The investor's key task is to resist emotion and to make decisions based on analysis rather than on intuition and fear. This thesis becomes particularly relevant in the context of the new wave of automation: when AI agents are increasingly used to support or perform financial processes, human oversight and analytical discipline take on even greater weight.
The financial sector has experienced several technological revolutions, including electronic banking, compliance automation, the rise of algorithmic trading, and digital customer service channels. Each has been accompanied by forecasts that new technology would radically transform the finance profession.
With artificial intelligence, the situation looks somewhat different. Increasingly, we are not dealing merely with a tool that helps a person work faster. Businesses are adopting solutions capable of performing actions on their own: analysing documents, screening counterparties, initiating payments, and interacting with corporate systems without an employee's constant involvement.
This is why the subject of AI agents attracts such interest today — not only among technology companies, but also among CFOs, accountants, auditors, and business owners.
Over the past two years, businesses have been actively adopting generative AI. At first, companies used it to prepare documents, search for information, and automate correspondence. More recently, a further stage of development has emerged. Digital assistants are giving way to so-called agentic AI systems, capable not only of answering questions but of carrying out actions independently.
For the finance function, this means potentially significant changes. Where the algorithm once helped an employee reach a decision, it now increasingly has the ability to initiate processes, screen counterparties, approve documents, and even participate in the movement of funds.
From Assistant to Digital Colleague
Agentic AI is generally understood to mean a system capable of independently completing an assigned task, making intermediate decisions, and interacting with other digital systems without continuous human involvement.
Goldman Sachs Asset Management, for example, has described agentic AI as a further stage in the evolution of automation, noting that it goes beyond traditional automation:[1]
“… agentic AI assumes responsibility for complete business processes, autonomously making decisions and managing exceptions …”
For now, such solutions are used extensively within large corporations, but the trend towards wider adoption is becoming increasingly visible. Lloyds Banking Group uses AI to automate fraud investigations and to handle complex customer inquiries.[2] J.P. Morgan, for its part, notes in its research that corporate treasuries increasingly view artificial intelligence as one of the priority technologies of the coming years.[3]
These technologies are becoming more accessible to small and medium-sized businesses as well. Services are already appearing on the market that can automatically verify documents, analyse expenses, generate payment orders, and monitor accounts receivable.
The Risks Technology Vendors Rarely Mention
AI solution presentations often focus on gains in productivity and efficiency. Yet as autonomy grows, a question arises that has traditionally interested CFOs far more: who will answer for a mistake.
Consider the following scenario. An AI agent is granted access to the payment system, checks the counterparty, generates the documents, and initiates a transfer of funds. It later emerges that the bank details had been substituted by fraudsters, or that the transaction itself violated the company's internal regulations. Who bears the responsibility? The answer depends on the applicable legal and organisational framework. To manage such risks, the final stage requires rigorous controls and oversight.
An algorithm does not itself bear legal responsibility. Responsibility still rests with the organisation and the people responsible for its use, subject to the applicable legal and regulatory framework. This is precisely why the European Union, under the AI Act, devotes such attention to transparency and to the ability to explain the actions of automated systems.[4] In parallel, the European Securities and Markets Authority (ESMA) has issued a statement (ESMA35-335435667-5924) confirming that existing MiFID II requirements apply to firms' use of AI in retail investment services, including requirements relating to organisational arrangements, conduct of business and acting in the best interests of clients.[5]
The more authority an algorithm is granted, the more important it becomes to understand the logic behind its decisions and to preserve the possibility of human intervention.
Deepfakes Are Becoming a Financial Risk
A few years ago, deepfakes looked more like a technological experiment. Today the picture has changed. Banks and financial institutions increasingly identify deepfakes as a potential fraud risk.
According to the World Economic Forum[6], the use of generative AI is already substantially changing the nature of cyber threats. Modern models can produce convincing voice messages, video calls, and digital identities, while specialised software can also be used in attempts to circumvent remote customer identification procedures.
The limitations of AI systems can also be illustrated by differences in how individual models respond to the same task. In practice, even advanced systems can produce errors, including errors that may appear convincing to the user.
In finance, the consequences of such errors may be more significant than an inaccurate answer in an ordinary chat. If an algorithm is entrusted with screening counterparties, executing payments, or identifying clients, the question of trust becomes a question of risk management.
This is exactly why the advance of artificial intelligence calls not for the removal of human involvement but, on the contrary, for a robust and comprehensive system of control. The more authority algorithms receive, the greater the responsibility of the person making the final decision must be.
What Businesses Should Consider
The financial sector has experienced several large-scale waves of digital transformation. There have been periods when businesses moved to digitise virtually every process: working groups were formed, KPIs were defined, and new systems were rolled out. Over time, however, it became clear that digitalisation by itself does not solve problems that already exist inside a business.
There is a simple analogy. If a car's suspension is worn out and a wheel is about to fall off, installing a modern sensor will not make the vehicle sound. At best, the system will warn you in advance that the fault is becoming critical. But it cannot repair the wheel.
With artificial intelligence, the situation is much the same. It is a tool for analysis and automation, but it cannot compensate for weak business processes, poor-quality data, or the absence of internal controls. What is more, AI may amplify such problems at scale and potentially do so faster than a human can.
The introduction of AI can therefore be viewed not only as an IT project but also as a project involving changes to management processes. Before delegating counterparty screening, payment approval, or risk analysis to an algorithm, it is important to define the boundaries of its authority, the control procedures, and the allocation of responsibility. These considerations can help organisations establish appropriate controls around the use of the technology.
Data matters no less. Even the most sophisticated model depends on the information it receives as input. If that data is incomplete or contains errors, the quality of the resulting output may be affected.
Sources
1. Goldman Sachs Asset Management, “Agentic AI Explained”, 19 May 2026. https://am.gs.com/en-us/advisors/insights/article/2026/agentic-ai-explained
2. Lloyds Banking Group, “2026: The year of Agentic AI, and a new era for finance”, February 2026, https://www.lloydsbankinggroup.com/insights/2026-the-year-of-agentic-ai-and-a-new-era-for-finance.html; and “Lloyds Banking Group deploys agentic AI to strengthen real-time fraud protection”, press release, 8 June 2026.
3. J.P. Morgan, “Treasurers in Europe shift priorities: Polling data highlights evolving strategies amid volatility” (Treasury Outlook 2026), 22 June 2026. https://www.jpmorgan.com/insights/banking/emea-treasurers-outlook
4. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (AI Act).
5. ESMA, Public Statement on the use of Artificial Intelligence in the provision of retail investment services, ESMA35-335435667-5924, 30 May 2024. https://www.esma.europa.eu/sites/default/files/2024-05/ESMA35-335435667-5924__Public_Statement_on_AI_and_investment_services.pdf
6. World Economic Forum, in collaboration with Accenture, Global Cybersecurity Outlook 2026, January 2026. https://www.weforum.org/publications/global-cybersecurity-outlook-2026/
Skanestas Investments Limited is a Cyprus Investment Firm authorised and regulated by Cyprus Securities and Exchange Commission (CySEC) under license CIF 251/14. This article is a marketing communication provided for general information only. It does not constitute investment advice, a personal recommendation, an offer to buy or sell any financial instrument, or an independent investment research. Investing involves risk, including possible loss of capital. Past performance and any examples referred to are not indicative of future results. Please seek independent professional advice before making any investment decision. This article is not an offer to provide investment services in any country where Skanestas is not authorised to do so, or where such an offer would be prohibited by applicable law. Information is current as at the date of publication and may change without notice.
