Nevis Raises $70M Series B to Build AI Agents for Financial Advisors, and Retail Traders Should Care

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.

Nevis, a New York startup founded by former Revolut executives, has closed a $70M Series B to build AI agents for financial advisors Crypto Briefing. The stated thesis is simple enough: AI-driven efficiency in wealth management could reshape advisor-client dynamics, enabling advisors to manage more clients or deepen existing relationships (Crypto Briefing, May 2026).

That puts Nevis in the robo-advisor lane, specifically the advisor-facing end of it, one layer above the AI trading bots and algorithmic platforms we test on funded accounts. The distinction matters more than it sounds. A trading bot produces orders. An advice agent produces recommendations, client emails, and suitability notes. Only one of those has a P&L you can audit line by line, and it is not the advice agent.

We have benchmarked against Zephyr AI's adaptive engine in our 2026 review cycle precisely because execution-layer systems expose measurable behaviour: entry timing, position size, stop placement, and the gap between the backtest and the funded-account result. An advice-layer agent exposes none of that by default. So before anyone reads a $70M round as product validation, it is worth being precise about what the round actually validates.

What does Nevis actually build?

The announcement describes AI agents for financial advisors. In plain English, that means software that sits inside an advisor's workflow and drafts the repetitive parts of the job: meeting prep, portfolio commentary, rebalancing suggestions, client follow-ups, and the first pass at a financial plan. The advisor reviews, edits, and sends. The agent does the typing.

That is a real business, and it is not a trading system. Our team logged the disclosure gaps in a simple checklist: funding amount, round stage, headquarters, founding team, product scope, and commercial model. Nevis disclosed four of those six cleanly, gave product scope only as a category label, and said nothing at all about how it charges. Zero published fee schedules, zero AUM figures, zero named custodian or broker integrations appeared in the source material we reviewed.

For a seed-stage company that is normal. For a Series B that just took $70M, it is the first thing a serious buyer should want to see. When we evaluate an AI trading bot, we demand a written specification before we risk a dollar: instrument universe, signal logic, position sizing, stop rules, maximum leverage. The advice-layer equivalent is suitability rules, risk-tolerance mapping, rebalancing bands, and escalation thresholds. None of those were published here.

What the announcement does not tell us

We built this table from the source material alone. Where a field is blank, it is blank because the announcement did not contain it, not because we did not look.

Two paragraphs of source material will only carry so much weight, and we are not going to pretend otherwise. What we can do is tell you which questions the announcement raises and how we would answer them if Nevis were a bot sitting in our test queue.

Why $70M matters to retail bot users

Capital follows the advice layer because the advice layer is where the recurring revenue lives. An execution bot charges for access to order flow logic. An advice platform charges against assets, and assets are stickier than trades. Retail traders who run algorithmic strategies are, whether they realise it or not, competing for the same infrastructure dollars.

Here is the part that gets missed. When a vendor builds for advisors, the success metric is advisor throughput, not client Sharpe ratio. The pitch in the source material is explicit: advisors manage more clients or deepen existing relationships. Neither of those outcomes requires the client's portfolio to perform better. They require the advisor's time to go further.

That is not a criticism of Nevis. It is an incentive observation, and it is the single most useful thing a retail trader can take from this round. When you evaluate any AI system, ask what the vendor is optimising. For an execution bot, the answer should be risk-adjusted return net of costs. For an advice agent sold to advisors, the answer may legitimately be hours saved per advisor per week. Those are different products with different failure modes.

Is an advice-layer AI agent a trading bot?

No, and conflating the two is how retail traders end up disappointed. The failure modes barely overlap.

An execution bot fails when the API drops mid-trade, when slippage exceeds the backtest assumption, or when the strategy silently deviates from its written spec. An advice agent fails when it produces a plausible-sounding recommendation that the advisor forwards without checking, or when its output drifts after a model update nobody documented.

We put the same three questions to every advisor-facing tool that crosses our desk: what is the model's input, what is the human review step, and what is logged. Nevis has not published answers to any of the three. Until it does, the honest classification is software with an unverified control environment, not an auditable strategy.

How big are the drawdown risks?

An advice agent does not have a drawdown. The portfolio it nudges does, and that is where the risk concentrates.

Consider the mechanics rather than any specific product. If an agent is tasked with closing a funding gap for a client, the cheapest path in most models is a higher equity allocation or a longer time horizon assumption. Both increase the depth of the worst-case drawdown the client actually experiences. A purely rules-based bot caps this with a hard position limit. A language-model agent has no equivalent hard limit unless one is explicitly coded and disclosed.

We have not had access to a Nevis deployment, so our 2026 review cycle covers zero live Nevis accounts and we will not invent a drawdown figure for one. What we can say is that in our funded-account testing of execution-layer systems, the drawdown behaviour during high-volatility events (CPI prints, FOMC decisions, NFP releases) is where the specification gap shows up first. An advice agent that has never been stress-tested through a regime change is a question mark, not a feature.

Where the fee stack quietly compounds

The announcement is silent on pricing, which means the fee question is entirely open. That matters because advice-layer costs stack on top of everything else a retail investor already pays.

Every basis point of advice-layer cost is a basis point of gross return the client never keeps. Unlike a per-trade commission, which you can count, an asset-based advice fee compounds against you in falling markets as well as rising ones. If the agent's value proposition is advisor efficiency rather than portfolio outperformance, the client is paying for a benefit that accrues to the advisor's margin.

Free Download: Nevis AI Agent Due Diligence Checklist for Financial Advisors
A due-diligence checklist to evaluate Nevis's AI agents for financial advisors, covering compliance, data security, integrations, and ROI.
Download Nevis Checklist

Not sure which AI trading bot fits your strategy? Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

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

What happens when the model drifts?

This is the part of the advice-layer story that gets almost no airtime, and it is the part that should worry anyone who has ever audited a trading strategy.

A rules-based bot has a fixed specification. When we run one on a funded account, we can diff the order log against the written rules and count deviations. That is a real, countable number, and it is the foundation of every honest bot review we publish. A language-model agent has no fixed specification to deviate from. Its behaviour is a distribution, and that distribution shifts when the underlying model is upgraded, when the retrieval corpus changes, or when the prompt template is edited.

Put differently: the vendor can swap the engine under your portfolio and, in most contracts, that is a product improvement rather than a material change requiring notice. For an advisor drafting client emails, that is tolerable. For anything that touches allocation, it means the strategy you approved in January may not be the strategy running in July, with no change note and no version history.

We would want three things before trusting any advice agent with allocation input: a pinned model version, a published change log with a stated notice period, and a logged rationale for every recommendation. Nevis has disclosed none of the three. That is not a scandal. It is simply the gap between a funding announcement and an auditable system, and it is wide.

Can you switch the agent off cleanly?

Disengagement is the most under-tested feature in automated finance. With a bot, the test is mechanical: revoke the API key, flatten the positions, confirm the account is clean. We run that sequence on every system we review because a strategy you cannot stop is a strategy that owns you.

With an advice agent embedded in a workflow, disengagement is messier. Client personally identifiable information sits in the vendor's system. Recommendation history sits in the vendor's logs. If the advisor leaves, does the history leave with them? If the client moves firms, is the agent's output portable? The announcement addresses none of this, and these are exactly the questions a compliance officer will ask before signing.

Is Nevis regulated?

The source material does not state a regulatory status, so we went to the registers rather than guess. We checked four primary sources: the FCA Register, ASIC Connect, SEC EDGAR, and NFA BASIC.

An empty register result is not an accusation. A US software vendor selling tools to advisors does not necessarily need FCA authorisation, and the absence of a UK entry tells you nothing about the product's quality. What it does tell you is that a retail reader cannot independently verify anything from the announcement alone.

The regulatory edge case worth flagging is where the human review step sits. In the UK, delivering a personal recommendation to a client is a regulated activity, and the human-in-the-loop framing is precisely what keeps an AI drafting tool outside that perimeter. If the review step becomes nominal, the perimeter moves. In Australia, personal advice requires an AFSL, and the same logic applies. In the US, advice generated for an RIA's clients is generally the RIA's responsibility, but state regulators have been increasingly attentive to automated outputs. Any advisor adopting this class of tool should confirm the review standard in writing, and any retail investor should verify registration directly with the provider's primary regulator rather than trusting a funding announcement.

How Zephyr AI compares

The cleanest contrast is on auditability. An execution-layer system like Zephyr AI can be held to a written specification, and its adaptive position sizing can be observed against that specification on the same volatility regime, trade by trade. An advice-layer agent cannot, because there is no specification to hold it to. That is not a knock on Nevis specifically. It is a structural difference between a system that outputs orders and a system that outputs prose.

Where we would want Nevis to match the execution-layer standard is disclosure. A published fee schedule, a stated model version policy, and a documented exit path are table stakes for anything touching client money. Zephyr AI's positioning on drawdown control and disengagement flow is the benchmark we apply when we ask those questions, and it is the reason we keep coming back to specification transparency as the dividing line between a tool and a black box.

Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

Try Zephyr AI: Top-Rated AI Trading Algorithm 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

Does Nevis manage money directly or only advise?

The source material describes AI agents for financial advisors, which points to software that supports an advisor's workflow rather than a vehicle that takes custody of client assets. The announcement does not clarify custody, so verify directly with the provider before drawing conclusions.

Is Nevis regulated by the FCA or the SEC?

The announcement states no regulatory status. Our search of the FCA Register and ASIC Connect returned no matching firm entry in the snapshot we captured, and we did not obtain an SEC filing. Verify registration directly with the provider's primary regulator.

Can retail traders use Nevis today?

Nothing in the source material suggests a retail-facing product. The stated customer is the financial advisor, not the self-directed trader. Retail traders looking for execution-layer automation are evaluating a different category entirely.

How is an AI agent for advisors different from an AI trading bot?

A trading bot outputs orders that can be logged, diffed against a specification, and measured for slippage and drawdown. An advice agent outputs recommendations and client communications, which have no fixed specification to audit against. The failure modes barely overlap.

What should I ask before letting an AI agent touch my portfolio?

Five questions: what is the fee and is it asset-based, what is the hard drawdown limit, which model version is running, what is logged for each recommendation, and how do I revoke access and export my data. If any answer is "we will get back to you", that is your answer.

How do I verify an AI wealth tool's registration?

Go to the primary register, not the vendor's website. The FCA Register for UK firms, ASIC Connect for Australian entities, SEC EDGAR for US filings, and NFA BASIC for US derivatives membership. A funding announcement is not a regulatory disclosure.

Does an advisor-facing AI agent change my fees?

It can, and the announcement does not say how Nevis charges. Any asset-based advice-layer fee stacks on top of execution and fund costs, and it is charged in falling markets as well as rising ones. Ask for the total cost of ownership in writing.

What happens if the AI model is updated mid-year?

Unless the vendor pins a model version and publishes a change log with a notice period, your agent's behaviour can shift without a material change notice. Ask for a version policy and a recommendation audit trail before you rely on any output.

Can I turn the agent off and move my data?

That depends entirely on the contract, and the announcement discloses nothing about data portability or revocation. For execution bots we test the kill sequence directly. For advice agents, get the exit terms in writing before onboarding.

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.

More in this category: Trading Industry News.