Every few weeks, there is a lot of hue and cry about what AI would do to humankind. The general tirade goes like: AI will replace bankers; AI will replace advisers; AI will replace the stock exchanges; the payment systems, perhaps money itself. Some of these claims come from people with something to sell. Others come from people who are genuinely worried. Either way, the question underneath them is a fair one, and anyone who owns a portfolio or plans a retirement deserves a straight answer.
I have a small stake in the question. I have built an AI-assisted trading workflow of my own, in which an AI handles the login, the scanning and the execution, while I set the strategy and the rules and approve each session with my own one-time password. I have also watched the capabilities of these tools improve at a pace that surprises even those who work with them every day. So, I am neither a sceptic nor a prophet. I am a practitioner trying to think clearly.
The honest starting point is that nobody knows how this will unfold. What we can do is lay out the plausible scenarios, understand the risks in each, and work out what a sensible person should do now, whichever way it goes. That is what this article tries to do.
What is happening today
It helps to begin with what is real, before moving to what is speculated.
AI is already deep inside the financial system. Banks use it to detect fraud and assess credit. Trading firms use it to execute orders and manage risk. Lenders use it to read documents. Chat assistants answer customer queries, and a growing number of investors use AI to read annual reports, compare funds and test ideas.
Regulators have noticed. A committee set up by the Reserve Bank of India published a framework for the responsible and ethical use of AI in the financial sector. SEBI's framework for retail algorithmic trading, tightened how automated trading may be accessed. It requires client-specific API keys tied to fixed IP addresses, two-factor authentication, unique identifiers for algo orders, and registration of algorithms above a certain speed through a broker. Brokers are made responsible for the algo activity on their platforms, and exchanges are to have kill switches to halt a malfunctioning algorithm.
Payments are moving too. NPCI was reported to be preparing a framework, called the Unified Agent Protocol, under which AI agents could make small UPI payments on a user's behalf within limits set in advance, without asking for approval every time. I have not seen the final details, so I will treat this as a direction of travel, not a settled fact. But the direction is clear: software that does not merely advise us about money but acts with it.
So, the question is not whether AI will touch finance. It already has. The question is how far it goes, and who stays in charge.
Scenario one: AI as the assistant
The first scenario is the least dramatic, and in my view the most likely over the next several years. AI becomes a very capable assistant inside the existing system.
The structure stays the same. Banks remain banks, exchanges remain exchanges, regulators remain regulators, and the rupee remains the rupee. But the work inside them is increasingly done by AI. Customer service is faster. Loans are underwritten in minutes. Fraud is caught earlier. Advisers and investors prepare in an hour what used to take a week. Research that was once available only to institutions becomes available to anyone with a phone.
In this world, humans keep the accountability. A bank still answers to the regulator, an adviser still answers to the client, and a person still signs off the decision. The AI does the legwork, and the human owns the result. Most of the regulation now emerging in India and elsewhere assumes exactly this model.
For ordinary people, this is mostly good news: cheaper services, better information and broader access. The main risks are over-reliance, errors that are presented with great confidence, and a new kind of fraud, which I will come to later.
Scenario two: AI agents that act for us
The second scenario goes a step further. AI does not only assist; it acts on our behalf.
Imagine an agent that pays your bills on time, hunts for the cheapest insurance renewal, moves idle cash into a better-yielding account, rebalances your portfolio when it drifts from your plan, and negotiates a better rate on your loan. You set the goals and the limits, and the agent handles the execution. Many of us already use a small version of this when we set up automatic payments or a systematic investment plan. The difference is that an agent can make judgements within the limits.
In this world, money moves between machines. Your agent talks to a merchant's agent, a bank's agent or a lender's agent. Transactions become faster and, for the most part, invisible.
The gains are real: time, convenience, and fewer mistakes caused by forgetfulness or emotion. But there are new dangers. An agent with too much authority can make an expensive error at machine speed. A compromised agent can drain an account before anyone notices. And a person who delegates everything may slowly lose the habit, and the ability, of understanding where their own money goes.
The key design principle is delegation with limits: how much, for what, and with what ability to stop it. The more authority we hand over, the more important those limits become.
Scenario three: markets run by machines
The third scenario concerns the markets themselves. Algorithmic trading is already a large part of volumes in many markets, and AI will add more. If most trading comes to be driven by AI systems that learn from similar data, the character of markets could change.
The International Monetary Fund has already talked about how central banks can contain the financial stability risks from AI. It points to several concerns. One is speed, since decisions move toward real time with little room for human intervention. Another is herding: if many AI-driven funds respond to the same signals, they can amplify swings. An IMF Blog claims that future flash crashes may arise less from coding errors and more from many AI systems reacting in parallel. A further concern is opacity, since complex models make it harder to explain market behaviour under stress. And there is concentration, because critical services often depend on a handful of cloud, data and model providers, so a disruption at one could affect many institutions at once.
I find the herding point the most interesting, because it echoes something very human. Markets have always been driven by crowds, and crowds tend to move together. If AI systems are trained on similar data and optimised for similar goals, they may behave like a very fast, very large crowd. Prices could be efficient most of the time and then, on rare occasions, fall sharply and without warning.
Flash crashes are not a hypothetical. In May 2010, the US market lost about a thousand points on the Dow within minutes and then recovered most of it, in an episode driven largely by automated trading. In India, in October 2012, the Nifty plunged by well over ten per cent in a few minutes after an erroneous order, and was halted by circuit breakers before recovering. Those were failures of far simpler systems than the ones now arriving.
India has also seen a live debate about what very large algorithmic players can do to a market. In July 2025, SEBI issued an interim order against the US trading firm Jane Street, alleging that it had manipulated the Bank Nifty index on expiry days, using large trades in the underlying market to move the index while holding options positions that would benefit. SEBI impounded the alleged unlawful gains and temporarily barred the firm from the market, though the restriction was lifted later after the amount was placed in escrow. Jane Street denies the allegations and says its trading was conventional index arbitrage. As I write, the matter is still being contested before the Securities Appellate Tribunal, so I will not comment on who is right.
But the episode makes a point that applies whatever the outcome. A very well-resourced, very fast algorithmic participant can be large enough, relative to the market, that its trading becomes a force in itself. Add more capable AI, and the questions multiply: how do regulators watch for behaviour that no human designed in detail, how do they tell manipulation from legitimate arbitrage, and what happens to an ordinary participant who is on the other side of a trade when the market is moved for a few minutes? Surveillance will have to become as capable as the systems it watches.
For an individual investor, this does not mean the markets will be unusable. It means that rare, sharp moves may become more frequent or more abrupt, and that strategies depending on a calm, orderly market deserve extra caution.
Scenario four: disruption of the middlemen
A fourth scenario concerns the intermediaries. If an AI system can give personalised advice, compare products, prepare a tax return and execute an investment at almost no cost, what happens to the many people and institutions whose business is to provide these services?
Some will be squeezed. Services that are mostly about processing information, such as basic product comparisons, simple research, routine paperwork and standard recommendations, are the easiest to automate, and their prices are likely to fall. Intermediaries who add little beyond access will find it harder to justify their fees.
But I would not rush to declare the end of the adviser or the bank. Money is not only a technical matter. It is also a matter of trust, accountability and human circumstances. A person deciding whether to retire early, how to divide an inheritance, or what to do after a bereavement is not just seeking information. He is seeking judgement, reassurance and someone who will be answerable if things go wrong. I think those parts of the work will become more valuable, not less, as the technical parts become cheap.
Alongside this, new forms of money and infrastructure are developing: digital currencies issued by central banks, stablecoins, and the tokenisation of assets. These could make settlement faster and the system more programmable. They are changes to the plumbing, and AI will make them easier to use. But they still sit inside legal frameworks and regulation.
Scenario five: full replacement
Now to the boldest claim: that the conventional financial system will be replaced wholesale by an AI-based one.
I think this is unlikely, at least in any near-term timeframe, and the reason is simple. The financial system is not just technology. It is a set of promises backed by law, courts, central banks and the authority of the state. A rupee is money because the law says it is, and because people trust that the system around it will honour its commitments. A deposit is safe, to the extent it is, because of regulation, supervision and insurance. A contract is enforceable because there are courts.
AI can run an enormous amount of the machinery inside such a system. It can make it faster, cheaper and more efficient. But it cannot replace the underlying trust and legal accountability, because someone has to be responsible when something goes wrong, and that someone has to be a person or an institution that can be held to account. Regulators in many countries have said so directly, which is why the emphasis in the frameworks I mentioned is on human oversight, governance and accountability.
So, my view is that the plumbing will change profoundly, the foundations will change slowly, and the idea of an AI replacing the whole system overnight is more a story than a forecast. I could be wrong, and that is why preparation matters.
Scenario six: shocks along the way
There is one more scenario, and it is not a destination but a set of bumps on the road.
Any technology that spreads quickly through a financial system creates new ways for it to fail. The IMF also highlights cyber threats, since generative AI makes attacks faster, more scalable and more convincing. We have already seen deepfake voices and videos used to deceive people into transferring money. A large outage at a major cloud or model provider could affect many institutions simultaneously. A flaw in a widely used model could spread through many firms at once. And there is the broader economic question: if AI changes the demand for certain kinds of work, incomes in some professions could be disrupted. Economists disagree about the speed and scale of this, and I will not pretend to settle that debate.
None of these outcomes is certain. But they are plausible enough that a sensible person plans for them, in the same way one buys insurance without expecting a fire.
Will the philosophy of each asset class change?
There is a deeper question underneath all of this. Beyond who runs the system, will AI change what equities, bonds, crypto and derivatives fundamentally are?
My own view is that the basic philosophies will survive, because each asset class answers a lasting human need, and AI changes how well and how fast that need is met, not the need itself. But how each one behaves in practice could change a good deal. This is reasoning, not forecast, and nobody really knows.
Equities
An equity is a claim on the future earnings of a business. AI does not change that. It changes what the businesses look like, and how quickly prices absorb news.
A company with few employees and a high degree of automation could earn the profits of a much larger one, so yardsticks such as revenue per employee may stop being useful. Public information will be priced in faster, and edges based on speed or on reading documents will shrink, because everyone has the same assistant. The harder question is the moat. If AI makes it cheap to copy a product or service, what stops competition from eroding profits? Judging that is now the central skill of equity investing. And the edge that remains is temperament and time horizon, meaning the ability to hold through noise.
Bonds
A bond is a promise to repay, backed by the issuer's capacity and the legal system. AI will make credit assessment finer and faster, so borrowers may be priced more individually and more continuously, and tokenised bonds could trade and settle more easily. But the philosophy is trust in the issuer and in the courts, and AI cannot replace either.
The risk I would watch is that if many lenders use similar models, they may tighten or loosen credit all at once, making credit cycles sharper.
Crypto
This is where the philosophy is most open. Crypto began as an idea about money without a trusted middleman, with scarcity and independence as its virtues. AI offers it a possible new role. If software agents are going to pay each other, programmable digital money is a natural fit, since machines do not need a bank account the way humans do.
The counterargument is that conventional rails are adapting quickly, as the reported UPI plans for agent payments suggest, so the new role may be served without crypto at all. So, the philosophy could shift from ideology, the "store of value" and distrust of institutions, toward plumbing for machine-to-machine payments. That outcome is far from settled. AI also makes crypto scams more convincing, which could hurt trust in the space.
Derivatives
A derivative transfers risk from someone who does not want it to someone who does, and that function stays essential. AI could widen it, with more tailored hedges for ordinary investors, cheaper pricing and better risk management for institutions. It is even possible that new risks, such as income disruption in particular professions, will create demand for new ways of hedging.
But AI could also widen the gap between well-equipped professionals and retail speculators, since it helps the market-maker and the quant far more than someone buying short-dated options on expiry day. And there is a systemic worry: if many AI systems hedge in the same way at the same time, they could amplify a move.
Where something deeper might shift
I see three places where the change could go beyond the surface.
The first is price formation. If most trading is done by AI agents reacting to each other, prices may reflect machine reactions more than human judgement. Markets could be efficient most of the time and then very abrupt on rare occasions.
The second is ownership. If businesses can be run largely by AI, the question of who owns them and who is accountable becomes harder. I think law and regulation will keep a human or an institution responsible for a long time.
The third is where the edge lies. When information and speed are cheap, an advantage comes from patience, discipline and an understanding of your own behaviour. That is a very old idea, and AI may make it more important, not less.
Overall, ownership, lending, risk transfer and the movement of value will all remain. The faster, cheaper and more machine-driven the system gets, the more it matters that the individual investor keeps his own plan, risk limits and judgement.
The pre-AI age: how financial innovation arrived
It helps to remember that the financial system has been reshaped before, and often. Let me call everything up to now the pre-AI age. Looking at how the changes of that age arrived tells us something about what the AI age might bring.
In the pre-AI age, the landscape was changed by central banks becoming the anchor of national money systems, by the Bretton Woods arrangement in 1944 and then its collapse in 1971, when the dollar's link to gold was ended and the world moved to money backed by trust in governments. It was changed by credit cards and ATMs, by the first electronic stock quotation system in 1971, by listed options in 1973, by index funds in the 1970s and exchange-traded funds in the 1990s, by securitisation, and by online brokerage. In India, the arrival of UPI in 2016 made instant digital payments an everyday habit for hundreds of millions of people. And in 2009, Bitcoin began operating, the most striking change of recent times and the one that set off the whole crypto industry.
If you look at these changes together, a few patterns repeat.
They lowered cost and widened access. An index fund let an ordinary person own the whole market cheaply. UPI made a payment as easy as sending a message. Each innovation took something that was once expensive or exclusive and made it available to many.
They created new instruments. Listed options, securitised loans and exchange-traded funds were things that simply did not exist before. Each opened new ways to take, spread or transfer risk.
They shifted where trust sits. Gold gave way to government promises. Paper gave way to electronic records. Bitcoin proposed that trust could rest in code and mathematics instead of an institution. Each change moved the question of whom, or what, we are relying on.
They created new risks, often noticed late. Securitisation was a clever way to spread loan risk, and it also helped spread the problems that led to the financial crisis of 2008. Innovations tend to be celebrated first and understood later.
They often came from outside, and regulators followed. Bitcoin was created by an anonymous person, grew outside the banking system, went through waves of enthusiasm, speculation and collapse, and was met by regulators only after it had become large. Fraud and failures marked each stage.
The AI age: what might arrive
Now apply the same patterns to the AI age. A few possibilities seem plausible to me. None is certain, and some may not happen.
A personal fund manager for everyone. Just as the index fund put the whole market within reach, a capable AI adviser could give ordinary people the kind of planning and portfolio oversight that used to be reserved for the wealthy. This could be the biggest widening of access since online brokerage, with the risk that people trust it blindly.
New instruments built around AI itself. The machinery of AI needs enormous amounts of computing power, data centres and chips, and lenders are already financing these. It is easy to imagine more of the financial world forming around them: markets for computing capacity, loans secured on hardware, instruments linked to the revenues of AI models, and contracts that let businesses hedge the cost of computing. The more important something is to the economy, the more likely it is that someone will build a market in it.
Personalised risk products. If AI can assess an individual's circumstances closely, insurance and hedging could become far more tailored, covering risks, such as a particular profession's exposure to automation, that are hard to price today.
A machine-to-machine economy. When agents pay and negotiate with other agents, constantly and in tiny amounts, new payment forms may be needed, and the line between money, data and service may blur.
A new basis for trust. Credit has long depended on history and collateral. AI could widen it by reading alternative data, bringing credit to people the system has overlooked. But it could also entrench bias in ways that are hard to see, and it raises the question of who audits a decision made by a model that no one fully understands. Trust may come to rest on how well systems can be checked and explained.
New systemic risks of its own. If the lesson of securitisation is that a clever risk-spreading tool can hide a shared weakness, the equivalent in the AI age is many institutions relying on similar models, similar data and a few providers. That is the herding and concentration risk the IMF has flagged.
Innovation from outside, and a hype cycle. Perhaps the most useful lesson from crypto is about how new things arrive. The next big financial innovation may well come from developers and start-ups, not from established institutions, and regulators may arrive after it has already grown. We should expect the full cycle: genuine breakthroughs, wild speculation, a good deal of fraud, and some spectacular failures. Claims of "AI-powered" returns are already a common sales pitch, and I would treat them with the same caution that investors should have applied to the early days of any new craze.
The difference between the two ages is speed. Innovations of the pre-AI age took decades to spread. Software can spread in months, which means that both the benefits and the mistakes may arrive faster than our institutions, or our own habits, are used to.
What is in store for humans
If I step back from the scenarios, a pattern emerges. As AI takes over more of the execution, the scarce human skills shift.
Execution becomes cheap. Judgement does not. Information becomes abundant. Knowing which information to trust does not. Technical analysis becomes automated. Understanding a client's fears and aspirations does not.
I expect the human role in finance to move toward setting goals and limits, making judgement calls under uncertainty, building trust, and taking accountability. These are skills that grow with experience and character, and they are the ones that cannot easily be copied.
I see this in my own small experiment. The AI handles the repetitive tasks with speed and discipline. But I still define the strategy, set the risk limits, decide when to pause because conditions have changed, and approve each trading session personally. The system executes the plan. The plan is mine. That division of labour seems to me a good template for the wider world.
How to protect what you already have
Whichever scenario unfolds, a few principles have served investors through every technological change, and I believe they will serve through this one.
Diversify properly. Do not put everything in one asset class, one sector, one country or one currency. This holds especially for the theme everyone is excited about. Whether or not the valuations of AI-linked companies prove justified, concentrating your wealth in a single story is a risk. Investors in the technology boom of the late 1990s were right about the technology and still lost heavily because they paid too much and held too little else.
Own productive assets for the long term. If AI does raise productivity, much of the gain will flow to the owners of businesses. A broad, diversified holding of equities, held patiently, lets you share in that growth without having to guess which company will be the winner.
Keep liquidity and avoid leverage. Fast markets and sharp moves are hardest on those who are forced to sell. An emergency fund, money set aside for near-term needs, and a refusal to borrow to invest, give you the ability to sit through a shock.
Protect your human capital. For most people, the largest asset is not the portfolio. It is their ability to earn. In a period of rapid change, the most valuable investment is probably continued learning. Learn how to use these tools in your own field and develop the judgement and relationship skills that AI is least able to replace. Think about whether your income depends on a single skill or employer and build a second string.
Strengthen your security. The rise of AI makes fraud more convincing. Treat any unexpected request for money or information with suspicion, even if the voice sounds like your son or the video looks like your bank manager. Agree on a family code word. Use two-factor authentication everywhere. Never share one-time passwords. Verify through a channel you already know, and slow down when someone insists on urgency.
Put limits on any authority you delegate. If you let software act for you, whether a trading automation or a payment agent, set firm limits on amounts, frequency and scope, keep a way to switch it off, and review its activity. Start small. In my own workflow, the rules include a mandatory stop loss, caps on position size and a daily loss limit, and the principle that without a rule there is no trade.
Use AI as an assistant and verify. AI can read a balance sheet in seconds and explain a complex product in plain words. But it can also be wrong with total confidence. Use it to widen your understanding and check anything important against primary sources or a qualified professional.
How to plan for growth in uncertain times
Protection is half the story. The other half is continuing to build.
Keep investing regularly, regardless of the headlines. Uncertainty is not a reason to stop, because the markets have always been uncertain. It is a reason to be disciplined about how much risk you carry.
Decide your allocation in advance and rebalance by rule, not by emotion. When a hot theme runs far ahead of everything else, a rule that trims it protects you from your own enthusiasm.
Stay curious. The people who will do best are likely to be those who learn how these tools work, use them thoughtfully and keep adapting. You do not need to become a technologist. You need to become comfortable asking good questions of a very capable assistant.
And revisit your plan regularly. Ask yourself a few scenario questions. What if my income fell sharply for a year? What if a major market fell by a third? What if my main account were frozen for a week? If the answers make you uneasy, that is where to strengthen the plan.
Keep the ability to navigate by yourself
I will end with something from my years in the Navy.
Modern ships and submarines carry extraordinary navigation systems. They are accurate, fast and reliable, and nobody would dream of doing without them. But every crew is still trained to navigate without them, with the help of charts, instruments and dead reckoning, because electronic systems can fail, be jammed or mislead. The best navigators use the technology fully, while keeping the ability to find their way without it.
I think that is the right attitude toward AI in finance. Use it, and use it fully, because it is a remarkable tool and it will keep improving. But keep the ability to understand your own money, to make your own decisions and to take your own responsibility. Keep a little cash and a little knowledge that does not depend on any system. Keep your plan simple enough to explain to your family.
The future of finance will almost certainly involve far more AI than today. Whether that future is kind to individuals will depend less on the technology than on how prepared the people living in it are. The tools will change, but the principles of diversification, patience, discipline, security and continuous learning will not. Technology will do more of the work. We still own the plan.
Disclaimer: This article reflects personal views for educational purposes only. It describes possible scenarios and is not a forecast, investment advice or a recommendation to buy or sell any security. Regulations and technologies are changing quickly, and details mentioned here may have changed. Please consult a qualified financial planner or adviser before making decisions about your own situation.