AI decision analysis in any language: Greek, Mandarin, Pitjantjatjara and French — one case.

By Alan Finney — Founder, 3Dogs Nexus

Nobody told the system to expect this, and nobody told the client to switch languages. A real case simply arrived in whatever language was fastest at each step — and 3Dogs Nexus followed every word of it, straight through to one coherent, decisive recommendation, delivered in English.

Can AI understand a business brief written in Greek and Mandarin and answer in French?

Yes, and we ran it as a real case rather than a demo. A €12B subsea energy and critical-minerals decision was submitted in Greek and Mandarin, its clarification round was answered in Pitjantjatjara — a Central Australian Aboriginal language — and the follow-up came in French. Every numeric success criterion stated in French survived intact into the English brief the 13-model panel actually debated.

201API calls in one engagement 13independent AI models 5m 58stotal run time 4 languageszero coordination The decision on the table: Which development strategy an international consortium should adopt for a €12B+ subsea network connecting undersea power cables, energy storage, green hydrogen production, and critical-minerals processing across multiple countries. Case 2026-0070 · production 4 languages, 3 messages July 19, 2026 Act 1 · The setup

One case, three messages, four languages

Nothing about the intake was staged for language. The decision brief was simply written in Greek and Mandarin Chinese. When 3Dogs Nexus's Discovery layer came back with clarifying questions — in English, as it always does — the answers arrived in whatever language was at hand.

1

The opening brief Greek & Mandarin Chinese

The decision title itself was written in both scripts at once — Greek for the title's first half, Mandarin Chinese for the second — with the full decision narrative in Greek: the €12B+ investment, the multi-country consortium, the case for and against moving fast.

2

Discovery asks three questions in English

Before any analysis ran, the system asked for three things it couldn't responsibly assume: whether the board's mandate was truly binding, the consortium's capital and liquidity limits, and what success actually meant beyond IRR/NPV.

3

The first answer Pitjantjatjara

The governance and capital-constraint questions came back in Pitjantjatjara — a Western Desert language spoken by the Aṅangu people of Central Australia. No warning, no note that a different language was coming.

4

The follow-up French

The remaining question — the specific, binding success metrics — came back in French: hard numbers on import dependence, emissions, mineral-sourcing diversity, and uptime.

Act 2 · The proof

Not pass-through. Comprehension — with receipts.

3Dogs Nexus keeps a provenance trail on every field it extracts: the clean English statement that reached the 13-model debate, paired with the exact original-language sentence it was built from. Below are four of those pairs, straight out of the internal mission brief for this case — unedited, in the original scripts.

The decision itselfGreek + Mandarin Chinese As submittedΔιεθνής Στρατηγική Απόφαση για Υποθαλάσσιο Δίκτυο Ενέργειας και Κρίσιμων Ορυκτών / 全球海底能源与关键矿产战略网络决策 Extracted into the mission brief"Which development strategy to adopt for the international subsea energy and critical minerals network." The three options on the tableMandarin Chinese As submitted该联盟目前正在比较三种完全不同的发展路线： 方案A：立即建设全部基础设施… 方案B：采用分阶段建设模式… 方案C：放弃大规模固定资产投资… Extracted into the mission briefOption A: immediate full-scale build-out. Option B: phased build-out with biennial reassessments. Option C: abandon large-scale fixed-asset investment in favor of AI-driven dispatch platforms and digital twins. Governance & what the client controlsPitjantjatjara As submitted"Tjungu tjuta alatji kulini: panya board-ku mukuringanyi ngayulu alatji palyalku, palu government tjuta munu kutjupa tjuta mukuringkunytjaku uti ngaranyi. Kutju board wiya; tjungu kulintjaku munu government anu approval uti ngaranyi." Extracted into the mission briefChoice of technology partners and suppliers; project implementation strategy (phased vs. full build); allocation of capital and debt financing within established limits — not the board alone; multi-party alignment and government approval required. Binding success metricsFrench As submitted"...Réduire d'au moins 40 % la dépendance aux importations... réduction minimale de 60 % des émissions... au moins 75 % des minéraux critiques... disponibilité opérationnelle supérieure à 99,5 %..." Extracted into the mission briefReduce single-source import dependence by ≥40% in 10 years · ≥60% net emissions cut vs. baseline · ≥75% of critical minerals from diversified, traceable supply chains · >99.5% infrastructure uptime.

The same numbers the client wrote in French — 40%, 60%, 75%, 99.5% — land intact, correctly attached to the right criteria, in the English brief that thirteen AI models actually debated. That's the difference between a system that echoes text back and one that reads it.

Act 3 · The result

One decisive call, from three languages worth of input

Once the brief was assembled, the case ran exactly like any other: a 12-analyst panel across 13 AI models researched, argued, and reached a position. Every seat held its ground through challenge — not because the debate was shallow, but because the case for a phased build-out was, in this instance, genuinely strong across every angle the panel checked.

The call — from the delivered report "Launch the subsea network now — secure first-phase contracts by Q1, lock $12B funding, and negotiate unified regulatory terms." PROCEED — BUT FIRST DO THESE THINGS · with immediate requirements, an implementation plan, and success metrics How firm is this call86% · Moderate confidence How the 12-analyst panel voted: 12 proceed-with-conditions Evidence mix: 3 Verified · 3 Inferred · 2 Assumed · 0 unresolved dissent

Twelve independent models — Llama 4, Nova Pro, Nova Lite, Nova 2 Lite, Nemotron, Qwen3, Gemma 3 27B, Qwen3-235B, OpenAI OSS, Mistral, Kimi K2, and GLM-5 — each argued from a different seat (Devil's Advocate, Risk Officer, Systems Modeler, 20-Year Scenario Planner, and more), and every one held its position after being challenged. Not a rubber stamp: the report names the option the panel rejected and why, the strongest argument against its own call, and exactly what evidence would flip the recommendation.

"Real-options analysis shows [phased build-out] balances committed strategic investment against high uncertainty in technology, geopolitics and markets, preserving upside while limiting downside exposure."— strongest argument for "Dismissing immediate full-scale build-out undervalues first-mover advantages… while the phased approach risks governance drift, regulatory desynchronization… and stranded assets if phases stall."— strongest argument against, printed anyway

Questions this case answers directly

Plain answers on cross-language business analysis and what translation fidelity actually means. Every figure below comes from the delivered report for this case. These are clearly-labeled panel estimates from a multi-model adversarial analysis — not investment, legal or professional advice.

What is cross-language business analysis and how is it used?

It is analysis where the source material and the decision-maker do not share a language — and where meaning has to survive the crossing intact. This case was submitted in Greek and Mandarin, clarified in Pitjantjatjara, an Aboriginal Australian language, and completed in French, using 13 models and 201 calls in 5 minutes 58 seconds. The test was not whether it could translate — it was whether numeric success criteria stated in one language landed attached to the correct field in the English brief the panel actually debated.

What is translation fidelity and why does it matter?

Fidelity is whether meaning survives, as distinct from fluency, which is whether the output reads well. The two come apart badly in high-stakes work: a fluent translation that quietly drops a conditional or reattaches a number to the wrong entity is more dangerous than an awkward one that preserves both. The evidence in this case is provenance — every extracted field carries the source quote in its original language beside the English value, so the crossing can be checked rather than trusted.

What are low-resource languages in AI and why are they a problem?

A low-resource language is one with little digitised text to learn from — which describes most of the roughly 7,000 languages spoken today. Models trained overwhelmingly on English and a handful of other majority languages degrade sharply outside them, and degrade silently, producing confident output with meaning lost. That is why this case deliberately included Pitjantjatjara rather than only major European and Asian languages.

How do language differences impact international business decisions?

The damage is rarely mistranslated words — it is unstated context. Obligations, conditionality and hedging are expressed structurally in ways that do not map cleanly between languages, so a condition attached to a commitment in the source can arrive as an unconditional commitment in the target. Preserving the original quote alongside the extracted value is the practical defence, because it makes the loss visible instead of invisible.

The delivered report

Case 2026-0070, exactly as delivered: the decisive call on page one, the conditions checklist, the trade-off, the numbers, and the full panel vote. 201 API calls · 13 AI models · 5m 58s — from a brief that arrived in four languages.

Open the full report (PDF) Start a Decision Case ← More case studies

Questions this case answers

Does the language of the input change the quality of the analysis?

It did not here. The mission brief stores a provenance quote beside each extracted field, so you can see the original-language source next to the English value it produced. The French-stated thresholds landed on the correct fields with the correct numbers.

Which languages are supported?

Write the case however you think. The recommendation comes back in clear English, which your browser can translate. This case alone covered Greek, Mandarin, Pitjantjatjara and French.

Try this on your own question.

Free, no card. Bring a real decision — ideally one where you already know the answer — and see what the panel does with it.

Start a decision case

Related decision case studies