Adestto AI
Bots & Tech

Designing an AI trading system: the 6-decision framework

Designing an automated trading system is a sequence of architecture decisions: data, objective function, build vs buy, guardrails, governance. The management framework.

Mahugnon9 min read

Designing a serious automated trading system has almost nothing to do with "finding the right indicator." It's a sequence of architecture decisions — the same ones a leader makes for any critical software product: what data, what objective, what guardrails, what governance. Here is the six-decision framework we use to think about an AI system applied to markets.

Decision 1 — The problem, not the tool

Before talking about a model, define the problem precisely: which market (Forex, gold, indices), what horizon, what edge hypothesis. A system that doesn't know what it's looking for ends up "finding" noise. This boring step eliminates 80% of bad ideas before they cost money.

Decision 2 — Data and its honesty

A system's quality is capped by the quality of its data. Realistic spread, slippage, costs, stress periods included: a backtest on idealized data lies. The management rule: test a system on what it will actually face, not on a perfect world. (See honest backtesting.)

Decision 3 — The objective function

A system optimizes exactly what you measure. If the target is raw profit, it will learn to bet big and blow up one day. A healthy objective function is risk-adjusted: it penalizes drawdown and volatility. Choosing the metric is choosing the behavior.

Decision 4 — Build vs buy

Every organization makes this calculation: build in-house or buy? Building offers control but demands maintenance capacity over time (a system is never "finished"). Buying delegates maintenance but forces you to judge the vendor's credibility: is their method transparent, or a black box? There's no universal right answer — only an answer consistent with your resources.

Decision 5 — The guardrails

An autonomous system without guardrails is a liability, not an asset. The minimum protections:

  • Circuit breaker: daily/weekly loss limits that unplug the system, independent of the model.
  • Risk bounds: capped sizing, no martingale.
  • Drift detection: an alert when the system leaves its statistical envelope.

Guardrails aren't a technical detail — they're what turns a prototype into a system you can leave running.

Decision 6 — Governance and iteration

Who monitors? How often do you re-evaluate? When do you unplug a system that's become obsolete? A mature AI system lives in a loop: measure, evaluate, adjust within a framework, repeat. That's precisely what separates an adaptive bot from a static one left to rot.

The common thread: trust is designed

These six decisions share one point: they're about risk governance, not model sophistication. A modest but well-framed system beats a brilliant one without guardrails. That's the angle Adestto AI applies to its Expert Advisors and its reinforcement-learning R&D: published method, active protections, zero promise of gain.

To see how it translates in practice, read how our bots learn from their mistakes or explore the MT5 bots guide.


Educational content. Adestto AI is a software and educational-content publisher — not a broker or an investment adviser. No return is guaranteed; trading involves risk.

Partager

The conversation continues

A reaction, a question about this article?

Join the Adestto community on Telegram — share your take, your test score or your questions.

Join the channel

About the author

Portrait of Mahugnon H.

Mahugnon H.

CEO · Adestto AI · from Montréal

CEO of Adestto AI. Builds AI analysis tools for the markets and publishes an editorial journal to learn both without kidding himself.

Go further

Learn to trade step by step with the Academy.

3 structured modules, from beginner to advanced. No promise of returns.

Discover the Academy

More into the tools? Discover the Adestto Autopilot