By CRYPTOVERSE Legal Consultancy
A client opens an investment app, answers twelve questions and receives a recommended portfolio. A few taps later, the app sends instructions to a broker. Each month, software checks whether the portfolio needs rebalancing.
The founders may describe the business as a technology platform. The Financial Services Regulatory Authority (FSRA) will ask a more important question:
What investment service is the platform providing to the client?
In Abu Dhabi Global Market (ADGM), delivering a service through software does not change its regulatory character. A robo-adviser may need an FSRA Financial Services Permission (FSP) for investment advice, discretionary asset management, arranging investments or a combination of activities.
The boundary is especially important for founders building an AI investment app, a digital wealth management platform or a white-label portfolio engine. Each may use similar technology while performing a different legal role.
What is “digital investment management” in ADGM?
The FSRA’s Supplementary Guidance – Digital Investment Management, updated on 6 November 2025, describes digital investment management as investment management services delivered using algorithm-based tools with limited or optional human interaction between the client and provider. It covers both fully digital services and hybrid models in which clients can speak with a human adviser.
The guidance expressly includes modern AI techniques, such as machine learning and large language models, within its discussion of algorithm-based tools. It excludes activities that use algorithms solely for back-office support.
The distinction is functional. A spreadsheet that helps an authorised adviser prepare reports is different from an app that gives clients tailored investment recommendations or exercises discretion over their portfolios.
Does every investment technology company need an FSRA licence?
No. The FSRA guidance distinguishes a business providing investment services from a company that only supplies technology to a wealth manager. A firm that solely white-labels or sells the technology, without itself providing the regulated investment service, is not treated as a Digital Investment Manager requiring an FSP on that basis.
But the label “technology provider” is not enough. A regulatory assessment should establish:
- Who contracts with the end client;
- Who determines the portfolio recommendation or investment decision;
- Who controls the model and its outputs;
- Who communicates advice to the client;
- Who sends transaction instructions; and
- Who is responsible when the model produces an unsuitable result.
If an app provider performs regulated activities itself, supplying software to another firm will not necessarily describe its entire business.
The three main permission questions
The FSRA guidance identifies three regulated activities commonly found in robo-advice and digital investment management models.
| What the platform does | Likely permission question |
| Recommends a portfolio or particular investment transactions to a client. | Advising on Investments or Credit |
| Decides, with discretion, when to buy, sell or rebalance investments for the client. | Managing Assets |
| Facilitates investment transactions, for example by passing a client-authorised instruction to a broker. | Arranging Deals in Investments |
The FSP must match the actual workflow, including what happens after the client sees a recommendation. Two apps can display the same portfolio screen but need different permissions because one waits for a client’s decision while the other automatically trades on the client’s behalf.
1. The app recommends a portfolio
Suppose a user enters their income, investment horizon and risk tolerance. The app responds:
“Based on your answers, we recommend that you invest 60% in Portfolio A and 40% in Portfolio B.”
The platform is doing more than displaying market data. It is making a recommendation based on the user’s circumstances. The FSRA identifies this type of conduct as potentially falling within Advising on Investments or Credit.
2. The app rebalances without asking each time
Now suppose the client authorises the platform to keep the portfolio within agreed parameters. The system decides when to trade and sends instructions to a broker without seeking separate approval for every rebalance.
That raises the Managing Assets permission question: the platform is exercising discretion over the client’s investments. The FSRA gives discretionary portfolio rebalancing as an example.
The fact that an algorithm makes the immediate decision does not remove the manager’s responsibility for that decision.
3. The app helps execute the client’s choice
A platform may recommend an investment and, once the client agrees, pass instructions to a broker to buy it. The FSRA identifies that workflow as an example relevant to Arranging Deals in Investments.
A founder should map exactly who initiates, transmits and completes each transaction rather than assume that broker execution resolves the platform’s permission requirements.
Does a digital asset manager need all three permissions?
Not always. The FSRA guidance states that a Digital Investment Manager authorised to Manage Assets does not need separate permissions for advice or arranging where those activities are undertaken incidentally as part of its investment management activities.
The qualification matters. If the firm also operates standalone advisory accounts or arranges transactions separate from its discretionary management service, additional permissions may be required.
A business model should therefore distinguish the client journeys. One group of users may receive fully managed portfolios; another may only receive recommendations and make their own decisions. The permission analysis can differ for each journey.
What if a third-party custodian holds the assets?
Using a regulated custodian does not automatically settle the platform’s licensing position for Virtual asset funds advisory services.
The FSRA guidance describes models where a Digital Investment Manager holds Client Assets through an account with a third-party banker or custodian, and models where the client establishes a direct relationship with a regulated custodian. In the latter case, the platform may need permission for Arranging Custody, unless a relevant exclusion applies.
The chosen custody model also affects operational responsibilities, client disclosures and the prudential analysis. It should be designed with the permission application, rather than added as a late integration decision.
“The algorithm decided” is not a compliance answer
A robo-adviser needs people who can explain and control its model.
The FSRA guidance expects board and senior management oversight of the algorithm’s design, performance, deployment and security. It discusses competent personnel, documented model logic, testing, source-code safeguards, error detection, contingency plans and ongoing reviews for inaccurate or biased outputs. For a sufficiently complex model, the FSRA may require third-party validation of its performance.
Consider a risk-scoring model that interprets a client’s answer of “I can tolerate losses” as approval for a highly volatile portfolio. If the questionnaire is poorly designed, the resulting recommendation may be inconsistent with the client’s actual circumstances.
A defensible model governance process should be able to answer:
- What information did the client provide?
- How did the model use it?
- Why did the model produce this recommendation or trade?
- How was the model tested before and after launch?
- Who can intervene if the output is wrong?
The newer guidance specifically identifies risks such as model drift, model poisoning and adversarial manipulation. The controls need to evolve with the technology the firm actually uses.
How does an online app assess suitability?
A short digital questionnaire can be convenient for users. It must still gather enough reliable information to support suitable advice or discretionary decisions.
The FSRA guidance links Digital Investment Managers to the suitability rules in the Conduct of Business Rulebook (COBS). It expects an assessment of the client’s needs, objectives and financial situation and, where relevant, risk tolerance, knowledge, experience and understanding of the risks.
The guidance highlights several practical safeguards for online questionnaires:
- Ask questions proportionate to the products’ complexity and risk;
- Identify clients for whom the service may be unsuitable;
- Follow up on inconsistent answers;
- Provide an appropriate route to human assistance where needed; and
- Prompt clients to update information as their circumstances change.
For example, a client who says they need their capital back in six months should not simply be routed into an unsuitable long-term investment because they clicked “high risk tolerance” on a later screen.
What must the platform explain to clients?
A digital investment service should clearly disclose what it does, how the algorithm is used, its relevant assumptions and limitations, and circumstances in which its outputs may fail or be changed. The FSRA guidance discusses these disclosures alongside the wider COBS requirements, with particular attention to Retail Clients.
A product screen that says “AI-powered investing” tells the client very little. More useful disclosures explain, in plain language, whether the firm is advising or managing, which assets it may select, what human oversight exists, what fees apply and who holds client assets.
The manner and timing of disclosure matter. A warning presented only after a client has committed funds is unlikely to serve the same purpose as information available when the client is deciding whether to use the service.
What about cyber risk, outsourcing and AML?
Digital investment managers depend heavily on technology, client data and third-party systems. The FSRA guidance therefore addresses data security, technology governance, business continuity, outsourcing and anti-money laundering and targeted financial sanctions controls.
A platform should be able to identify what happens if its model, cloud provider, broker connection or client onboarding service fails. It also needs to determine who monitors transactions and financial crime alerts in an automated customer journey.
These questions belong in the FSRA application and operating model, not solely in a vendor contract.
Is there a special low-cost “robo-adviser licence”?
A prospective applicant should be cautious with that description. The FSRA guidance maps the business to its actual regulated activities, each with associated prudential and operational requirements.
The guidance also states that the FSRA may consider modifying or waiving prudential or other requirements case by case where technology improves supervision, risk management or regulatory outcomes. This is a regulatory discretion, not an automatic capital discount for using an app or AI.
Capital planning should therefore start with the proposed activities, custody model and current prudential rules, then account for any specific relief only if the FSRA grants it.
A practical licensing test for founders
Before launching a robo-adviser or AI investment platform in ADGM, map the customer journey from beginning to end:
- Input: What financial and personal information does the client provide?
- Output: Does the platform present information, or recommend an investment for that client?
- Decision: Does the client approve each transaction, or can the platform decide and trade?
- Execution: Who sends instructions to the broker?
- Custody: Who holds or arranges custody of the assets?
- Control: Who owns, tests, monitors and can stop the algorithm?
- Responsibility: Which entity contracts with the client and stands behind the service?
The answers determine the likely FSRA permissions and the controls needed to support them.
The central lesson
An ADGM robo-adviser is regulated according to the investment services it provides, whether those services are delivered by a human, conventional software or AI. A technology company selling tools to an authorised firm has a different position from a platform that recommends investments, arranges trades or manages portfolios for clients.
For founders, the strongest starting point is an honest map of what the product does. That map should drive the FSRA licensing strategy, client disclosures, suitability process, algorithm governance and custody design before the platform goes live.
CRYPTOVERSE Legal Consultancy advises on ADGM fund setup and licensing services, regulatory perimeter assessments, digital investment management models and the governance documentation needed for technology-driven financial services.
Disclaimer: This article provides general information based on the FSRA’s Supplementary Guidance – Digital Investment Management, version 2 dated 6 November 2025, and the applicable ADGM regulatory framework. It is not legal advice on a particular product, algorithm, client journey or licence application.
FAQs
1. Does building a robo-adviser automatically require an FSRA licence in ADGM?
Not automatically — it depends on what the platform actually does. A firm that only supplies technology to an authorised wealth manager, without providing the investment service itself, isn’t treated as a Digital Investment Manager on that basis alone.
2. What’s the difference between a robo-adviser that recommends and one that manages?
A platform that recommends a portfolio and lets the client decide raises the Advising on Investments or Credit question. A platform that can rebalance or trade without seeking fresh approval each time raises the Managing Assets question — the key factor is who holds decision-making discretion.
3. Does using AI or machine learning change the licensing analysis?
No. The FSRA’s guidance explicitly covers modern AI techniques like machine learning and large language models within its discussion of algorithm-based tools — the underlying regulated activity is what matters, not the technology used to deliver it.
4. What governance does the FSRA expect over an investment algorithm?
Board and senior management oversight of the model’s design, performance, and deployment — including documented logic, testing, error detection, contingency plans, and ongoing review for issues like model drift or biased outputs. For complex models, third-party validation may be required.
5. Is there a special reduced licensing category for robo-advisers?
No dedicated “robo-adviser licence” exists. The FSRA maps the business to its actual regulated activities and associated capital and operational requirements, though it may consider modifying certain requirements case by case where technology improves risk management.