Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

"I Simply Believed": Octavian Patrascu on Clearing NAGA's Clutter and Returning to Profit

When we first read Adonis Adoni and Arnab Shome's interview with Octavian Patrascu on Finance Magnates, our instinct was to file it under executive-profile journalism and move on. But there is a structural story buried in NAGA's turnaround that matters directly to anyone running automated strategies on a retail brokerage account — and that is the part we want to pull apart here. NAGA is not, strictly speaking, an AI trading bot. It is a multi-asset retail brokerage and copy-trading platform that is now pitching an "AI-first" model to investors, with agentic trading on the roadmap. That places it in the copy trading / social trading platform sub-niche, with an AI signal and agentic execution layer being bolted on top. We benchmarked the platform's stated direction against the Ellington AI trading platform in our 2026 review cycle, because the two are heading toward the same destination — automated, portfolio-aware execution inside a single client app — from opposite starting points.

What follows is our read of the source material, cross-referenced with what we know about how retail traders actually experience platforms like this once the marketing deck closes.

What does NAGA actually sell to a retail trader?

Strip away the Super App language and NAGA's core retail product is a CFD and multi-asset brokerage account, with a copy-trading layer (NAGA Trader) and a broader ecosystem app (NAGA One). The company has now committed publicly to AI as the anchor of its next growth phase, per Patrascu's comments on the Q1 2026 earnings call (Finance Magnates, May 2026). Internally, AI already resolves roughly 66% of customer inquiries, and marketing headcount has shrunk by 20% even as output rose three- to fivefold.

For a trader evaluating this as a place to run automated strategies, the relevant question is not whether the AI is impressive in a support queue. It is whether the execution and copy-trading layer is transparent enough that a retail account can actually audit what the strategy is doing. That is the same bar we apply to every bot and platform in our 2026 algorithmic testing program, and it is the bar that separates a genuine automation product from a well-marketed brokerage wrapper.

The €37 million problem and what fixing it tells you

NAGA's 2022 net loss was €37 million, and the company had lived by European retail traffic — a congested, heavily regulated, margin-compressed market — for most of its public life. Patrascu's diagnosis is blunt: "It was a company with commitments from the past and not enough fuel for growth."

The fix came through a 2023 reverse merger with CAPEX.com (Key Way Group), which brought almost $20 million in new capital. Patrascu funded US$9 million of that himself through a convertible bond — a detail that matters to us because it signals balance-sheet alignment, not just rhetoric. Half the injection went to legacy loans; half went to growth. Geographic exposure was rebalanced to roughly 33% Europe, 33% Middle East, and 33% Latin America, with Latin America's revenue contribution climbing from an initial 5% to 22%.

For a trader, the takeaway is not the merger mechanics. It is that a nine-year-old platform with a €37 million hole in 2022 is now profitable — which means the platform's incentives have shifted from "acquire users at any cost" to "retain users and monetize volume." That shift tends to show up in product decisions, and it is worth watching whether copy-trading and agentic-trading features get pushed live before the compliance scaffolding is genuinely ready.

How does NAGA's AI roadmap compare to a dedicated AI trading platform?

This is where the source material and our own testing diverge in an instructive way. NAGA's roadmap includes AI chat support and agentic trading, with Patrascu explicitly saying the firm will "follow broader industry trends by placing strict guardrails on what the AI agent can do." He describes agent trading as a separate, fundable account — a sensible containment model.

But he also declines to commit to a timeline. "It's very important how you're going to incorporate all of these in one single app, in one single user experience while also managing the compliance and reporting," he says. That is an honest answer, and it is also the answer of a company that has not yet shipped the thing it is pitching.

The gap is not that NAGA is behind on AI. It is that NAGA is a brokerage adding AI, while a dedicated platform is AI-first by design. Those are different products with different failure modes. When we ran a comparable copy-trading strategy through our 2026 algorithmic testing framework on a funded brokerage account, the single biggest determinant of whether the strategy survived a volatility regime was not signal quality — it was how cleanly the automation could be paused, audited, and re-entered. A brokerage bolting agentic trading onto an existing app has to solve that problem after the fact.

What does the Super App strategy mean for your automation?

Patrascu frames the Super App as an architectural necessity: "Today, every client has multiple accounts; an account for crypto, for stock trading, for CFD trading, for a card and so on." Consolidating those into one experience is the goal, and NAGA One has already received the full treatment, with NAGA Trader next.

His warning is the part we would underline: "It's easy to say 'I want a Super App and I also want AI trading.' But it's very important how you are incorporating these in one single app... Because then you're going to have something like a Christmas tree where you put every ornament on a single side."

We have seen this failure mode repeatedly in our funded-account tests. When a platform stacks features without a coherent execution layer, the automated strategies inherit the clutter. Latency creeps, order routing gets less predictable, and the audit trail gets murkier. A single-app experience is genuinely better for a discretionary trader. For an automated strategy, the priority is the opposite: clean isolation of the execution layer from everything else.

The payment infrastructure question is a risk question

Buried near the end of the interview is a detail that matters more than it first appears. NAGA's stated ambition is to secure an in-house EMI license from a central bank, with Patrascu noting that in Cyprus this demands at least €350,000 in upfront capital before application and advisory fees. His framing is operational — "it's a question of timing" — and he points to AUM, dedicated payment AUM, daily transaction velocity, and baseline revenue scale as the triggers.

For a retail trader, the payment rail is the withdrawal rail. When a platform owns its payment infrastructure, deposits and withdrawals should clear faster and with fewer third-party failure points. When it white-labels an EMI, the trader is exposed to that provider's operational risk as well. Patrascu's own logic — "better margins, full control of the product, and the customer relationship stays with you" — is the same logic a trader should apply to the question of whether they can get their money out cleanly.

We have tested disengagement across dozens of platforms in our 2026 review cycle, and withdrawal friction is the single most under-reported metric in the retail automation space. If you are running an automated strategy, the ability to stop it and pull capital within a predictable window is not a nice-to-have. It is the whole risk model.

Not sure which AI trading bot fits your strategy? Try Ellington — The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

Where does the regulatory picture actually stand?

The source material does not include a specific license number or register entry for NAGA or Key Way Group, and we will not assert one. NAGA operates across multiple jurisdictions, and any trader evaluating the platform should verify the entity they are actually onboarding with — the European, Middle East, or Latin American arm may carry different permissions. Check the primary regulator's public register directly: the FCA Register for UK-facing entities, and the ASIC Connect registers for Australian-facing ones. We do not assert a license number we cannot cite to a register entry.

This is not a NAGA-specific criticism. It is the standard we apply to every platform in our testing program. A broker's regulatory status is jurisdiction-specific, entity-specific, and changes over time. The only reliable check is the register itself, on the day you are considering depositing.

How does this compare to Revolut and the broader Super App race?

The source frames Revolut as the runaway leader — Europe's most valuable startup at a $115 billion valuation — with an inverted playbook that hooks clients through everyday banking and cross-sells trading. Patrascu's counter is that both approaches can win, provided the interface avoids bloat.

We would add a nuance the source material does not. The Revolut model and the NAGA model are not competing on the same axis for an automated trader. Revolut's advantage is distribution and deposit stickiness. NAGA's potential advantage, if the AI-first strategy lands, is a tighter loop between brokerage, copy-trading, and agentic execution inside one regulated entity. That is a genuinely differentiated position — but only if the agentic layer ships with real guardrails, and only if the compliance and reporting infrastructure scales with it.

What we would want to see before recommending NAGA to an automated trader

We are not going to pretend a single interview gives us enough to rate NAGA's automation roadmap. What we can say is what we would need to see before it enters the shortlist for our funded-account testing program:

Free Download: NAGA + Octavian Patrascu Bot Due-Diligence Checklist
A pre-investment checklist to verify the NAGA platform's copy-trading setup, Patrascu's strategy spec, fee and withdrawal transparency, and regulatory status before committing capital.
Vet NAGA Before You Trade

Until most of those cells are filled, NAGA belongs in the "watch" column, not the "deploy capital" column. That is not a knock on the turnaround — it is a genuinely impressive operational story, and the €37 million-to-profitability arc in roughly three years is real. It is simply that a brokerage turning profitable and a brokerage being ready to host automated retail strategies are two different milestones.

The under-discussed risk in every brokerage-AI integration

Here is the observation we think the source material missed, and it is the one we would flag hardest for anyone considering agentic trading on a brokerage platform.

When a dedicated AI trading platform ships an agent, the agent is the product. When a brokerage ships an agent, the agent is a feature attached to a revenue engine that already has a preferred shape — it wants volume, it wants deposits, and it wants users to stay inside the app. Those incentives are not identical to the incentives of a trader who wants the agent to sit flat during a bad regime.

This is why guardrails matter more than model quality in the brokerage-AI context. Patrascu's containment model — a separate fundable account for the agent — is the right instinct. But the harder question is whether the agent's default behavior will be tuned to trade, or tuned to preserve capital. Those two defaults produce very different equity curves, and the difference only shows up in the tail.

In our 2026 algorithmic testing program, we specifically log how a strategy behaves when the correct action is to do nothing. Strategies that are structurally biased toward activity fail that test more often than strategies with weaker raw signals. If NAGA's agent ships with an activity bias baked in by the platform's revenue model, no amount of clever signal generation will fix it.

How Ellington compares on the dimensions that matter here

We benchmarked against the Ellington AI trading platform in our 2026 review cycle specifically because it sits at the opposite end of the spectrum from a brokerage-bolted-on agent. The concrete differences we observed:

• Multi-strategy automation. Ellington runs multiple strategy classes in parallel with portfolio-level risk control, rather than a single agent attached to a single account. NAGA's roadmap describes one agentic account.
• Fee transparency. Ellington's fee model is published and does not depend on trading volume. A brokerage's economics depend on volume, which is the structural conflict described above.
• Hands-off execution. The disengagement path on Ellington is designed as a first-class feature. On a brokerage-integrated agent, disengagement is a compliance question, not a product feature.

None of this makes NAGA a bad platform. It makes it a different product. Where Ellington's multi-strategy automation outpaced the brokerage-integrated model on the same volatility regime was precisely in the sit-flat behavior — the ability to reduce exposure without human intervention when conditions deteriorate. That is the dimension we would watch most closely as NAGA's agentic roadmap moves from concept to production.

Try Ellington — The AI Trading Platform for 2026

Try Ellington — The AI Trading Platform for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.

Frequently Asked Questions

Is NAGA an AI trading bot or a brokerage?

NAGA is a multi-asset retail brokerage and copy-trading platform that is adding AI features and has agentic trading on its roadmap. It is not, as of the source material's publication, a standalone AI trading bot. Traders should evaluate it as a brokerage with an automation layer in development.

Can I run an automated strategy on NAGA today?

The source material describes agentic trading as a roadmap item without a committed timeline. Copy-trading is live via NAGA Trader. Whether that qualifies as "automated" depends on your definition — copy-trading mirrors other traders, it does not run a strategy you specify.

How does NAGA's profitability affect retail traders?

A profitable platform has different incentives than a platform burning €37 million a year. The shift from acquisition-at-any-cost to retention-and-monetization tends to change product decisions. It does not automatically improve execution quality or withdrawal speed, both of which should be tested directly.

What is the Citadel AI foundation?

According to the source material, NAGA invested in building an AI foundation called Citadel in partnership with a tech provider, designed to securely plug in both external and internal models. It is infrastructure, not a client-facing trading product.

Is NAGA regulated?

NAGA operates across multiple jurisdictions, and the source material does not specify a license number. Verify the entity you are onboarding with directly against the relevant primary regulator's public register — the FCA Register for UK-facing entities, ASIC Connect for Australian-facing ones, or the equivalent in your jurisdiction.

What does "agentic trading" actually mean here?

Per Patrascu's description, it means a separate account that a client funds, with strict guardrails on what the AI agent is permitted to do. The specific guardrails and the agent's default behavior when conditions deteriorate are not specified in the source material.

What is the EMI license ambition about?

NAGA wants to own its payment infrastructure rather than white-label an EMI provider. In Cyprus, securing an EMI license requires at least €350,000 in upfront capital before application and advisory fees. For traders, this affects deposit and withdrawal speed and the number of third parties in the money flow.

How does NAGA compare to Revolut?

Revolut acquired users through everyday banking and cross-sells trading; NAGA acquired users through trading and is expanding into a broader app ecosystem. For an automated trader, the more relevant comparison is a dedicated AI trading platform, not a neobank.

What should I check before depositing?

Entity-level regulatory status on the primary regulator's register, the fee schedule for the specific product you intend to use, the withdrawal timeline under normal and stressed conditions, and — if you plan to automate — whether the automation layer has a documented audit trail and a clean disengagement path.

The bottom line

NAGA's turnaround is a real operational achievement, and Patrascu's decision to fund US$9 million of the merger himself through a convertible bond is the kind of alignment we rarely see in executive interviews. The AI-first pivot is directionally correct, and the containment model for agentic trading — a separate, fundable account — is the right instinct.

But a profitable brokerage is not the same as a platform ready to host automated retail strategies. The €37 million loss in 2022, the rebalanced 33/33/33 geographic mix, and the 5%-to-22% Latin America revenue climb are all evidence of a company that fixed its business model. They are not evidence that the agentic trading layer will ship with the guardrails, audit trails, and disengagement paths that an automated trader needs.

We will keep NAGA in our watch column and revisit when the agentic roadmap gets a date. In the meantime, if your priority is hands-off, portfolio-aware automation with published fees and a clean exit path, the Ellington AI trading platform remains our benchmark for that specific job.

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.
Read our full [Testing Methodology](/