AI Agents Are Taking Over Finance and Compliance: What Goldman Sachs and BNY Signal for Every Business

In the first week of February 2026, Goldman Sachs announced it is deploying autonomous AI agents — built on Anthropic's Claude — to handle core accounting, compliance, and operational finance functions. Days later, BNY (America's oldest bank, formerly Bank of New York Mellon) revealed its internal AI platform "Eliza," which integrates multiple AI models with the firm's data and compliance frameworks across its entire workforce.

These aren't chatbots answering customer questions. These are autonomous agents making decisions, processing transactions, flagging regulatory issues, and generating reports — with minimal human intervention.

If you're running a UK business and thinking "that's just Wall Street," think again. What starts in investment banking filters down to mid-market within 18 months. Here's what you need to know.

What Goldman Sachs Is Actually Doing

Goldman's deployment isn't experimental. They're using AI agents for:

• Reconciliation and accounting: Agents autonomously match transactions, identify discrepancies, and process corrections across multiple systems
• Compliance monitoring: Real-time scanning of communications, transactions, and filings for regulatory violations
• Operational finance: Automating report generation, audit preparation, and regulatory submissions
• Exception handling: Instead of routing every anomaly to a human, agents assess severity, apply rules, and escalate only genuine concerns

The key word is "autonomous." These agents aren't suggesting actions for humans to approve — they're executing within defined guardrails. A human reviews outcomes, not inputs.

BNY's "Eliza" Platform: AI for Everyone

BNY's approach is equally significant but different in philosophy. Rather than deploying specialised agents, they've built a platform that:

• Integrates open-source and commercial AI models with internal data
• Provides every employee access to AI tools tailored to their role
• Required all staff to complete a 10-hour AI training programme
• Connects AI capabilities directly to compliance and governance frameworks

This is the "AI-native workplace" model — where AI isn't a separate tool but an integrated layer across every function. BNY is spending billions on this transformation, and they're not alone.

Why This Matters for UK Businesses

The Compliance Burden Is Real

UK businesses face a uniquely complex regulatory environment: HMRC's Making Tax Digital requirements, FCA regulations for financial services, GDPR for data handling, and sector-specific compliance obligations. Most mid-market companies handle this with a combination of spreadsheets, manual checks, and expensive advisors.

AI agents can fundamentally change this equation:

• Making Tax Digital (MTD): Agents that continuously reconcile your accounting data, flag discrepancies before submission deadlines, and auto-generate compliant returns
• Anti-money laundering (AML): For businesses handling transactions, agents that monitor patterns in real-time rather than retrospective batch analysis
• Employment law compliance: Agents tracking working time directives, holiday accrual, and right-to-work documentation automatically
• GDPR and data governance: Continuous monitoring of data handling practices, consent management, and breach detection

The Cost Advantage

Goldman Sachs isn't doing this for innovation points. They're doing it because compliance is expensive. The average mid-market company spends 5-8% of revenue on regulatory compliance. AI agents can reduce that by 40-60% while improving accuracy.

For a UK company turning £5 million, that's potentially £150,000 to £240,000 annually in savings — more than enough to fund the AI implementation itself.

Speed of Detection

Traditional compliance is reactive. Something goes wrong, someone notices (eventually), an investigation happens, and a fix is applied. AI agents flip this to proactive:

• Continuous monitoring instead of periodic audits
• Real-time flagging instead of quarterly reviews
• Automated remediation instead of manual correction cycles

One UK accounting firm reported reducing their audit preparation time from 6 weeks to 4 days after deploying AI agents for pre-audit reconciliation.

How UK SMEs Can Start Today

You don't need Goldman's budget. Here's a practical roadmap:

Phase 1: Automate the Obvious (Month 1-2)

Start with the finance functions that are high-volume, rules-based, and error-prone:

• Invoice processing: AI agents that read invoices, match them to purchase orders, flag discrepancies, and queue payments
• Expense management: Automatic categorisation, policy compliance checking, and VAT reclaim identification
• Bank reconciliation: Daily automated matching with intelligent handling of partial matches and timing differences

Tools: Xero's AI features, Dext, ApprovalMax, or custom agents via platforms like n8n or Make

Phase 2: Compliance Monitoring (Month 3-4)

Layer in compliance-specific agents:

• Contract review: Agents that scan new contracts for non-standard terms, missing clauses, and regulatory requirements
• Regulatory change tracking: Automated monitoring of FCA, HMRC, and sector-specific regulatory updates with impact assessments
• Employee compliance: Right-to-work checks, training certification tracking, and policy acknowledgement monitoring

Phase 3: Strategic Finance (Month 5-6)

Graduate to higher-value applications:

• Cash flow forecasting: Agents that combine historical patterns, current pipeline data, and market signals to predict cash positions
• Tax planning: Continuous optimisation of tax positions across corporation tax, VAT, and personal tax for directors
• Financial reporting: Automated management accounts with narrative commentary and variance analysis

The Guardrails Question

The obvious concern: do you want AI making financial decisions without human oversight?

The answer isn't binary. Goldman's approach uses a "graduated autonomy" model:

• Level 1 (Fully autonomous): Routine transactions under £10,000, standard reconciliations, report generation
• Level 2 (Autonomous with notification): Unusual patterns, larger transactions, compliance flags — the agent acts but notifies a human
• Level 3 (Human approval required): New counterparties, regulatory submissions, anything above defined thresholds

This graduated approach works for SMEs too. Start with Level 1 for low-risk tasks, and gradually expand the autonomy envelope as trust builds.

What This Means for Finance Teams

Let's be direct: AI agents will reduce the number of people needed for routine finance and compliance work. A team of 5 doing bookkeeping, reconciliation, and basic compliance could become 2 people plus AI agents.

But those 2 remaining roles become significantly more valuable. They're:

• Overseeing agent performance and tuning parameters
• Handling genuinely complex cases that require judgement
• Providing strategic analysis rather than data processing
• Managing relationships with auditors, regulators, and advisors

The UK businesses that handle this transition well will invest in upskilling their finance teams, not just deploying technology. BNY's mandatory 10-hour AI training for all staff is the right model.

The Bottom Line

When Goldman Sachs and BNY deploy AI agents for their most regulated, highest-stakes functions — that's a signal. If autonomous agents are trusted enough for Wall Street compliance, they're certainly capable enough for your invoice processing and MTD submissions.

The question isn't whether AI agents will transform business finance. It's whether you'll be the one deploying them or the one competing against businesses that already have.

For UK businesses, the regulatory complexity that feels like a burden is actually an opportunity. The more complex the compliance landscape, the greater the advantage AI agents provide. Companies that move now will have compounding efficiency gains by the time their competitors start looking into it.

Start small, build trust, expand gradually. The Goldman Sachs playbook works at every scale.

---

Caversham Digital helps UK businesses implement AI automation that delivers measurable results. Get in touch to discuss how AI agents could transform your finance and compliance operations.