Voice & Multilingual Research Capability Brief

VoiceInsights Africa — Voice & Multilingual Research Capability Brief

Voice-first collection, the six-axis language model, transcription and translation governance, human review and low-literacy access — each at its registered status.

Document
VIA-BRIEF-VOICE-MULTILINGUAL
Version
1.0
Audience
Public · Procurement
Published
2026-09-21
Last reviewed
2026-09-21
Language
English (canonical)

The document below is the full text. It is the same content as the PDF, from one source.

VoiceInsights Africa

Voice & Multilingual Research

VoiceInsights Africa

Any voice, any language, answered on six separate questions

“Every Voice. Every Language. Every Insight.”

Version1.0
Published2026-09-21
DocumentVIA-BRIEF-VOICE-MULTILINGUAL
AudiencePublic · Procurement

Why voice-first research

VoiceInsights turns community and field evidence into traceable, decision-ready institutional intelligence.

Who a written instrument excludes

A questionnaire that must be read excludes anyone who cannot read it, in a setting where that is common rather than exceptional. Voice removes the literacy requirement from participation without removing the record.

What a voice answer is worth

Audio is retained with the consent record, bound to the response it belongs to. A transcript is an artefact derived from it, with its own reviewer — not a replacement for the recording.

The full profile. This brief is drawn from the Company & Capability Profile 2026, which carries every capability, its registered status and its limits in full. Where the two are read together, the profile governs.

Voice collection workflows

6 registered routes, three of which carry audio.

Collection routeStatus
Phone CallProven
WhatsAppProven
SMS / Feature-Phone FallbackProven
Web Platform LinkProven
Offline Mobile ApplicationProven
EmailProven

A phone call reaches a respondent with no smartphone and no data. WhatsApp carries text and audio. SMS is the feature-phone fallback. The offline field application records audio where a network does not exist at all.

The six questions a language must answer

AxisWhat it asks
TextCan a respondent answer in writing in this language?
VoiceCan a respondent answer by speaking in this language?
TranscriptionCan that speech be turned into text?
TranslationCan it be translated for analysis?
AnalysisCan free text in this language be analysed?
OutputCan a deliverable be produced in this language?

Each is answered on its own for every language in the registry, because a language can pass one and fail the next. They are never added together into a single number.

What we refuse to say

What we will not tell you about language. No speech-recognition accuracy figure is published for any African language, because none has been benchmarked. The platform performs no automatic language detection — a declared language is what the caller supplied. The number of registered languages is not a capability figure, and the six axes are never added into a single “languages supported” number.
Registered is not the same as production-ready. Registration, measurement and product editions are three different counts, and the live state of every language on every axis is published on the website and read from the capability registry rather than restated in a document where it would age. A buyer asking “do you support my language?” gets six answers, not one.

Translation and transcription governance

Translation is an artefact

A translation is a governed artefact with a recorded reviewer, not something generated silently in the pipeline. A translation that loses a distinction makes a claim the platform cannot support, in a language the reviewer who wrote the original will never read.

Human review

AI may assist transcription and translation; every contribution is candidate intelligence until a named reviewer accepts, modifies or rejects it.

The interface language is a separate question again: which languages your team can work in is not which languages you can collect in, and collection reaches further than analysis does. Three different answers, kept apart.

Evidence, privacy and engagement

Consent in the language it was given

Consent records reflect the collection context, the language, the scope and the withdrawal conditions. Audio is stored under access control with retention and deletion following law, consent and contract.

Applicable law

Implementations are designed to support obligations under Tanzania’s Personal Data Protection Act, No. 11 of 2022, alongside the applicable agreement. Where a programme falls under the EU GDPR or another regime, those obligations are assessed and documented during implementation. No independent assessment of GDPR alignment has been carried out, and none is claimed.

Contact us

“Every Voice. Every Language. Every Insight.”

Tanzania — Head Office

Ali Hassan Mwinyi Road, Kisutu
3rd Floor, Haidary Plaza
P.O. Box 77139
Dar es Salaam, Tanzania

Follow us

LinkedIn
X

VoiceInsights Africa · VoiceInsights Workspace
Voice & Multilingual Research Capability Brief · Version 1.0 · voiceinsightsafrica.com
VIA-BRIEF-VOICE-MULTILINGUAL · Published 2026-09-21 · English (canonical)
Derived from the Company & Capability Profile 2026 and the governed capability registries. 126 master claims audited; 39 excluded as future or withheld.