Finance Firms That Don’t Use AI in 2026: Who Holds Back?

The Pulse

Finance firms that don’t use AI still exist in 2026, but the phrase is becoming misleading. Among established banks, insurers, asset managers and fintechs, complete avoidance is increasingly unusual because machine learning is already embedded in fraud detection, cybersecurity, credit analysis, operations and customer service.

The 2026 Global AI in Financial Services Report from Cambridge Judge Business School found that 81% of surveyed financial-services firms were adopting AI at some level, although only 40% had reached advanced scaling or transformation. The European Central Bank separately says more than 85% of significant banks under European banking supervision already use AI. [Cambridge 2026 Global AI in Financial Services Report] [ECB AI adoption among European banks]

The more useful divide is therefore not AI versus no AI. It is how much decision-making a financial institution is willing to delegate. Some firms use traditional machine learning quietly in the background while restricting generative AI. Others allow internal copilots but keep AI away from customer advice, credit decisions and autonomous transactions. The real 2026 story is the depth of AI delegation.

Core Significance

Why it matters:

• A reliable list of major finance firms using zero AI is difficult to produce: Many financial institutions have used statistical machine learning, fraud models, anomaly detection and automated risk systems for years even if they restrict ChatGPT-style tools or autonomous agents.
• Finance firms distinguish traditional AI from generative AI: The American Bankers Association reported in 2026 that banks view traditional AI such as automation, analytics and pattern recognition as relatively mature, while generative AI remains earlier-stage and more tightly controlled because of data security, accuracy and regulatory concerns. [ABA 2026 bank AI adoption survey]
• Holding back is not always technological conservatism: For customer advice, lending, insurance pricing, regulated decisions and autonomous transactions, a slower rollout can be rational when the cost of an incorrect AI decision exceeds the productivity gain from automating it.

Deep Context: Are there really finance firms that don’t use AI?

There are financial institutions with little or no meaningful AI deployment, but public evidence rarely supports naming major firms as complete non-users. A company may prohibit public generative AI tools while still running machine learning in fraud systems, anti-money laundering, cybersecurity, document classification or forecasting.

The Bank of England and FCA’s last published full AI survey illustrates the point. In the 2024 benchmark, 75% of responding financial firms were already using AI and another 10% planned to use it within three years. Insurance firms reported 95% adoption and international banks 94%, while financial-market infrastructure firms had the lowest reported adoption at 57%. [Bank of England and FCA AI in financial services survey]

The Bank of England and FCA launched another sector-wide AI survey in June 2026 covering machine learning, foundation models, generative AI and agentic AI. That alone shows how quickly the definition of AI adoption is changing: a binary question about whether a bank “uses AI” now hides several very different maturity levels.

The five levels of AI adoption in finance

Level 0 is no meaningful AI. Workflows remain primarily manual or rules-based, with conventional software, spreadsheets and human judgment handling most decisions. This category still exists, but it is increasingly concentrated away from large, technology-intensive institutions.

Level 1 is invisible AI. The institution uses machine learning for fraud detection, AML alerts, anomaly detection, forecasting, cybersecurity or scoring, but employees and customers may rarely interact directly with an AI interface.

Level 2 is controlled generative AI. Employees use approved AI for research, summarisation, drafting, software development, document review and internal knowledge work while sensitive data and customer-facing outputs remain restricted.

Level 3 is workflow AI. AI becomes part of underwriting, servicing, claims, compliance, financial analysis, customer support or advisory workflows, usually with humans retaining approval rights for high-impact decisions.

Level 4 is agentic finance. AI can reason, access tools, retrieve customer or market data, execute multi-step processes and potentially take actions within predetermined permissions. This is the layer creating the hardest questions around accountability, identity, security and human control.

That spectrum is more useful than asking whether a finance firm uses AI. Two banks can both answer yes while one runs a fraud model in the background and the other lets AI agents coordinate complex workflows across multiple systems.

Where low-AI finance still exists

The strongest evidence points toward uneven adoption rather than one identifiable group of non-users. Cambridge found fintechs ahead of traditional financial institutions in advanced AI adoption, with 47% of fintech respondents at advanced stages compared with 30% of incumbents. Only 14% of financial-industry respondents considered AI transformational to organizational strategy and competitive advantage.

Smaller institutions can also face a different economic equation. AI implementation requires data engineering, vendor management, security review, staff training and governance capacity that larger banks can spread across more customers and transactions. The cost of building those capabilities can make broad AI adoption less attractive for smaller institutions with narrow product lines.

That does not mean community banks are ignoring AI. A 2026 American Bankers Association survey found that roughly half of surveyed community banks planned to use new technologies such as AI to reduce costs, while separate ABA research characterized bank adoption as uneven and cautious rather than absent.

Relationship-heavy businesses are another area where adoption can remain selective. Private wealth advice, complex institutional transactions, bespoke corporate finance and other high-value interactions depend heavily on trust, context, negotiation and accountability. AI can support those professionals without necessarily becoming the final decision-maker.

Why some finance firms still hold back

The first constraint is data. Cambridge’s 2026 financial-services research found data availability and quality was a major adoption barrier for 40% of industry respondents. Among AI vendors serving financial institutions, 72% cited data quality and completeness problems, 46% cited legacy and siloed systems, and 41% cited data-sharing restrictions.

The second constraint is risk. The same study found 74% of financial-services respondents identified data privacy and protection among the main AI risks, while 70% cited hallucinations or unreliable outputs and 55% cited loss of human oversight. These concerns become more serious when AI touches lending, investments, insurance or customer money.

The third constraint is third-party dependence. The Bank for International Settlements’ Financial Stability Institute warns that advanced AI intensifies existing financial-sector concerns around data privacy, quality and security while increasing dependencies on external providers and concentration among major technology vendors. [BIS FSI AI data use in financial services]

The fourth constraint is regulatory accountability. In the United States, the OCC, Federal Reserve and FDIC updated their model-risk approach in 2026 while explicitly noting that generative and agentic AI are rapidly evolving and warrant further consideration. For banks, uncertainty does not remove responsibility for governance, third-party controls and safe deployment. [OCC updated model risk management guidance]

Customer trust creates another boundary

The case for AI is strongest when the task is repetitive, information-heavy and easy to verify. The case becomes weaker when a customer expects judgment, accountability or reassurance around a major financial decision.

The FCA’s 2026 Mills Review found meaningful consumer interest in agentic finance, with 20% of surveyed UK retail financial-services consumers saying they would be likely to use AI capable of acting autonomously within preset goals. But the same work found concerns around trust and control, reinforcing why firms cannot treat autonomous financial services as a pure technology rollout. [FCA Mills Review on AI in retail finance]

As covered in our AI finance regulation analysis, financial institutions do not only need AI systems that perform well. They need systems whose decisions, risks and controls can be defended when customers, auditors or regulators challenge them.

The gap is increasingly about intensity, not access

Research across the wider economy shows the same pattern. ECB analysis published in June 2026 found that more than 70% of euro-area firms used AI at least occasionally, but only 7% reported intensive use. Among firms trying to deepen adoption, 40% cited shortages of AI skills, 28% questioned the usefulness of current technology for their needs, and 26% cited incompatibility with existing systems. [ECB intensive AI adoption analysis]

That is a better framework for finance too. Buying access to an AI assistant is easy. Integrating AI into core lending, risk, advice, trading, compliance or payment processes requires data, controls, infrastructure, skills and executive accountability.

As covered in our AI enterprise governance 2026 analysis, the deeper AI moves into business operations, the more governance becomes part of the deployment architecture rather than a policy document added afterward.

Data Insights

By the numbers:

The datasets below measure different financial markets and stages of adoption. They should not be combined into one universal AI adoption rate.

• 81% of surveyed financial-services firms are adopting AI at some level: Cambridge’s 2026 global study found 40% at advanced scaling or transforming stages, while only 14% said AI was already transformational to organizational strategy and competitive advantage.
• Fintechs are deeper adopters than incumbents: Advanced AI adoption reached 47% among surveyed fintechs compared with 30% among incumbent financial institutions, showing that the divide increasingly concerns deployment depth rather than whether firms have experimented with AI at all.
• More than 85% of significant European banks already use AI: ECB supervisory data shows fraud and cybercrime detection among the most common uses, followed by marketing, chatbots and credit scoring.
• 40% of surveyed finance firms report increased profitability from AI while 43% report no change: Cambridge’s findings show why cautious institutions can look at the same technology and reach different investment conclusions. AI value is emerging, but it is not yet universal.
• 20% of UK retail-finance consumers say they are likely to use autonomous AI within preset goals: The FCA estimates that share represents roughly 11 million UK adults, suggesting demand for agentic finance is developing even while trust and control remain concerns.

Table 1: What “not using AI” means in finance

Adoption level	What the firm is doing	Typical examples	2026 position
Level 0: No meaningful AI	Primarily manual and deterministic software processes	Spreadsheets, rule-based workflows, human review	Increasingly uncommon among large financial institutions
Level 1: Invisible AI	Uses traditional machine learning without broad GenAI access	Fraud detection, AML, anomaly detection, forecasting	Mature and widespread
Level 2: Controlled GenAI	Allows approved assistants inside defined internal boundaries	Research, drafting, coding, document review	Rapidly expanding
Level 3: Workflow AI	AI participates directly in business processes	Servicing, underwriting support, claims, compliance	Growing with stronger controls
Level 4: Agentic finance	AI can use tools and take bounded actions	Multi-step service, operations and personal finance agents	Early and tightly governed

Table 2: Which finance workflows resist AI most

Finance workflow	AI adoption tendency	Main barrier	Human role
Fraud and cyber detection	Established	False positives, adversarial attacks, monitoring	Investigate and escalate higher-risk cases
Internal research and document work	Rapidly growing	Data leakage, hallucinations, source verification	Review and approve outputs
Customer support	Growing	Accuracy, disclosure, complaints and escalation	Handle exceptions and sensitive interactions
Credit and underwriting	Controlled	Explainability, fairness, consumer impact	Approve significant or exceptional decisions
Investment and wealth advice	Selective	Suitability, liability, trust and fiduciary expectations	Provide judgment, context and accountability
Autonomous transactions	Early	Permissions, fraud, authentication and irreversible actions	Set limits and approve high-impact actions

The Business Case: When holding back from AI makes sense

Finance firms should not adopt AI simply because competitors are deploying it. The correct starting point is a specific workflow where AI can reduce cost, improve speed, strengthen risk detection or expand service without creating disproportionate new risk.

A low-volume, highly bespoke workflow may not justify a large AI implementation. If experienced professionals already complete the work efficiently and every case requires substantial judgment, integration and governance costs can exceed the automation benefit.

The same logic applies when data is not ready. Connecting a sophisticated model to fragmented customer records, inconsistent product definitions and poorly governed historical data does not create intelligence. It creates faster inconsistency.

Where the task is repetitive and measurable, holding back becomes harder to justify. Document review, fraud analysis, compliance research, software development, knowledge retrieval and internal operations offer bounded workflows where institutions can preserve human accountability while testing measurable productivity gains.

The next step should be controlled expansion rather than immediate autonomy. A financial institution can move from AI recommendations to AI-assisted execution and only then toward autonomous actions once permissions, audit trails, incident response and human override have been proven.

As covered in our AI agents enterprise news 2026 analysis, the risk profile changes sharply when AI gains the ability to act rather than simply recommend. Finance firms should govern autonomy separately from model intelligence.

When caution becomes a competitive problem

Caution stops being rational when a firm repeatedly performs high-volume information work manually while competitors automate the same task safely. At that point the institution is not protecting itself from AI risk. It is accumulating operating-cost and capability disadvantages.

The cost can also appear through vendors. Core banking, CRM, fraud, cybersecurity, productivity and cloud platforms are embedding AI whether financial institutions build models themselves or not. A firm that does not develop internal AI competence may eventually depend more heavily on vendors to explain and control systems already inside its technology stack.

That is why the ABA’s 2026 bank research reached an unusual conclusion: many bankers viewed doing nothing with AI as a greater long-term risk than cautious adoption. The concern was not only missed productivity. It was loss of internal expertise, greater vendor dependence and declining competitiveness.

Expert Nuance: AI avoidance is often boundary-setting

The most important distinction in this search query is between refusing AI and refusing a specific type of AI deployment.

A bank can ban employees from pasting customer data into public generative AI tools while simultaneously using machine learning for fraud detection. A wealth manager can use AI to summarize research while refusing to let the system provide final investment advice. An insurer can automate document extraction while keeping underwriting decisions under human authority.

Those organizations are not anti-AI. They are setting boundaries around data, autonomy and liability.

This matters because financial AI is moving toward more consequential actions. AI that recommends the next step is easier to govern than AI that executes the step. A research assistant can be reviewed before its work matters. An agent moving money, changing account permissions or triggering customer actions requires identity, authorization and rollback mechanisms.

The Financial Stability Board’s 2026 consultation on responsible AI adoption makes the same shift visible at the system level. It argues that financial institutions need organization-wide governance across the AI lifecycle as rapid adoption introduces or amplifies risks that can affect resilience and financial stability. [FSB responsible AI adoption in finance]

The better strategic question for finance firms is therefore not whether to use AI. It is which decisions remain human, which tasks become AI-assisted, which workflows can be delegated, and how much autonomy the institution can safely supervise.

Strategic Outlook

1. Expect true no-AI finance firms to become rarer: AI is being embedded inside core software, security, analytics and productivity platforms, which means institutions can adopt AI indirectly even without building an internal AI program.
2. Expect the adoption debate to move toward autonomy: The important distinction will increasingly be whether AI can only analyze and recommend or whether it can execute transactions and workflows under delegated authority.
3. Expect low-risk internal workflows to move first: Research, compliance support, document processing, software development and knowledge retrieval allow firms to build AI capability without immediately placing models in control of customer money or regulated decisions.
4. Expect vendor concentration to make non-adoption harder: As a small number of cloud, model and software providers become embedded in financial AI, institutions will need enough internal expertise to govern dependencies even when the underlying AI is supplied by someone else.
5. Expect human judgment to retain a premium in high-trust finance: AI can make advisers, bankers and analysts faster, but liability, negotiation, client context and accountability give humans a durable role in consequential financial decisions.

Key Question Answered

Are there finance firms that don’t use AI in 2026?

Yes, some finance firms still use little or no meaningful AI, but there is no reliable public list showing that major financial institutions completely avoid artificial intelligence. Sector surveys instead show that AI adoption is already widespread among banks, insurers, asset managers and fintechs.

What varies much more is the depth of adoption. Some firms use traditional machine learning for fraud, risk or cybersecurity while restricting generative AI. Others allow internal copilots but prohibit AI from making customer-facing decisions. Advanced institutions are beginning to embed AI directly into workflows and experiment with autonomous agents.

The firms most likely to hold back are therefore not necessarily firms rejecting AI entirely. They are institutions limiting AI to lower-risk tasks because of data quality, legacy systems, skills shortages, privacy, explainability, regulatory risk, customer trust or uncertain economics.

FAQ

1. Which finance firms don’t use AI?

There is no authoritative public list of major finance firms that use no AI at all. Large banks and insurers increasingly use AI somewhere in their operations, even if they restrict generative AI or autonomous systems. Lower adoption is more likely to appear in smaller institutions, selected infrastructure firms and relationship-heavy businesses with limited automation economics.

2. Why do some finance firms avoid AI?

The main reasons include poor data quality, legacy-system incompatibility, privacy and security concerns, hallucinations, explainability requirements, regulatory uncertainty, skills shortages, vendor dependency and uncertainty over whether the financial return justifies implementation cost.

3. Do community banks use AI?

Many community banks are exploring or adopting AI, but adoption is uneven and generally more cautious than at large technology-intensive institutions. Common starting points include policy analysis, compliance support, document review, internal research, fraud detection and other workflows where AI supports rather than replaces human judgment.

4. Will AI replace finance jobs?

AI is more likely to automate individual finance tasks than eliminate every role built around them. Repetitive research, document processing, reconciliation and analysis can be automated heavily, while client relationships, negotiation, accountability, complex judgment and regulated approvals continue to require substantial human involvement.

5. Is avoiding AI still a viable strategy for finance firms?

Selective restraint can be rational when a use case has high liability, poor data or weak economics. Avoiding AI across the entire institution is increasingly difficult to sustain because AI is becoming embedded in cybersecurity, fraud prevention, software, analytics and enterprise platforms used throughout financial services.

The Takeaway

The finance industry is not dividing into firms that use AI and firms that do not.

It is dividing by depth of delegation. Some institutions use AI invisibly for fraud and analytics. Others allow controlled generative AI inside the workforce. More advanced firms are redesigning business processes around AI, while a smaller frontier is experimenting with agents that can act across financial systems.

That makes the literal question, which finance firms don’t use AI, less important than it first appears. The strategic question is where each institution draws the boundary between human judgment, AI assistance and autonomous action.

Holding back can still make sense where data is poor, liability is high or automation produces little economic value. But complete avoidance is becoming its own risk as competitors build internal AI expertise and vendors embed AI into the infrastructure financial firms already depend on.

The strongest finance firms will not necessarily automate the most. They will be the ones that know exactly where AI produces value, where humans still matter, and where giving a machine more authority creates more risk than advantage.