STORIES OF IMPACT / Articles by Endurance / My AI Journey

My AI Journey by Stephen Curry

Summary

In 2025, my journey as a banking executive plunged me into the heart of AI’s transformative power, a force reshaping our industry with breathtaking speed. Through candid conversations with peers, dynamic forums hosted by Deloitte and Grant Thornton, and hands-on explorations, I witnessed AI evolve from a distant promise to a core driver of banking operations. Agentic AI, capable of independently orchestrating data, processes, and models, captivated me with its ability to revolutionize customer interactions, streamline back-office workflows, and sharpen risk management. Real-world examples, like a major bank automating information flows to trading and risk management and streamlining analysts' workflow and securities filings, or Bank of America’s Erica delivering hyper-personalized financial guidance, brought this potential to life. Tools like Tabnine and Github CoPoliot slashed development timelines, while Deloitte’s Zora AI transformed financial reporting with real-time dashboards. Yet, challenges like data silos, AI biases, and workforce apprehension loomed large, underscoring the need for robust data governance and precise prompts—reminiscent of my high school days crafting Microsoft BASIC commands to unlock meaningful outcomes.

Looking to 2026, I envision AI propelling banking into a new era of hyper-personalized customer experiences, operational efficiency, and real-time risk management. Imagine clients receiving tailored financial advice or seamless cross-selling offers, boosting loyalty and revenue, while automated compliance and reporting free teams for strategic innovation. Success stories inspire this vision: a major investment banks’ AI-driven platforms cut costs and automated 40% of analyst tasks, while JPMorgan’s Contract Intelligence system saving thousands of hours on contract reviews. However, success hinges on clean, high-quality data and tools to link prompts to reliable information. Workforce reskilling and AI education, as seen in SMU’s forward-thinking curriculum, will prepare employees and future leaders to thrive. By balancing innovation with ethical safeguards, banks can harness AI to redefine customer engagement, streamline operations, and chart a bold, sustainable future.

My Journey into AI Transformation in Banking: 2025 Insights and 2026 Opportunities

As a banking executive in 2025, I’ve been swept into a whirlwind of exploration about artificial intelligence (AI), the most transformative force in a generation, which is reshaping our industry at breakneck speed. Over recent months, I’ve engaged in candid one-on-one conversations with fellow banking leaders, attended dynamic forums led by Deloitte and Grant Thornton, and delved into hands-on discussions about implementing AI. Banks embracing AI are reaping remarkable efficiency gains, uncovering deeper operational insights, and elevating customer experiences—promises that felt distant just a year ago. Here’s my personal story of discovery, interwoven with real-world examples, technical insights, challenges, and a vision for what lies ahead in 2026. Even though I was an early adopter personally, the progress has far exceeded my expectations.

A Deep Dive into AI’s Current Role

My journey kicked off with private discussions with industry peers who vividly illustrated AI’s central role in banking today. They highlighted that AI has moved beyond a visionary concept to a core operational driver, with roughly half of financial institutions already using it in 2023 and estimates suggesting nearly three-quarters will by 2026. What captivated me was the emergence of agentic AI—intelligent systems that operate independently, adapting to fresh data without constant human guidance. These AI agents act as orchestrators, pulling together data, processes, and AI models. For instance, in autonomous vehicles, an AI agent processes sensor inputs (data), applies navigation rules (process), and uses deep learning (AI model) to make real-time driving choices. This adaptability is revolutionizing banking, from customer interactions, back-office workflows, risk management functions, trading and even research.

Agentic AI differs from traditional AI by its ability to autonomously pursue goals, make decisions, and take actions in complex, dynamic environments without constant human intervention. While traditional AI typically focuses on specific tasks—like pattern recognition, data analysis, or predictive modeling—within predefined parameters, agentic AI leverages advanced reasoning, planning, and adaptability to achieve broader objectives. It can interpret high-level goals, break them into actionable steps, and interact with external systems or environments to execute plans, often learning and adjusting in real-time. For example, traditional AI might analyze customer data to recommend products, whereas agentic AI could independently manage an entire customer service workflow, negotiating, scheduling, and resolving issues. This autonomy, enabled by large language models, reinforcement learning, and contextual awareness, makes agentic AI more proactive and versatile, aligning closely with human-like problem-solving.

Another eye-opener was AI’s influence on software development. Tools like GitHub Copilot and Tabnine auto-generate code, troubleshoot errors, and propose optimizations, with industry estimates suggesting up to 40% reductions in development timelines. At a Deloitte forum, I witnessed how generative AI, agentic AI, cloud computing, and natural language processing (NLP) empower banks to analyze unstructured data, personalize offerings, and automate intricate tasks at scale. Presenters shared practical examples, from fraud detection to automated trading. Grant Thornton’s sessions deepened this perspective, underscoring AI’s power to boost efficiency and innovation while cautioning about risks if not carefully managed. These forums grounded my understanding in tangible applications, fueling my curiosity for what’s next.

Opportunities Await in 2026

Through these conversations, I began envisioning AI’s potential for 2026. Industry leaders painted a picture of hyper-personalized customer experiences driven by agentic AI. Picture a client receiving customized financial advice, instant portfolio tweaks, and seamless cross-selling offers that feel natural, not forced—building loyalty and boosting revenue. Operational efficiency also loomed large: automating routine tasks like compliance reviews and data processing could free our teams for higher-value strategic work. Risk management stood out as well, with AI systems monitoring market shifts and credit risks in real-time to keep us ahead of the curve. Financial reporting could become a strength, with AI replacing cumbersome manual processes with dynamic, real-time dashboards. Regulatory compliance, often a burden, could be streamlined, cutting costs and ensuring precision.

These possibilities rely on agentic AI’s capacity to function as a virtual workforce, independently tackling complex tasks. But a critical insight emerged: none of this works without robust, high-quality data.

Success Stories That Inspired

Practical discussions brought these ideas to life through compelling success stories. A Grant Thornton-hosted breakfast with a major Investment Bank's CIO was a revelation. One leader described their AI-powered trading and risk management platforms, which analyze market data in real-time to predict trends and support trade execution with precision. This system also processes news, earnings reports, and market signals, automates trading decisions, enhancing returns and populates risk management dashboards. The firm has also automated nearly 40% of tasks traditionally handled by analysts in investment banking, trading, and research, including data collection, financial modeling, client presentations and due diligence. This has reduced operational costs and improves talent utilization.

Their automation of securities filings was equally striking. AI drafts IPO prospectuses and regulatory documents, cutting preparation time and dropping per-filing costs from $10,000 to under $500. Tools similar to JPMorgan’s Contract Intelligence (COiN) scan filings for risks and errors, reducing legal costs. These tools can save millions annually by reducing reliance on external legal support. These examples resonated, showing how banks can adopt similar tools to streamline credit agreements, compliance, and deal preparation.

Deloitte’s Zora AI, highlighted at their forum, was another standout. Built on NVIDIA’s platform, Zora automates expense tracking, procurement, and financial reporting. Deloitte’s finance team uses it to oversee payroll and marketing expenses, achieving a 25% cost reduction and a 40% productivity surge. Zora’s real-time monitoring of project costs and progress, flagging delays or overruns instantly, could revolutionize how we manage technology and operations projects. Its cloud-based integration with existing systems sparked ideas for modernizing bank infrastructure.

Hewlett Packard Enterprise (HPE), partnering with Deloitte, showcased the future of financial reporting. Their Zora AI deployment on HPE Private Cloud AI transforms static finance reviews into live digital dashboards, minimizing manual work, boosting accuracy, and accelerating decisions. In a hands-on session, we explored how similar tools could reduce banks reporting costs and empower leadership with sharper and real time insights. These stories weren’t just motivating—they offered a roadmap for our own transformations. Bank of America’s Erica brought agentic AI to life. In a one-on-one, I learned how Erica delivers tailored financial guidance (and tried it myself!). For each customer, Erica can analyze spending patterns, savings goals, and credit usage to create a personalized budget, alerting them to overspending and recommending investments based on her risk tolerance. In 2024, Erica managed over 2 billion interactions since 2018, including 676 million in 2024, with over 98% of clients getting the answers they need within 44 seconds. proving hyper-personalization drives engagement and revenue. This inspired ideas for developing AI assistants to strengthen customer connections.

Data and the Art of the Prompt

A recurring lesson was the critical role of robust data. At a Grant Thornton session, a major investment bank executive stressed that clean, accurate, and well-organized data is AI’s foundation. Poor data— incomplete, biased, or fragmented—can lead to flawed loan decisions or trading errors. A Deloitte FinanceAI event checklist reinforced data’s primacy, urging banks to assess quality and integrate sources for AI. My own experience at Bank of America, building a large-scale data warehouse and sponsoring data science uses in the 1990s echoed this, as data gaps often derailed insights. As we learned, executive management must prioritize and resource data governance efforts, including data remediation. Fortunately, this is becoming more feasible with tools that can help cleanse and integrate data to maximize AI’s potential. Banks are also tapping data brokers and synthetic data (artificially generated data that mimics real-world data using algorithms and simulations) to enrich their data pools, improving decision-making, customer experiences, and market positioning.

A real-world example of AI data collection and management strategies in banking is JPMorgan Chase’s implementation of its COiN (Contract Intelligence) platform, detailed in a case study published by Superior Data Science. To support this AI system, JPMorgan sources data from internal loan documents, external legal databases, and third-party financial APIs to create a comprehensive dataset of contract terms and market conditions. They employ automated data pipelines using cloud-based ETL tools (e.g., AWS Glue) to ingest, clean, and preprocess millions of documents in near real-time, ensuring scalability. Data quality is maintained through anomaly detection and schema validation, catching inconsistencies in contract language or missing clauses. To enrich the data, metadata like borrower credit profiles and macroeconomic indicators are integrated, enhancing the AI’s ability to predict default risks. They them collaborate with partner institutions to train models on decentralized, privacy-sensitive data without centralizing it. The bank uses a data lake (e.g., Snowflake) for efficient storage and retrieval of structured and unstructured data, while fairness-aware algorithms and expert reviews mitigate bias in loan term interpretations. Data versioning ensures reproducibility, and feedback from loan officers refines the dataset, aligning the system with evolving business needs, ultimately reducing processing time from thousands of hours to seconds while improving accuracy and compliance.

The concept of “the prompt” was another revelation, reminiscent of my high school experience crafting BASIC commands. A prompt is a big step forward from coding though - it is the natural-language instruction, like asking Erica, “Review Sarah’s transactions and suggest a savings plan.” Its clarity and precision are vital—vague prompts produce off-target results. Just as learning Microsoft BASIC in high school required precise commands to achieve results, modern AI relies on tools like vector databases, like Pinecone, which act as digital librarians and RAG frameworks, such as LangChain to connect prompts to data seamlessly. These technologies mirror the logic of BASIC—structuring inputs to unlock meaningful outputs—making AI both accessible and impactful for tasks like travel planning or workplace adoption. Strong repositories ensure real-time, reliable data, while well-designed prompts align AI with our goals. According to MBO Partners, generative artificial intelligence (Gen AI) has been the fastest-growing technology application ever. ChatGPT weekly has more than 100 million users, with Open AI’s Gen AI competitors (Anthropic, Google, Baidu, etc.) also reporting millions of users. In their studies, Gen AI has the potential to automate 60-70% of the average workers activities.

Challenges like data silos, where legacy systems fragment data, can complicate AI utilization and complicate real time analytics. Latency in processing large datasets can hinder performance, and inconsistent data erodes trust. Vendors are addressing these: Snowflake unifies data silos, Databricks ensures robust governance, Google Cloud’s Vertex AI supports prompt design, Microsoft Azure AI enhances scalability, and Palantir integrates data for financial institutions, including past work with JPMorgan. Oracle’s AI tools enhance enterprise automation and decision-making. These solutions are vital for connecting AI engines, prompts, and data, critical components for every bank’s AI strategy.

Cost Savings That Matter

The financial case for AI crystallized through my discussions. Legal document review can see up to 80% cost reductions with tools which extract terms and flag risks instantly, eliminating hours of manual labor. For a bank reviewing 10,000 documents yearly, this could save millions. Operations automation offers 20–30% savings, as seen in Deloitte’s Zora AI and Goldman’s analyst tools, potentially saving a mid-sized bank $5–10 million by automating half its back-office tasks. Financial reporting promises 15–25% savings with platforms like HPE’s Zora AI, which could save a large bank $2–5 million annually. Beyond savings, real-time insights—unimaginable a few years ago—are transforming and real time decision-making. These figures, shared in practical sessions, are shaping investment priorities.

Deploying AI Platforms

Forums and discussions revealed a range of AI platforms in action. Generative AI, like Bloomberg’s tools, analyzes news and social media for market insights, producing reports and summaries. Agentic AI, such as Deloitte’s Zora AI and Salesforce’s Agentforce, automates procurement, reviews, and project oversight. Machine learning platforms, like Goldman Sachs’, forecast trends and assess risks. Robotic process automation (RPA) streamlines compliance and operations, as seen in Goldman’s filing processes. Chatbots like Erica offer 24/7 support, resolving 98% of inquiries in seconds. These platforms, often paired with cloud systems like AWS or HPE Private Cloud AI, ensure scalability and real-time processing, a key focus in technical discussions.

Navigating the Risks

AI’s risks were a constant theme. Biased data can lead to unfair outcomes, like discriminatory loan decisions, a concern raised at Grant Thornton. Banks are countering this with bias detection, transparent data sources, and ethical frameworks like Deloitte’s Trustworthy AI™. Cybersecurity threats, such as AI-driven fraud or deepfakes, require security-by-design and continuous monitoring, as noted in a one-on-one.

AI hallucination—where models produce convincing but false outputs—poses some of the most serious risks, undermining trust in critical applications or even executing flawed trades. This can manifest as plausible but incorrect content, overconfidence, or amplified biases. Mitigation strategies include human-in-the-loop validation, where human review ensures accuracy, and prompt engineering to instruct AI to avoid speculation. Post-processing, like fact-checking against trusted databases and requiring AI responses to provide detailed reference citations, bolsters reliability. Regular model audits, through ongoing evaluations and real-world monitoring, help identify and correct hallucination patterns, enhancing AI trustworthiness. Regulatory and audit challenges arise when AI decisions lack transparency, so requiring detailed citations can be critical.

Another risk is unmanaged adoption. A recent KPMG survey found 44% of workers admitted to potentially unauthorized or inappropriate uses of AI, and 46% copped to uploading sensitive company information and intellectual property to public AI tools. KPMG concluded that this represented a significant gap in governance and raises serious concerns about transparency, ethical behavior and the accuracy of AI-generated content. In other words, employees are taking the AI transition into their own hands; companies delaying adoption are at risk of having unmanaged adoption and all the risks that go with it, including data insecurity, inadequate controls and more.

The biggest risk for most companies though is more direct - evolving too slowly to participate in the upside of AI. These tools are becoming critical for staying competitive in a digital, data-driven industry, and being adopted rapidly. Institutions slow to adapt risk falling behind entirely, surrendering domination to those who have already taken the lead.

Coming Soon to Financial Institutions

Over the next three years, AI will enable institutions to deliver hyper-personalized services that strengthen client relationships. By leveraging machine learning to analyze transaction data, behavioral patterns, and market trends, banks will be able to offer tailored financial products, such as customized loan terms, or automated mortgage applications, all in real time. Imagine deploying AI-driven chatbots, refined through generative AI, to handle complex client inquiries with the finesse of a seasoned banker. These tools could boost client satisfaction and retention, potentially increasing revenue.

AI will is already beginning to revolutionize underwriting by incorporating non-traditional data, such as payment histories or even client digital footprints, to create accurate credit models. This could expand your institutions lending reach while improving default predictions by 20-30%. For risk management, AI-driven stress-testing and portfolio optimization will enable precise scenario modeling, critical for regulatory compliance.

Then there is the impact in the investment community, which will be very sizeable. The evolution of predictive trading and portfolio management will result in AI applications that will replace fund managers, a transformation that is already underway. These algorithms analyze market sentiment and news in real time, optimizing portfolios with institutional-grade precision. For retail clients, AI platforms can democratize access to sophisticated strategies, enhancing market performance. Platforms like Truewind.ai can be utilized today to pilot trading algorithms.

Blockchain and digital assets will act as powerful catalysts for AI-driven transformation, amplifying efficiency, transparency and innovation. This will likely come in many forms, including some of the following – 1) migration to stablecoin payment systems to streamline and democratize an instant payments, 2) fully digital tokenized client profiles which will feed AI models with real-time, confidential but consented data, 3) smart contracts which will automate compliance tasks and streamline document management, and loan agreements stored immutably, and 4) tokenized securities transactions simplifying and accelerating the settlement process.

Where to Begin Your Journey

Now that you have witnessed the transformative impact AI is having elsewhere, it’s time to take the plunge. Begin with an engaging AI overview for your Board and C-suite, followed by a more in-depth program rolled out across the organization. Leadership should zero in on specific targets—like boosting customer experience, strengthening fraud prevention, refining risk management, or automating repetitive processes. Form dedicated teams to evaluate your data’s quality, accessibility, and security. Solid, well-organized data is the foundation of effective AI. Involve key players from different departments to align goals and address issues such as regulatory compliance, data protection, and ethical considerations. Kick off with small-scale pilot projects, led by those closest to the challenges you’re tackling. Experiment with AI solutions, assess their impact, and fine-tune before expanding. Build your team’s capabilities by recruiting AI and data science experts or upskilling existing staff with finance-relevant expertise. Continuously track AI performance, optimize workflows, and scale what works. Consider specialized AI tools for financial institutions, such as FinChat.io for investment insights, Truewind.ai for accounting tasks, or Datarails for financial planning. Stay vigilant about regulatory changes and ensure your AI initiatives adhere to industry standards to minimize legal risks.

By starting with these steps, your institution can confidently unlock AI’s potential.

Workforce Implementation

Embracing AI in the workplace can be filled with challenges that demand thoughtful strategies to foster enthusiasm and adoption. Employees often hesitate, wary of complex tools that feel disconnected from their daily tasks or fearful of job displacement. To overcome this, companies must create an environment where AI is seen as a partner, not a threat. Encouraging employees to experiment with AI in their personal lives can ignite excitement—take, for example, my recent experience using AI to plan a family Spring Break vacation. With careful prompting, I narrowed down lodging options, booked a perfect sailing charter, and discovered unique restaurants, all with remarkable ease. These personal wins can inspire employees to champion AI at work, showing colleagues how it amplifies their potential.

To build on this enthusiasm, companies should designate AI Champions at every level of the organization—enthusiastic advocates who guide and inspire their peers. Comprehensive, role-specific training is essential, with hands-on workshops and accessible resources empowering employees to use AI confidently, while Champions provide ongoing support. At Deloitte’s FinanceAI event, I witnessed banks like US Bank plan AI adoption by fostering super users and training programs.

Trust is equally critical; companies must transparently explain how AI functions, address ethical concerns like bias or privacy, and ensure secure access to tools. Starting with small, high-impact pilot projects allows firms to refine their approach, building confidence before scaling up. Leadership plays a vital role, actively modeling AI use and aligning it with broader business goals to demonstrate its value. Regular feedback loops and clear success metrics—such as time saved or improved outcomes—keep momentum alive. By blending personal experimentation, robust training, and transparent leadership, companies can transform AI adoption from a challenge into an inspiring, shared success, creating a workplace where employees and AI thrive together.

Futhermore, PwC suggests thinking about agentic workflow as a fundamental part of your workforce strategy. Agentic workflow will, for example, involve new management roles responsible for integrating digital workers into workforce strategies, then monitoring and governing them. With both digital and human workers on the job, for instance, you can plan for greater agility and shift resources more quickly to meet changing demands.

The Future Workforce & the College Experience

Workforce disruption, with estimates of 300 million jobs at risk globally, demands reskilling, but even more important is ensuring college graduates are ready for this evolution. This is even more immediate for me, as my daughter begins her master’s in accounting at SMU this summer. I was delighted to connect recently with Helmuth Ludwig, a Professor of Practice at SMU’s Cox School of Business. Over lunch, he shared how he infuses his strategy and entrepreneurship courses with a dynamic focus on AI, guiding students to explore its practical applications in sparking innovation, refining decisions, and driving business transformation. Drawing on his experience as Siemens’ former CIO, Ludwig challenges his students to weigh AI’s strategic promise against real-world hurdles like costs, ethics, and scalability. SMU’s commitment to AI, underscored by an $11.5 million NVIDIA partnership to integrate AI and STEM across disciplines, prepares students like my daughter to excel in a tech-driven world. Visionaries like Ludwig are equipping the next generation with the tools to lead. All of this acquires even greater importance to me as the father of a 12-year-old. For him, I need to focus on fostering skills, mindsets, and habits that will empower him to thrive in a world where AI is a partner, not a mystery. For all my children, the 21st century will present a lifetime of learning and continual adaptation.

A Vision for the Future

This journey has reshaped my perspective as a banking leader (and a father). AI isn’t just a tool—it’s a catalyst for reimagining our industry. Furthermore, every company that makes things or provides services—whether banking, lawnmowers or construction machinery—will be shifting from manual to autonomous or semi-autonomous systems in the next few years. In banking, we can deliver hyper-personalized experiences, optimize operations, strengthen risk management, revolutionize reporting, and streamline compliance, all while slashing costs. Success stories from Goldman Sachs, Grant Thornton, Deloitte, HPE, and Bank of America light the path, but robust data and precise prompts are the foundation. Vendors like Snowflake and Databricks are bridging critical gaps, enabling us to fully harness AI.

The workforce impact will be significant. New jobs will be created, some jobs will be lost, every job will be changed. AI will streamline routine tasks, reducing operational staffing needs by 20% or more and enabling broader management spans. This shift demands new skills—AI management, data engineering, and prompt design—that we must foster through reskilling. College students should prioritize courses in logic, data science, technical writing, statistics, and humanities, with disciplines like philosophy gaining new relevance for critical thinking and prompt design. Our employees, and children, will soon need to partner with AI, with more resources freed to take on strategic roles that fuel innovation.

Reflecting on my journey—from intimate leader sessions to vibrant forums and practical discussions—I’m invigorated by the possibilities. By balancing innovation with responsibility, we can steer our banks toward a future where AI drives sustainable growth, empowers our people, and redefines what’s possible. To paraphrase Elon Musk, the “future is going to start looking like the future”.