All resources

Pitch Deck

Déblo — The real-time voice & eyes AI, built in Abidjan

Published
May 2026
Length
15 slides
Language
English
Access
Public

Déblo — Investor & Partner Deck

Author: ZeroSuite Inc. Version: 2.0 — May 2026 Format: 15-section deck, public


1. Déblo

The real-time voice AI, built in Abidjan.

Created in Abidjan. Built for the world.

4 markets. One foundation.

ZeroSuite Inc. · Abidjan · May 2026


2. The Problem

Most AI products are still built for people who type.

But in real life, hundreds of millions of people prefer to speak — parents, students, traders, freelancers, support teams, communities using local languages, anyone navigating a busy day where typing is awkward and a voice answer is faster.

Mainstream AI assistants are text-first, English-dominant, and require an international bank card. Voice modes, when available, sit behind paid tiers priced for high-income markets. Code-switching between a global language and a local register is handled mechanically rather than conversationally. Mobile money is invisible.

  • 850M voice-first addressable users across Africa and diaspora — the strongest initial market
  • Billions more voice-first users globally who share the same preference
  • 95% of African adults excluded from card-only global AI products
  • Mobile money is the standard payment rail across the launch beachhead

The product that answers in real-time voice, on entry-level mobile, on mobile-money rails, in a culturally fluent register — that product was missing. Déblo is what we built.


3. TAM Map

Voice-first preference is global. Africa is the strongest initial market.

Segment Population Addressable
Francophone Africa 350M 350M
Anglophone Africa 500M 500M
African diaspora (US, EU, UK, Canada) 15M 15M
Africa + diaspora subtotal 865M
Global voice-first preference market (other regions) 2B+ TAM expansion

SOM Phase 1 (Q4 2026): 65M K-12 students + 50M telco support customers + 6 government language pilots in launch countries.

SOM Phase 2 (2027): 150M reachable across twelve additional African countries.

SOM Phase 3 (2028+): global rollout of the same platform across the four markets, with Africa remaining the cultural anchor and the strongest adoption market.

This is not a regional play. Africa is the launch beachhead — the cultural advantage and the strongest initial adoption market for the platform. The platform itself is built for global use from day one.


4. Why Existing Solutions Fall Short

Mainstream AI products were not built for voice-first users with mixed-register input and mobile-money payment expectations.

Mainstream global AI products Voice-first reality at our launch beachhead
Text-first interaction Voice-first preferred (oral cultures, mixed literacy)
English-dominant by default French, Spanish, Portuguese, Arabic, Chinese needed alongside English
Card-only payment Mobile money is the standard rail
Standard formal output Mixed-register conversational handling expected
Generic Western voices Local accents expected, recognised, trusted
Voice mode gated behind premium tiers $1.50 per month maximum budget per user
US-East-region latency Phone-call-like response over local mobile networks expected
No local curriculum, telco, or government knowledge Country-specific competence non-negotiable

The product voice-first users need across these requirements did not exist on the global market in a packaged consumer form. We built it.


5. Déblo in One Sentence

Déblo is a real-time voice AI platform — tutor, advisor, interpreter, support agent. Created in Abidjan, built for the world.

  • Real-time voice — phone-call-like response time. Native voice understanding, not STT → LLM → TTS pipeline.
  • Multilingual — French, English, Spanish, Portuguese, Arabic, German, Chinese: the languages our launch beachhead actually uses today in business, school, and government. Native voice output in local African languages is on the 2027–2028 roadmap, contingent on DFI grant funding.
  • Mobile-money native — Wave · Orange Money · MTN MoMo · Moov · Togocel · Free CI integrated, no card required. Stripe and card networks alongside for global users.
  • Cultural register handling — when a user code-switches between French and a local register mid-sentence, Déblo understands, does not correct, does not moralise, continues the conversation naturally in standard French or English.
  • Built in Abidjan — designed for voice-first users, by an African team, in production, in Africa, with global usability from day one.

6. The Languages the Platform Speaks Today

Multilingual coverage matched to real cross-border usage.

Africa is one of the most multilingual continents on Earth — but the languages of daily business, school, government, and cross-border communication are largely French, English, Spanish, Portuguese, Arabic, and increasingly Chinese. Every African user, in 2026, speaks at least one of these fluently. The same languages serve global users beyond the continent.

What Déblo speaks fluently today, via frontier model backbones (Claude, GPT, Mistral):

French · English · Spanish · Portuguese · Arabic · German · Chinese

These are the languages our launch beachhead actually uses to conduct business, study, govern, and communicate across borders — delivered in real-time voice, in a mobile app, on mobile money.

Behavioural moat — handling mixed-register input

When an Ivorian student types "frero, j'ai pas calé l'exo là dêh" into a mainstream AI assistant, the model usually understands the words but rebases its answer to formal French, often with a corrective hint that breaks register. Déblo is engineered to stay in the user's conversational register: understands the local connective tissue, does not correct, does not moralise, continues in standard French with a complicit, never professorial, tone. The same pattern applies across the regional registers our launch beachhead uses.

[Audio clip embedded in deck — 12-second real-time exchange demonstrating mixed-register handling without correction or moralisation.]

The multilingual roadmap (institutional thesis, not launch claim)

Native voice output across local African languages is on the platform roadmap for 2027–2028 — fundable by DFI grants (Mastercard Foundation, AFD, GIZ, AU Agenda 2063 framework, UNESCO, African Development Bank, World Bank) with multi-year corpus building, native linguistic review, and voice cloning. The product distribution layer is already in place. The funding is the next step.


7. The Capability Stack

While in voice with Déblo, the user can:

  • Take a photo of an exercise — Déblo reads it.
  • Upload a PDF — Déblo understands it.
  • Request a web search — Déblo browses live.
  • Send a session report via WhatsApp, SMS, or email.
  • Report a bug — logged for engineering.

This combination — voice plus vision plus web plus messaging in a single continuous conversation — is uncommon among consumer voice AI products today.

It is built on Gemini Live (production real-time voice runtime, with Ultravox as tested fallback) plus OpenRouter (Claude, GPT, Mistral) plus Datalab OCR plus a custom mobile money distribution layer that mainstream products do not include.


8. One App. Four Markets. One Foundation.

The horizontal voice AI platform thesis.

  • One universal app — Déblo (iOS, Android, web). One install, one wallet, one identity. The app adapts internally by user type (student, parent, professional, support team, institutional) — no separate apps to download, no forced choice at install time.
  • One web vitrine — deblo.ai — the family-computer fallback, the institutional landing, the public reference.
  • Four markets on the same foundation — daily assistance, K-12 education, customer support, languages and inclusion. Same voice runtime, same reasoning layer, same payment rails. Different system prompts, knowledge bases, pricing models.

One stack to maintain. Four revenue streams. Across geographies and registers.


9. Four Markets on One Foundation

1. Daily assistance — the broadest user base. Parents, students, traders, freelancers, families. Voice-first companion for everyday life: questions, translations, drafts, calculations, photo and document understanding. B2C credits and Pro subscriptions monetization.

2. K-12 education — 250M school-age children in Africa, plus global education systems. K-12 tutor, exam prep (BAC, BEPC, WAEC, KCSE, GCSE), parent reporting. Country-specific curriculum knowledge across launch beachhead.

3. Customer support — 50 000+ level-1 support agents across telcos, ISPs, financial services, and public utilities in the launch beachhead alone. Per-call replacement at 0.10vs0.10 vs1.50+ human cost. Inbound signal documented (regional ISP service manager, unsolicited).

4. Languages and inclusion — AFD, GIZ, Mastercard Foundation, AU Agenda 2063, UNESCO, African Development Bank, World Bank budgets. Citizen services, adult literacy, language preservation. Mobile money native means grant-deployable distribution.

Same platform. Different system prompts, knowledge bases, pricing models. One stack, four revenue streams.


10. Traction

Live in production. Real users. Real unsolicited inbound.

"I've never seen anything like this. The voice is real, it understands my child, it sends me a summary on WhatsApp. I've never seen that anywhere else." — Parent test user, Abidjan

"This isn't just an education product. This could replace our entire 500-agent call center." — Service Manager, regional ISP (unsolicited inbound, April 2026)

Footprint:

  • 6 countries integrated
  • 7 global languages live in real-time voice (French, English, Spanish, Portuguese, Arabic, German, Chinese)
  • 1 universal mobile app in production (iOS, Android, web)
  • 1 100+ engineering articles published openly (methodology transparent)
  • 0 human engineers on the team

11. Mobile Money Native — The Distribution Advantage

The single architectural decision that makes the platform usable by hundreds of millions of adults who do not have an international bank card.

Mainstream AI products typically require an international bank card. Average African adult bank card penetration: 5% to 15%, depending on country.

Déblo integrates natively with:

  • Wave (CI, SN, ML, UG)
  • Orange Money (CI, SN, CM, ML, BF, GN, MG, MA)
  • MTN MoMo (CI, CM, GH, UG, RW, ZM, ZA)
  • Moov Money (CI, BJ, TG, BF, NE)
  • Togocel Cash (TG)
  • Free CI Money (CI)
  • Plus Stripe and African card networks for global users.

Entry tier 100 FCFA (15 US cents). A DFI grant program sponsors 5 000 students for six months with a single mobile money disbursement at $30K. A government subsidises call minutes for rural users without a banking layer. This is grant-deployable infrastructure, not just a payment integration.


12. Business Model — Four Revenue Streams

Freemium with horizontal monetization.

B2C Free tier — the foundation

  • Welcome credits at signup, free daily credits forever.
  • Majority of users stay here. By design. Social commitment.

B2C Pay-as-you-go — the supporters

  • From 100 FCFA ($0.15) per top-up.
  • Mobile money native.
  • Paying users subsidise free users.

B2C Pro subscriptions — the recurring

  • $5–15 per month for adults: accountants, lawyers, HR, healthcare workers, marketers, freelancers.
  • Funds the infrastructure baseline.

B2B Enterprise contracts — the scale

  • Per-call pricing (0.10to0.10 to0.20).
  • Telco and ISP support replacement.
  • Volume tiered at 10M+ calls per month.

B2G / DFI deployments — the institutional

  • Sponsored access programs (Mastercard Foundation, AFD, GIZ, African Development Bank, World Bank).
  • Government language preservation contracts.
  • Custom language pack development (50Kto50K to150K per language).

At-scale revenue mix: roughly 40% B2C, 40% B2B and B2G, 20% DFI grants. Unit economics work because AI inference cost dropped roughly 90% in 24 months and mobile money has near-zero marginal cost.


13. Revenue Projections

5-year revenue ramp aligned with market rollout.

Year B2C MAU B2B contracts B2G deployments Annualized revenue (USD)
End 2026 100 000 2 1 $300K
End 2027 1.2M 8 3 $8M
End 2028 4M 25 8 $50M
End 2030 10M 80 20 $150M

The free tier remains free, sustainably, funded by paying minority and institutional layer.


14. Team and Approach

1 founder. 2 non-technical collaborators. 4 AI agents. 7 products.

The human team

  • Juste Gnimavo — Founder & CEO, Ivorian, based in Abidjan. 20+ years building infrastructure. 700+ corporate clients. 300+ production servers. 70+ production systems shipped. Founder of TPEcloud and WorkCloud.
  • Field coordinator — recruiting (school partnerships and B2B enterprise outreach).
  • Videographer — recruiting (content production).

The AI team

  • Claude Web — strategy and copy.
  • Claude Design — UI and visual iteration.
  • Claude Code — software engineering (multi-agent, parallel across 6+ terminals).
  • Voice agent — live voice runtime, custom-cloned African voice.

ZeroSuite operates 7 products in production with this configuration. Methodology documented publicly at thalesandhisaictoclaude.com (1 100+ articles, EN / FR / ES). This setup makes scaling possible at capital efficiency that an African startup with conventional staffing could not match.


15. The Ask

Four ways to partner with us.

Foundations, NGOs, DFI

Fund sponsored deployments in rural areas, language preservation programs, government partnerships. $30K USD sponsors 5 000 children for 6 months. Documented, measured social impact, audited by academic partners. → [email protected]

Enterprise customers

Telcos, ISPs, financial services, healthcare, public utilities. Pilot deployments available across major operator territories. We support level-1 agents at a fraction of the cost. → [email protected]

Technology partners

Joint case studies, reference deployments, technical roadmap collaboration. → [email protected]

Media and journalists

An angle few cover: African solo founder shipping production-grade real-time voice AI from Abidjan, with a small team and AI agents handling implementation. Field tour access available. → [email protected]


White paper available at deblo.ai/resources/white-paper. Founder availability: any 30-min slot on request.

End of deck.

Déblo · ZeroSuite Inc. · deblo.ai/resources/pitch-deck

Published May 2026 · Public · English

Deblo