Goldman, JPMorgan lead banking’s shift to autonomous AI agents

By early 2026, Goldman Sachs had embarked on a transformative journey embedding autonomous AI agents into its core workflows through the OneGS 3.0 program, targeting six major processes such as accounting, trade reconciliation, and client onboarding. This strategic integration, developed in close collaboration with Anthropic engineers using the Claude model, aims not only to boost employee productivity and operational efficiency but also to free capital for investment in growth areas while maintaining strong financial returns. Industry leaders like Ben Horowitz underscore that such pioneering efforts mark just the beginning of enterprise AI adoption, signaling foundational shifts in banking operations driven by autonomous AI agents.

Goldman Sachs and other Wall Street giants like JPMorgan Chase and Morgan Stanley are redefining banking execution models by transitioning from AI as mere decision-support tools to autonomous agents capable of executing complex tasks and managing outcomes within defined objectives. JPMorgan’s deployment of long-running AI agents that screen market activity overnight has already contributed to a reported 20% increase in gross sales, illustrating a shift from cost-cutting to sustainable competitive advantage through AI-driven revenue growth. This evolution is further exemplified by banks treating AI agents as digital coworkers with assigned logins, managers, and performance reviews, fostering hybrid human-agent collaboration that enhances productivity and operational resilience.

Flagship AI initiatives such as Goldman Sachs’ OneGS 3.0 and GS AI Assistant, alongside Morgan Stanley’s digital assistants and UBS’s AI-driven advisor alerts, demonstrate a clear strategic focus on automating routine tasks while enhancing client engagement and revenue generation. These programs enable banks to shift towards higher-fee, data-driven business models, reducing reliance on balance sheet-intensive activities and unlocking new growth opportunities. Despite near-term elevated expenses due to AI investments, Goldman Sachs expresses strong confidence in achieving meaningful productivity gains and improved operating leverage, aligning with investor demands for clear ROI on technology spending.

While accelerating AI adoption, leading banks maintain cautious governance frameworks to mitigate regulatory and risk concerns, ensuring human oversight remains integral, especially in critical decision-making areas like portfolio management and compliance. This balanced approach, highlighted by Bain & Company’s Bhavi Mehta and KPMG’s Peter Torrente, involves embedding guardrails and defining hybrid roles where AI agents augment rather than replace human employees. Such prudence reflects the industry’s recognition that autonomous AI agents, though powerful, must operate within controlled environments to sustain trust and operational integrity.

By early 2026, Goldman Sachs had forged a strategic partnership with AI startup Anthropic, embedding their engineers onsite for six months to co-develop autonomous agents powered by the Claude model. This hands-on collaboration exemplifies how leading banks are integrating agentic AI into legacy systems to automate complex functions like accounting, trade reconciliation, and client onboarding, positioning generative AI as a cornerstone of multi-year strategies aimed at boosting operational efficiency and enhancing client experiences while reducing third-party dependencies.

Financial institutions increasingly rely on partnerships with cloud hyperscalers such as Google, Amazon, and Microsoft Azure to deploy scalable AI solutions that can gracefully interface with entrenched legacy infrastructures. As one industry expert explained, success hinges on a 'crawl, walk, run' maturity model that proves AI’s value incrementally, while cloud providers like Google emphasize a flexible approach that brings AI to the data—regardless of whether it resides in AWS, Azure, or elsewhere—thereby enabling seamless integration across heterogeneous environments.

Google’s DeepMind and its multimodal AI engine Gemini represent a technological leap tailored to the financial sector’s massive data complexity and agentic AI use cases, processing millions of tokens with advanced reasoning capabilities. This innovation underscores the critical role of specialized AI architectures designed to meet the demanding scale and nuance of banking workflows, reinforcing why financial institutions are increasingly leaning on such cloud-native AI platforms to navigate their digital transformation.

Backbase’s 2026 acquisition of Kasisto marks a pivotal moment in embedding agentic AI within banking platforms, combining conversational intelligence with robust governance, regulatory controls, and domain-specific knowledge. This move reflects a broader industry trend toward delivering end-to-end automation that transcends simple chatbots, emphasizing auditable decision logs, role-based approvals, and compliance-ready AI workflows that integrate tightly with core banking systems—key factors that will determine early adoption success as banks pilot these unified, resolution-focused AI solutions.

The deployment of autonomous agentic AI in banking operations introduces profound governance challenges centered on accountability, oversight, and liability. As these AI agents independently execute multistep workflows—ranging from fraud detection to incident resolution—banks must carefully delineate the scope of AI responsibilities to balance efficiency gains with stringent control mechanisms. With 46% of financial institutions reporting increasingly sophisticated fraud schemes and regulatory pressures intensifying, firms like Primitive emphasize robust governance frameworks that integrate continuous measurement and oversight to mitigate operational and compliance risks.

Effective governance in regulated financial environments now hinges on a dynamic two-loop model where AI agents and human experts engage in real-time collaboration to refine decision logic continuously. This interactive oversight is crucial as generative AI outputs—while often persuasive and professionally articulated—can contain factual inaccuracies or fabrications, complicating traditional compliance paradigms. By mid-2026, regulators such as the OCC and Federal Reserve acknowledged that existing model risk management frameworks do not fully encompass generative AI, underscoring the need for novel governance approaches that address the distributed and non-linear accountability emerging from AI-human decision interplay.

Banks adopting autonomous AI agents to accelerate incident resolution have achieved remarkable efficiency improvements, with resolution times dropping from over two hours to under 30 minutes and nearly 40% of incidents managed without human intervention. However, this operational leap intensifies the imperative for transparency, explainability, and reversibility of AI actions to satisfy regulatory scrutiny and maintain institutional liability. Leading practices involve layered AI agent models where compliance checks precede execution, ensuring that despite automation, ultimate accountability remains firmly with human governance structures, as highlighted by Matthew Cheung of ipushpull who asserts that 'governance will be the key to who wins in a world of messaging and AI agents.'

Goldman Sachs has taken a decisive step to bridge the AI talent gap by appointing Evan Kotsovinos, a seasoned AI leader from Google with a background in privacy, safety, and security, as partner and head of engineering for its $3.65 trillion asset and wealth management division. This role, which reports directly to COO Will Bousquette and CIO Marco Argenti, underscores the firm’s commitment to embedding advanced AI technologies deeply into investment performance, client experience, risk management, and operational efficiency strategies.

Beyond strategic hires, Goldman Sachs is investing in internal capability-building through initiatives like the CEO AI Academy, designed to prepare its leadership and teams to manage AI integration responsibly and effectively. This dual approach—combining external recruitment of top AI talent from tech giants like Google with comprehensive leadership development programs—reflects a holistic strategy to ensure that AI adoption is both innovative and aligned with the firm’s broader operational goals.

By mid-2026, Goldman Sachs has positioned AI and data at the core of its private equity value creation strategy, integrating advanced analytics and engineering capabilities to drive operational improvements and EBITDA growth. Initiatives like the CEO AI Academy and the firm's Engineering division exemplify this commitment, enabling portfolio companies to undergo rigorous diagnostics of data hygiene and technology infrastructure, while also upgrading management talent to future-proof investments. This approach leverages extensive networks and operating partners to scale value creation resources across multiple funds, reflecting Michael Bruun's emphasis on activating networks to outperform in complex geopolitical and market environments.

Simultaneously, Goldman Sachs and JPMorgan Chase have capitalized on the global AI boom to transform their banking revenue models, achieving record revenues driven by equities trading and investment banking activities linked to AI-related deals. This shift extends beyond external advisory and trading fees, as banks integrate AI internally to enhance operational resilience and efficiency. Goldman Sachs, in particular, is pivoting towards higher-fee, data-driven businesses, reducing reliance on balance sheet-intensive activities, exemplified by its acquisition of Industry Ventures and a $1.5 billion partnership with Anthropic to accelerate AI adoption across portfolio companies.

The evolution of private equity market structures through innovations like evergreen funds and continuation vehicles further complements AI-driven strategies by offering greater flexibility and long-term capital deployment options. These new structures, highlighted by Michael Bruun, enable firms to adapt to shifting market conditions while leveraging AI-enhanced valuation, risk assessment, and portfolio construction techniques. Together, these developments underscore a broader transformation in how banks and private equity firms harness AI not only to optimize operations but also to redefine business models and capture emerging market opportunities.

Internally, Goldman Sachs is deploying AI tools such as the GS AI Assistant to streamline workflows, reduce manual workloads, and boost employee productivity, allowing professionals to focus more on client engagement and strategic decision-making. This operational enhancement aligns with the firm's broader AI strategy to unlock new growth avenues by simplifying complex processes and fostering a culture of innovation amid a rapidly evolving technological landscape.

By 2026, the banking industry is undergoing a profound transformation as autonomous AI agents evolve from mere decision-support tools into fully autonomous execution engines that manage outcomes and continuously refine their performance. This agentic shift, described as moving from AI that supports decisions to AI that executes them, is not a superficial technology upgrade but a fundamental restructuring of how banking work is organized and delivered, driving a multi-year investment cycle focused on responsible, scalable adoption. Banks embedding these AI agents as governed digital labor into their core operations are poised to realize structural advantages including revenue growth, cost optimization, enhanced operational resilience, accelerated decision-making, and improved institutional knowledge retention.

Leading institutions like JPMorgan Chase are spearheading the deployment of long-running autonomous AI agents capable of managing complex workflows across multiple systems, signaling that security and governance challenges are nearing resolution. Derek Waldron, JPMorgan's Chief Analytics Officer, highlights the transition to AI agents that can operate for hours or even weeks, enabling productivity gains that extend beyond back-office functions into revenue-generating areas such as private banking, where AI tools have driven a 20% increase in gross sales and could expand client coverage by up to 50%. This strategic pivot from cost-cutting to building sustainable competitive advantages through in-house AI capabilities also challenges traditional software vendors, reshaping the competitive landscape.

The adoption of autonomous AI agents is now a global phenomenon, with major financial institutions including Morgan Stanley, UBS, BNY Mellon, Goldman Sachs, and Citi actively integrating these agents into client interactions, trading, compliance, and treasury functions. According to a June 2026 KPMG survey, over half of banks were piloting AI agents, reflecting a multi-year investment cycle where AI agents are treated as digital employees with logins, performance reviews, and human managers overseeing training and quality control. This evolution fosters hybrid workforce models where humans and AI collaborate closely, enabling advisors at UBS to spend 70% of their time engaging clients while AI handles routine tasks, all under strict human oversight to mitigate risks in customer-facing and critical decision areas.

In Asian markets, the acceleration of agentic AI adoption is tightly coupled with rigorous data governance and regulatory oversight, as institutions anticipate 20–30% reductions in operating costs by 2028 and have already reported productivity improvements exceeding 50%. However, the primary bottleneck remains data readiness—banks must implement disciplined customer data management practices including continuous cleansing, verification, and enrichment across complex multilingual and cross-border environments to ensure AI agents operate on accurate, auditable data. Regulators like Singapore’s MAS and the Hong Kong Monetary Authority have elevated data quality to a supervisory priority, enforcing standards that prevent AI from making erroneous decisions that could directly impact customers, underscoring that AI readiness is as much about data stewardship as technology deployment.