, data rights, security, and delivery format?

Demand signals

Public-sector research

0 A bid candidates and 13 B demand signals from a guarded public procurement watch.
Strongest signal: language-access and telephonic-interpretation benchmarking.
Adjacent signals: transcription, language panels, contact-centre modernization, WhatsApp channels, and AI/data services.
No registration, no contact, no tender-pack download, no bid.

Boundary

Evidence before claims

This pack is not evidence of revenue, production users, live Gemini, live Google Cloud, live WhatsApp collection, or a production dataset.
Public-sector demand signals are research only: not revenue, not a bid, not a buyer relationship, and not production proof.
Every paid pilot must define consent terms, review method, delivery format, data rights, and buyer approval evidence.
Voice Bench can run a problem interview now and move to paid delivery only after the relevant collection, QA, and provider gates are safe.
Use npm run email:buyer:pilot:check, npm run revenue:pilot-evidence:check, and npm run xprize:revenue:check before any revenue claim.

Paid pilot menu