Monstro USA blog

Why can’t a large language model make regulated financial decisions?

augustin.wolff@monstro.com (Monstro) — Mon, 28 Sep 2026 14:49:51 GMT

A large language model can't make a regulated financial decision because probabilistic intelligence is not institutional authority. The model produces the most likely output given what it has seen. A regulated decision has to be one the institution has authorized, can explain and can reproduce when asked. A better model improves the output, but it doesn't turn that output into a decision the institution can stand behind.

Probability is fine inside the model, as long as the institution keeps firm control over what actually gets decided.

What is the difference between an answer and a decision?

An answer responds to a question. A decision is a position the institution takes and has to be able to defend later.

If a client asks whether they can afford a home, an answer is a plausible sentence. A decision is a determination made against the bank's liquidity thresholds, suitability requirements, product criteria and approval structure, and the bank has to stand behind it when a regulator, a supervisor or the client asks why.

When a system produces an answer and presents it as a decision, nothing has really been automated. The decision is simply missing the accountability it needs.

Isn't this solved by a better model?

No. The limit is authority, and a more accurate model doesn't change that.

A more capable model gives better answers more often. It still doesn't hold the institution's rules, and it can't be the party that decides what the institution permits. It makes a better estimate, while a regulated decision needs an outcome the institution controls.

Something real has changed, though. AI can now handle far more complexity without becoming the institution's authority. That is what makes this approach workable today, and it also sets the limit on what the model should be asked to do.

What about putting the model behind the institution's rules?

That instinct is right, and the details decide whether it works.

If the rules are handed to the model as context and the model is asked to apply them, the model is still the one making the decision, just with more information in front of it. The outcome stays probabilistic. Ask the same question twice and the bank can't promise the same result.

The alternative is to use generative AI to help build the institution's assets ahead of time. Rules, calculations, decision paths and approved guidance are drafted with AI assistance, then reviewed, approved and executed deterministically. AI does a lot of the work of creating the intelligence, and the policy is settled before a client ever asks a question.

What should a system do when it lacks the facts to answer?

It should stop and say what is missing.

Consider a homebuyer whose income, cash, investments, liabilities and liquidity needs are all verified, while the planned renovation spend is unknown. The right move is to ask for that number instead of estimating it. We call this step “not yet, ask this first”.

Treat a missing fact as a reason to pause. A system that always produces an answer can't flag the one case where it shouldn't, which is why every fact it holds should carry its source, its date, a confidence level and a note of what is still unknown.

Is an LLM wrapper enough to deliver financial advice?

No. An LLM wrapper in front of existing systems is a demo, and a regulated institution can't deploy a demo as its architecture.

The difference shows up in what the institution owns once the model has finished. With a wrapper, the institution is left with a set of responses. A proper architecture gives it governed institutional assets instead: rules, calculations, decision trees, required information, approved guidance and action pathways that it owns and can inspect, change and reuse.

What does a deployable architecture look like?

It has four layers, and governance applies inside each of them while the system runs:

Layer	What it does
Verified facts	Records what is known about the client, with source, date, confidence and known gaps
Institutional assets	Holds the institution's rules, criteria, calculations, thresholds and approved guidance in executable form
Deterministic execution	Asks, decides, guides, escalates and acts, so the same facts and rules lead to the same governed decision
Individualized explanation	Writes the explanation for each client, while the institutional logic underneath stays the same

The simplest test of the whole design is one question: why did you tell her that? A deployable system can rebuild the answer from the facts, the rule, the calculation, the path, the model's contribution, the approval and the client output. There is more on how this works in practice on our governance page.

Frequently asked questions

Does this mean AI has no role in regulated financial decisions?

AI does a lot of the work. It interprets unstructured institutional knowledge, understands a client's context and drafts proposals. The constraint only applies to the final decision.

What does deterministic execution actually mean?

The decision and the guidance that follows are controlled and reproducible. The wording can still differ from client to client, because each explanation is written for the individual.

Who sets the thresholds?

The institution. The model works against thresholds the institution has defined and approved, and it never sets them itself.

Can a recommendation be explained after the fact?

Yes, and the architecture is designed around that. Every step can be reconstructed, and a person keeps the authority to review, escalate or override.

What does AI make it possible for your bank to become?

augustin.wolff@monstro.com (Monstro) — Mon, 28 Sep 2026 14:37:38 GMT

AI makes it possible for a bank to become the place where its clients make financial decisions, as well as the place where they keep their money. That goal is different from making today's operations faster or cheaper, and it calls for different investments. Most bank AI programs today improve the institution as it already is. The bigger opportunity is to change what the institution means to the people it serves.

Why is that a different question from “what can AI do for our bank?”

Both questions are reasonable, and they lead to different roadmaps.

Asking what AI can do for the bank you already have produces a list of improvements to things that already run: service costs, conversion rates, fraud detection, contact center handling times and the productivity of existing teams. Every item on that list is worth doing, and together they make the institution noticeably better.

Asking what AI makes it possible for the bank to become starts somewhere else. It looks at the parts of the bank's ambition that used to be out of reach and asks whether any of them can now be done. That list is shorter, and the items on it are bigger.

What does it mean to become the place where clients make financial decisions?

The financial decision relationship is the relationship an institution has when clients come to it to decide what to do with their money, in addition to holding, moving or borrowing it.

Think about a client buying a home. The client sees one decision, while the bank sees six product areas: home finance, cash and deposits, investments, insurance, tax and retirement. What the client wants to know is what they should do.

The order matters. The sequence runs from guidance to the decision, then to the transaction, and finally to the economics. That is why the relationship sits upstream of revenue: the institution that helps a client decide is usually the one whose products get chosen.

It works the other way too. A bank that loses the decision will, over time, tend to lose the transaction as well. Its products can look the same for years while the relationship that feeds them moves somewhere else.

Hasn't this been tried before?

Yes, many times. The ambition itself isn't new.

Customer relationship management, personal financial management, next-best action, financial wellness, digital advice and personalization were all serious attempts at the same goal. Each one ran into the same five problems.

Problem	Why it stopped earlier attempts
Fragmented data	What the bank knew about a client sat in systems that didn't reconcile
Unreliable intelligence	Outputs couldn't be trusted enough to put in front of a client
Individual complexity	Real financial lives didn't fit the segment logic these systems were built on
Institutional rules	The bank's own requirements couldn't be expressed in the system
Scale	What worked for a thousand relationships broke down at ten million

None of this came down to a lack of intent or effort. These were hard constraints, and working harder didn't remove them.

What has changed?

The ambition may now be achievable. Four capabilities that used to be out of reach are available together: understanding a client's context across their financial life, working with unstructured institutional knowledge, expressing policy in a form a system can run, and doing all of it at population scale.

AI can now handle far more complexity without becoming the institution's authority. Keeping intelligence and authority separate is what makes a system deployable in a regulated institution.

What has to be true for a bank to actually do this?

The important development is that an institution can now turn its own expertise into governed, executable assets and run them at a scale banking hasn't had before. How well AI holds a conversation matters much less than that.

It takes four things working together, with governance applied while the system runs:

Verified facts. The system knows what it knows about a client, including the source, date, confidence and gaps for each fact. When something is missing, it asks before going further.
Institutional assets. The bank's liquidity rules, financing criteria, suitability requirements, calculations, approval thresholds and approved guidance, in a form the system can execute. Where they apply, this includes Sharia requirements, and the institution decides what is permissible.
Deterministic execution. The system asks, decides, guides, escalates and acts. The same facts and rules lead to the same governed decision, and the thresholds come from the institution.
Individualized explanation. Each client gets an explanation written for them, based on institutional logic that stays the same from one client to the next.

You can read more about how this works on our governance and Financial Guidance pages.

What should a bank do first?

Pick a decision your clients already make where you aren't involved today, such as buying a home, handling a liquidity event or moving into retirement. Map what the bank already knows, what it already requires, and the point in that decision where the client goes somewhere else. The gap between those is where to start.

Frequently asked questions

Is this a chatbot?

No. A chatbot answers the question it's asked. A Financial Intelligence System works out what matters in a client's financial life, what should happen next and whether the institution's rules allow it, then explains the result to that client.

Does this replace bankers or advisors?

No. It expands professional capacity, so a bank can serve more relationships without growing headcount at the same rate, and its people get more time for judgment and client conversations.

Does a bank have to replace its core systems?

No. It adds Financial Intelligence without platform replacement, working as an intelligence layer across the systems the bank already runs.

How is this governed?

Governance is applied while assets are created and while they run. That includes rules and permissions defined by the institution, explainability, an audit trail, professional review and override authority.

What a governed AI system for banking actually requires

augustin.wolff@monstro.com (Monstro) — Mon, 28 Sep 2026 14:36:13 GMT

A governed AI system for banking needs four things in place at once: verified facts about the client, including what is still unknown; institutional assets that the bank owns and can change; deterministic execution; and an explanation written for each client. Governance also has to be part of how the system runs from the very first step. If one of the four is missing, what you have is a supervised system, which is a much weaker thing to put in front of a regulator.

Plenty of systems are described as governed today. Usually that means a person reviews what the model produced before a client sees it. Review is a useful control and banks should keep it. It still happens at the very end, and the reviewer rarely has any way to see how the output was put together.

Why isn't human review enough?

Review tells you whether an answer looks right, which is a different question from whether it was produced correctly.

A reviewer who approves a recommendation is judging a conclusion. They aren't checking that the liquidity threshold was applied, that suitability was assessed, that the figures were current, or that the same client with the same facts would get the same result tomorrow. At a hundred recommendations a day, the only way to scale review is to add reviewers, and governance has to scale without that.

Professional review, supervisory approval, escalation and override authority all still matter. They cover one part of governance, and the rest has to be built into the system itself.

What does “verified facts” mean in practice?

It means the system keeps track of what it knows about a client and what it still doesn't know.

Each fact comes with four things attached: where it came from, how recent it is, how confident the system is in it, and what is missing around it. People tend to overlook the last one, and it is often the one that matters most.

Take a client who is buying a home. The bank has verified their income, cash, investments, liabilities and stated liquidity needs, but it doesn't know how much they plan to spend on renovation. A system built to always produce an answer will estimate that figure. A governed system stops and asks for it first. We describe that step as “not yet, ask this first”.

A gap in the data is a good reason to pause. A system that always returns something gives you no signal in the one case where it shouldn't have answered at all.

What are institutional assets, and why do they matter?

An institutional asset is part of the bank's own expertise, written down in a form the system can execute. That covers liquidity rules, financing criteria, suitability requirements, product requirements, calculations, approval thresholds, approved guidance and the routes a client can take to act on it. Where they apply, Sharia requirements belong here too, as one more set of rules the institution defines. The model never decides what is permissible, because that decision stays with the institution.

This step is often skipped, and skipping it changes what the bank ends up owning.

Generative AI is very good at helping to build these assets. It can read unstructured policy documents, interpret them, suggest a structure and write a first draft. People review and approve the result before the system uses it. Generative AI shouldn't be making these rules up while a client waits for an answer.

A useful question for any vendor is what the bank owns once the model has done its work. Many setups leave the bank with a log of responses and little else. A governed approach leaves it with assets it can inspect, change and reuse, so its expertise is kept in a form the bank controls.

What does deterministic execution actually mean?

It means the same facts and the same rules lead to the same governed decision every time.

In practice the system works through five stages:

Ask for anything that is missing.
Decide against a threshold the institution has set.
Guide the client through the options, for example using less cash, borrowing more or selling some investments.
Escalate to a specialist when the institution requires it.
Route the client to the right place to act.

The threshold at the decision stage always comes from the institution, and the model has no say in it.

Deterministic execution doesn't mean every client reads the same words. An experienced investor and a first-time buyer should each get an explanation written for them. The institutional logic behind the decision is the same for both, and the model's job is to explain that logic clearly without changing it.

How do you prove any of this to a regulator?

You reconstruct a single recommendation from start to finish.

The question a regulator is likely to ask is a simple one: why did you tell her that? A governed system answers with a trace covering the facts it used, the rule it applied, the calculation, the path it followed, what the model contributed, who approved it and what the client saw. Provenance, approval, audit and human oversight cover the whole process, including everything that happens before a client sees anything.

If all a system can show is a log of what it said, you can see the output but not how the decision was reached, and governance can only look backwards.

Why is this hard to retrofit?

The four requirements are part of the architecture, so a review step added at the end can't supply them. Provenance has to be recorded from the moment a fact enters the system. Rules only become executable once someone has written them down in that form, and a decision can only be reproduced if its path is fixed in advance.

Putting an LLM wrapper in front of existing systems will get you a working demo. Demos are worth building, as long as nobody mistakes one for the architecture a regulated bank needs.

None of this asks a bank to invent a new control framework. The expertise already sits inside the institution, and the architecture makes it executable within the controls the bank already runs. There is more detail on our governance page, and on how this connects to Financial Guidance.

Frequently asked questions

Is “governed” the same as “compliant”?

No. Governed means the system operates within the authority and controls the institution has defined. That supports the bank's compliance responsibilities, and the responsibility itself stays with the bank. No system can make an institution compliant on its own.

Does this require replacing core banking systems?

No. It works as an intelligence layer across the systems the institution already runs.

Who owns the institutional assets?

The institution does. That is the main reason to build them as assets instead of prompts.

Where does human judgment fit?

At every stage. People define the rules, approve the assets, review escalations and keep the authority to override. The aim is to expand professional capacity while keeping professional judgment in charge.

What is a Financial Intelligence System?

augustin.wolff@monstro.com (Monstro) — Mon, 28 Sep 2026 14:34:12 GMT

Financial institutions hold the accounts, products, expertise and much of the information that shapes a client's financial life. But that does not necessarily mean they are present when the client decides what to do next. A Financial Intelligence System is designed to close that gap.

A Financial Intelligence System is the complete system that creates, governs and delivers Financial Intelligence across a financial institution. It brings together what the institution understands about each client, what the institution itself knows and allows, and Monstro's own financial reasoning. From that, it works out what matters in a client's financial life, why it matters and what should follow. The result is continuous, personalized, governed Financial Guidance, delivered to clients directly or through the people who serve them.

The question comes up in most of our first conversations with institutions, often from teams that have already assessed several AI vendors. This page sets out our full answer. Each section stands on its own, so you can go straight to the part most relevant to you.

What does a Financial Intelligence System do?

It helps a financial institution identify the moments that matter in a client's financial life and respond to them well, for every client and not only the few who have a dedicated advisor.

A client's financial life changes all the time. They start a new job, have a child, receive an inheritance, sell a business, move to another country or start caring for a parent. Each of those changes is a Guidance Opportunity, a moment where good guidance could make a real difference to the person. A Guidance Opportunity starts with what the client needs. A product conversation may follow, and often does, but it isn't the starting point.

For each one, the system determines what matters, why it matters and what should follow, within the institution's approved standards and controls. That becomes Financial Guidance: a clear explanation of where the client stands, the options open to them and a sensible next step. That step might be something the client can do in the app, or a conversation with a banker or advisor who arrives already knowing the context.

What goes into a Financial Intelligence System?

It combines three kinds of intelligence, each with a distinct role:

Client Intelligence is the system's understanding of an individual client's financial life. It evolves as circumstances change while preserving the context needed to understand what is known, where information came from and where important gaps remain. It works more like the understanding an experienced private banker builds over time than like a data record.

Institution Intelligence is everything the institution knows, permits and requires: its products and approved solutions, policies, permissions, compliance requirements, house views, financing criteria and approval thresholds. This is what makes the guidance belong to the institution. A client should receive guidance that reflects their bank's standards, whichever channel or employee they happen to use.

Monstro Intelligence is Monstro's proprietary financial reasoning. The system is designed to reason across financial domains that rarely sit together inside one institution, such as cash flow, investments, tax and retirement. Its role is to connect the client's situation with the institution's expertise.

When the three are applied together, the result is Financial Intelligence. The Financial Intelligence System is what makes that happen reliably and at scale, across the whole institution.

How does it make decisions an institution can stand behind?

This is usually the first thing a risk or technology team wants to understand.

Large language models are powerful at interpreting information and generating language. But general-purpose models generate outputs probabilistically at runtime, which makes them insufficient on their own for decisions an institution must consistently reproduce, govern and explain. A Financial Intelligence System therefore uses AI where it is strongest and keeps authority with the institution.

In practice that works in four steps:

The institution's expertise becomes executable. Generative AI helps turn policies, product rules, calculations and approved guidance into assets the system can run. The institution reviews and approves them before any client sees a result.
Facts are checked before anything is decided. If something important is missing, such as a client's tax position, the system asks for it instead of filling the gap with an estimate.
Decisions follow the institution's rules. For governed rules and decisions, this is what we mean by deterministic execution. The same facts and the same institution-defined rules lead to the same governed decision, and the thresholds come from the institution, never from the model.
The explanation is written for the person. When the same approved decision logic applies, the underlying governed outcome remains consistent while the explanation can adapt to the individual client.

Every step leaves a record, so anyone reviewing a piece of guidance later can see what it was based on, the approved logic that applied and what the client was shown. Bankers and advisors can review guidance, escalate it and override it.

A worked example

Take a client who has just sold the small business they ran for fifteen years. The proceeds arrive in their current account. At many institutions that deposit is simply a larger balance until someone happens to notice it.

A Financial Intelligence System treats it as a Guidance Opportunity. Client Intelligence knows the client's income has changed, what they already hold and what they owe. It doesn't know how much of the sale will be taxed or when the client plans to stop working, so it asks. Institution Intelligence adds the bank's liquidity rules, its suitability requirements and the point at which a wealth advisor has to be involved. Monstro Intelligence reasons across cash, tax, investments and retirement together, since for this client they are one decision.

The client then receives guidance in plain language: how much to keep easily available, what to consider before investing the rest, and an offer to discuss it with an advisor. The advisor who picks up the conversation sees the same reasoning the client was given, so the first meeting starts from where the client already is.

What changes for the institution?

A Financial Intelligence System changes the role a financial institution can play in a client's life.

Instead of waiting for a client to ask a question or choose a product, the institution can continuously identify where guidance is valuable and engage at the moment a decision is being made.

That creates value in several ways:

More Guidance Opportunities recognized across the client base
Deeper and more frequent client relationships
More consistent Financial Guidance across employees and channels
Greater reach without proportional increases in specialist headcount
More opportunities for deposits, lending, investments and other relevant institutional solutions

The objective is not to generate more interactions. It is to make the institution more useful when financial decisions are actually being made.

How does it fit with the systems a bank already runs?

The system works as an intelligence layer across the technology an institution already runs, so it adds Financial Intelligence without platform replacement. Guidance can reach clients through the institution's existing digital experiences, through the workflows its employees already use, or through a dedicated client application.

The architecture is designed to minimize unnecessary exposure of client information while preserving the context required to produce relevant Financial Guidance. The client relationship stays with the institution throughout.

Who is a Financial Intelligence System for?

It's built for regulated financial institutions, from retail banks to wealth managers and private banks. Different teams inside them tend to approach it from different angles.

In retail banking, the conversation often starts with the primary financial relationship: moving digital banking beyond transactions so the institution can identify important decisions earlier, provide relevant guidance and deepen the client relationship over time.

Wealth leaders often focus on reach. High-quality personalized guidance has historically depended on scarce advisor capacity. A Financial Intelligence System allows the institution's expertise to reach more clients without requiring professional headcount to grow proportionally.

For digital and innovation teams, the priority is moving past chatbots and one-off pilots to a capability the whole institution can run.

Risk, compliance and AI governance teams usually ask the most detailed questions, and rightly so. What matters most to them is that the rules and permissions are the institution's own, that every piece of guidance can be explained, and that people stay in control.

Why is this possible now?

Banks have pursued this goal for a long time. Personal financial management, next-best action, financial wellness programs and digital advice were all serious attempts. Each addressed part of the problem, but important limitations remained: data that didn't reconcile, intelligence that couldn't be trusted in front of a client, real lives that didn't fit segment logic, institutional rules that couldn't be expressed in software, and costs that grew with every relationship served.

Two things have changed. AI can now interpret far more unstructured institutional knowledge and client context than previous generations of software, which addresses several of those limitations at once. The architecture described above then lets an institution use that capability while keeping authority where it belongs. AI can now handle far more complexity without becoming the institution's authority, and that combination is what makes a Financial Intelligence System deployable.

Is this related to financial intelligence in anti-money laundering?

No, although the names overlap. In anti-money laundering, financial intelligence means analyzing transaction reports to detect money laundering and terrorist financing, usually by a national Financial Intelligence Unit. Monstro uses Financial Intelligence to mean something else entirely: the applied intelligence that helps clients make better financial decisions. The two are unrelated.

How is it different from the AI tools banks already have?

Many tools in the market are described in similar terms, so the distinctions are worth setting out. The table below compares the most common ones.

Category	What it does	What it does not solve on its own
Chatbot or virtual assistant	Provides a conversational interface	It does not by itself create the governed Financial Intelligence required to continuously identify and reason through client decisions
Next-best-action	Can prioritize an action or offer	It does not by itself understand and reason across the client's broader financial life
Customer 360	Unifies client information	Information alone does not determine what matters, why it matters or what Financial Guidance should follow
General-purpose LLM layer	Adds powerful interpretation and generation capabilities	It requires additional architecture for institution-defined governance, repeatability, explainability and control

A Financial Intelligence System combines client context, institutional intelligence and financial reasoning to continuously create governed Financial Guidance. It usually works alongside several of these tools, using the same data and reaching clients through the same channels, and adds the reasoning and governance that let an institution put its name to the result.

There is more on how this works in practice on our Financial Guidance and governance pages.