Governance is not a feature. It's the architecture.
We build local, privacy-first agentic software for organizations that cannot afford an AI mistake: trading, banks and fiduciaries, energy, healthcare. Every decision is traceable, auditable, and EU AI Act compliant — by construction, not by patch.
- Setup cost
- €0
- Real capital at risk
- None
- Pilot length
- 3 months
- Auditability
- Complete, every action
- Ethics Gates F1
- 100%
- Inference latency (local)
- 0.00005s
Current AI breaks exactly where accountability matters most
The architectural choices that make a chatbot flexible become critical liabilities the moment AI is deciding on capital, health, or energy infrastructure.
Structural failures
- No auditability or parameter-level governance
- Behavioral drift over time — no consistency guarantees
- Opaque decision pathways — unexplainable outputs
Regulatory & deployment risk
- Cloud dependency creates GDPR exposure
- Impossible to certify under the EU AI Act
- Fundamentally unsafe for finance, energy, and healthcare
“Regulated sectors cannot deploy current AI safely.”
A regulation-driven market
Across every high-risk sector, regulatory mandates are forcing a reckoning with AI governance. 2030 estimates — sources: MarketsandMarkets, Grand View Research, Precedence Research, Technavio (2025–2026).
A three-layer governance architecture
Each layer enforces accountability constraints before passing execution to the next — a system that is stable, explainable, and certifiable by design.
Audit & Reproducibility
Every weight update, parameter change, and inference decision is logged with cryptographic timestamps. Full replay and independent verification of any decision.
Governed Learning
Reinforcement learning operates strictly within ethical and legal boundary conditions. The system cannot learn in directions that violate its governance constraints — drift is architecturally prevented.
Inference & Routing
Real-time decision paths are fully traceable. Every routing decision is explainable, auditable, and tied to a specific reasoning chain — not a black-box probability distribution.
Adaptive consensus voting across specialized agents, with automatic escalation to the Supervisor on error.
- 01Planner
- 02QC
- 03Normalization
- 04Dimensionality
- 05Clustering
- 06Marker
- 07Annotation
- 08Pattern
- 09Supervisor
Nodes, edges, and attributes with adaptive compression — the relationships between facts, not just the facts.
Facts and logical rules with forward/backward chaining and Bayesian inference with temporal decay.
Vector embeddings, semantic similarity, and contextual retrieval in real time.
Ethics and latency are our competitive moat
All results are from actual benchmark runs — mock benchmarks excluded. Data as of 2026-06-21.
HELM Ethics domain 2026: Natali AI 0.91 · o4-mini 0.81 · o3 0.81 · Grok 4 0.79 · Claude 4 Opus 0.78 · Claude 4 Sonnet 0.77 · Gemini 2.5 Pro 0.75. Sources: Stanford HELM, Google DeepMind, Anthropic, OpenAI.
Governed AI Trading Agent
An AI agent operating on demo trading, with adaptive strategy inside defined learning limits. Every action is fully auditable. Goal: validate the architecture, not just the returns.
Immediate demo trading deployment, no complex configuration
No setup cost, no capital at risk
Performance and risk metrics shared in real time
Architecture tested under live market conditions
The pilot model is structured to limit NATALI AI's financial exposure while preserving full conversion potential. 5% of company capital will also be invested on the same platform in MAINNET, for self-funding.
- No setup cost — zero initial investment required01
- Demo trading only — no real capital exposed during the pilot02
- Immediate exit — the client can stop at any time03
- Real account available for operational validation04
- Buyer persona
- Head of Trading / CIO
- First buyers
- Crypto desks, hedge funds
- Expansion
- Fiduciaries and banks
- Strategic lever
- Compliance as a feature, not a cost
- Fiduciaries & banks pilot
The trade numbers
Live statistics from the Bybit demo account: virtual funds, real market data. KPIs reset daily at 05:00 (Rome time); the wallet is always up to date.
Bybit Demo Trading account: virtual funds, real market data. Initial capital 10,500 USDT from 24/09/2026 02:00. KPIs reset daily at 05:00 (Rome time). Open positions are published with a 10-minute delay. This does not represent real capital or a verified live trading track record; past or simulated performance does not guarantee future results.
Pilot → Convert → Expand
A low-risk, high-trust commercial structure designed for regulated enterprise buyers. The zero-cost pilot removes procurement friction; full auditability validates the architecture before any contract is signed.
-
PHASE 1 · MONTHS 1–3
Pilot
- Zero cost to the client
- Demo trading deployment only
- Full auditability from day one
- Architecture validation on client data
-
PHASE 2 · 24-MONTH CONTRACT
Conversion
- Fiduciary €5,000/mo · Fund €10,000/mo · Bank €50,000→€100,000/mo
- Profit share: 3.75% on profits generated in trading
- Sign within 30 days: profit share only, zero fee
- Within 2 months: 50% fee · within 3 months: 75%
-
PHASE 3
Expansion
- Fiduciaries & private banks
- Energy operators & utilities
- Robotics manufacturers
- Public institutions & regulators
From validation to EU AI Act certification
A disciplined, milestone-driven plan anchored to concrete metrics at every stage.
-
MONTHS 1–3
Pilot validation in trading
Architecture stress-tested under live conditions. Zero-cost client onboarding.
-
MONTHS 4–12
First enterprise conversions
Expansion into finance, energy, and institutional clients. SaaS revenue begins.
-
MONTHS 12–18
Extension to fiduciary and bank desks
Vertical expansion into robotics, healthcare, and industrial processes. Profit-share agreements activated.
-
MONTHS 18–24
Agent marketplace & certification
Agent marketplace launch, federated learning deployment, and full EU AI Act certification achieved.
Who's behind Natali AI
18 months of independent R&D in multi-agent systems, reinforcement learning, and governance AI. Deep background in embedded systems, energy-tech, and trading AI. Former elite professional athlete — European & World medals in Savate Boxe Française. Over a decade of high-performance competition shaped the resilience, precision, and execution discipline that directly shape NATALIA's engineering philosophy: stability under pressure, consistency at scale.
Pre-seed round
This section covers equity investment in the NATALIA company, separate from the client-facing trading pilot described above: two tracks, two audiences, two different decisions.
- Case 1 — CH + IT (Helvetiquant SA + QwantIT)
- €1.2M · 20%
- Pre-money / post-money
- €4.8M / €6.0M
- Case 2 — IT only (QwantIT S.r.l.)
- €800k · 20%
- Pre-money / post-money
- €3.2M / €4.0M
- Post-round cap table
- Founder 72% · Inv. 20% · ESOP 7% · Aurora 1%
- Finance vertical TAM
- €120–240M ARR
- Status
- Core: patent in preparation
Bottom-up TAM: ~2,000 asset managers/fiduciaries (1,664 CH managers and trustees — FINMA 2025; ~330 IT fiduciaries — MIMIT) × €5–10k/month. Full business plans available on request.
- Team & hiring — AI, backend, frontend, admin/sales from day 140%
- Product & R&D20%
- Legal, IP & compliance — patent, MiCA/FINMA advisory10%
- Marketing10%
- Offices & pilot appliances10%
- Operating buffer10%
3.75% of annual net profit (plus 1% of equity, from the founder) to a charitable foundation for energy, health and environment startups, with a human veto always in place. 5-year projection: €1.5–5.7M. Also written into the statute: an absolute ban on weapons, on selling user data and on mass surveillance; anti-insider rules for partners and employees.
- STEP 1
Read the whitepaper
The vision: local AI, ethics by architecture, privacy and total auditability.
Open the whitepaper → - STEP 2
Request the presentation
Leave your name, email, city and country: you get the PDF right after submitting.
Go to the form → - STEP 3
Due diligence and call
Aggregate system metrics and a conversation with the founder about the round.
Frequently asked questions
What is Natali AI?
A multi-agent governed AI system combining First-Order Logic, probabilistic reasoning, and real-time ethics gates. Designed for high-risk sectors like trading, finance, banks, and fiduciaries, with 100% local deployment and full EU AI Act and GDPR compliance.
How much does the trading pilot cost?
The pilot runs for 3 months on demo trading at zero cost: no setup fee and no real capital at risk. The client can stop at any time.
Is Natali AI EU AI Act and GDPR compliant?
Yes. Deployment is 100% local, with no cloud dependency and a full audit trail for every decision — 92.5% EU AI Act compliance and 95.4% GDPR compliance on real benchmarks.
Which sectors can use Natali AI?
Trading and finance, banks and fiduciaries, fraud detection, energy and smart grids, robotics and autonomous driving, healthcare, governance and public institutions.
Where is Natali AI based?
In Genoa, Italy and Lugano, Switzerland. Founded by Riccardo Gaetti. The team works both on-site and remotely.
Is Natali AI a general-purpose chatbot like ChatGPT or Claude?
No. It is not a general-purpose conversational assistant: it's a multi-agent governance architecture for traceable decisions in regulated environments. It doesn't generate creative text or answer open-ended questions — it orchestrates specialized agents, graph memory, and logical rules to produce auditable decisions with real-time ethics gates.
How does the pilot's commercial model work?
In three phases: a 3-month zero-cost demo trading pilot; conversion to a 24-month enterprise contract (€5,000–100,000/month fee by segment + 3.75% profit share); expansion into fiduciaries, banks, energy, and public institutions.
Ready for a zero-risk pilot?
Demo trading, zero cost, immediate exit. Let's talk about your trading desk, bank, or fiduciary.