Case study · AI research for finance

How to invest around the socio×geo×political forces of 2026.

A worked case study for a Romanian retail investor trading through BT Trade. It starts from the socio-economic, geopolitical and market forces that move a portfolio in 2026 (war, sticky inflation, a wobbling AI boom, a hawkish rate regime, Romanian politics), then follows them through to concrete, buyable decisions. Built from a Romanian + European-listed universe; the unit is % of capital across a €20–90k band, not a single number.

This is not financial advice.

Every insight and number here came from an AI research pipeline. See how it was built · why each call was made.

The takeaway: five strategies, one thesis

Everything in this study lands on five concrete portfolios, arranged in a 3×2 grid: three risk tiers × two capital bands, with one cell deliberately empty. ④ Balanced is the base case. Pick your cell (capital first, then risk) and tap any card for its split, its lines and its buy list.

The forces, at a glance

Every call starts from 23 dated factors across five layers: Romanian politics, global macro, the structural AI shift. The geopolitical layer is on the map below; tap any factor to expand.

Hover or tap a country, or a numbered factor. The full interactive map with the market-trends snapshot is on Factors →

All 23 factors, across five layers →

How to read this report

Three things to hold in mind before you dive in: what it's for, the world it's set in, and how it was built. Then follow the nav: Factors → Reaction Map → Strategies, with the reasoning on Method & Why.

The goal

Structure a mid-size portfolio (€20–90k) for a Romanian retail investor on BT Trade, 2026, and ship it as a reproducible worked example of using AI to research a market. Fictional and educational; not financial advice.

The context

Sticky RO inflation (10.4%, above almost every safe yield), a war-driven defense build-out, an AI melt-up now re-rating, a hawkish ECB and a Romanian sovereign under review. Universe: RO + EU-listed only. No US venues, no gold, no crypto.

The approach

A wide factor map → parallel deep research → a 23×11 reaction map → a four-voice persona debate → five strategies. Evidence over assertion: every material claim traces to a dated source. Method & why →

The factor reaction map

We mapped 23 factors (war, rates, AI, energy, policy) against eleven investment directions: who benefits and who suffers, cell by cell. This is the compact corner: the nine most decisive factors × the nine most-reactive directions, where green is a tailwind and rust a headwind.

++ strong benefit + benefit ± two-sided − hurt −− strong hurt

Read a row to see how one force ripples across the directions (EU rearmament lifts defense, leaves broad equity mixed, weighs on bonds). This 9 × 9 is a slice of the full 23 × 11 interactive grid, every factor wired to what the four personas say: the Reaction Map →

Key findings: where to dig deeper

The whole study compresses to four findings, each linked to where it's argued in full.

Central finding

Safe money loses to inflation

RO inflation (10.4% y/y) sits above almost every safe RON yield, and the end-2026 forecast is ~5.5%. Safe RON paper stays real-negative, so real assets (European equity) carry the mandate.

Why safe yields fall short → The one exception

Tax-free Fidelis EUR clears the bar

A 6.20% tax-free EUR coupon against a 5.5% forecast: the only line in this study that beats inflation after tax, in EUR, without equity risk. It anchors every tier's bond floor.

See the bond floor → What actually binds

The platform fee grid, not the AI ceiling

BT's ~€20 fixed per-order fee makes a €400 line cost ~5% at €20k but ~0.7% at €50k. Line count scales with capital, not risk appetite. That's why there are five strategies and one empty cell.

Why five strategies → The reasoning

Six decisions, a four-voice debate

Bond floor, RO-vs-Europe, the satellite, how much AI, how many lines, and gold/cash/crypto: each argued for and against by a skeptic, opportunist, conservator and macro strategist.

Read the decisions & debate →

Stress-tested: what if the AI bubble pops? We put the portfolio through a Burry-style bear case (datacenter capex ahead of revenue, depreciation flattering profits), and by mid-2026 parts of it are observed, not hypothetical (TSMC posted record revenue and fell anyway on capex guidance). The most-evidenced pop is a margin/write-down re-rating, and because a "diversified" core is already ~25–30% AI, the core and any AI sleeve fall together. The first-line defense is free: total-AI sizing, netted, kept well under the ceiling. That's exactly why the strategies sit at ~3.5–9.8% AI, not at 15%. See the full stress test →

This is a worked example: reasoned, conditional, and fully sourced; not financial advice. The reasoning behind every call is on Method & Why; the from-scratch rebuild is under Reproduce it.