KeyBank Focuses AI on Operations and Tech Over Client-Facing Chatbots

Ai4

KeyBank finds enterprise AI delivers maximum economic return when automating back-office document processing and employee workflows rather than deploying customer-facing conversational bots.

AI-assisted summary of When the Rubber Meets the Road: How the Consumer Bank is Realizing Value With KeyBank — original source. Check the original for full context.

Source published: 2026-08-19

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 19, 2026, this presentation features Keenan Moukarzel, Head of Consumer AI & Data at KeyBank. According to the video description, the session explores how established banks can generate real business value through AI by focusing on the operational challenges of people and process rather than purely technological solutions.

## TL;DR Keenan Moukarzel breaks down KeyBank's strategic framework for driving measurable value with AI across consumer banking, focusing primarily on technology and operations. He details three core execution pillars: unifying customer data for Next Best Action sales enablement, automating document-heavy workflows such as merchant statements and lending procedures, and supporting fast custom application prototyping. To scale impact safely, KeyBank implements a tiered delivery model covering broad employee copilots, vendor-embedded features, and prioritized custom builds under calibrated risk governance. The talk concludes with practical lessons on overcoming organizational resistance, maintaining tokenized data pipelines, and nurturing a dynamic project backlog.

## Atmosphere & style The presentation is refreshingly pragmatic, conversational, and energetic, cutting through abstract AI hype with grounded corporate realism. Moukarzel maintains an approachable, humorous tone—cracking jokes about charging audience members for slide photos and questioning the competitive value of finance spreadsheets. The session feels like an authentic field report from the banking trenches, balancing technical rigor and governance constraints with relatable anecdotes about grassroots innovation and frontline adoption.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=4J00q-G3DFo&t=4s) The core promises of retail banking come down to safeguarding funds, growing wealth, and facilitating transactions, with competitive differentiation increasingly driven by Technology and Operations rather than administrative functions. - [03:36](https://www.youtube.com/watch?v=4J00q-G3DFo&t=216s) KeyBank structures its operational and technical AI initiatives into three core pillars: Data and Business Intelligence, Document Management, and Fast Custom Solutions. - [03:36](https://www.youtube.com/watch?v=4J00q-G3DFo&t=216s) Banks hold massive proprietary transactional datasets, and generative AI accelerates achieving a true Customer 360 view to power Next Best Action recommendations for frontline bankers. - [03:36](https://www.youtube.com/watch?v=4J00q-G3DFo&t=216s) Lightweight, bespoke web-based visualizations built with AI can replace rigid, heavy Business Intelligence platforms like Tableau or Power BI with lower engineering overhead. - [07:16](https://www.youtube.com/watch?v=4J00q-G3DFo&t=436s) AI transforms heavy document workflows across banking by instantly analyzing competitor merchant fee statements, navigating complex wealth management trust documents, and surfacing branch standard operating procedures. - [09:02](https://www.youtube.com/watch?v=4J00q-G3DFo&t=542s) Internal AI coding tools enable business line employees to prototype their own functional applications rapidly, though formalizing production pathways for citizen-developed tools remains a challenge. - [10:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=625s) AI tooling is deployed in three tiers: enterprise-wide readily available general AI, vendor-embedded capabilities (such as Salesforce or Tableau Data Stories), and internally developed custom AI solutions. - [10:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=625s) General-purpose enterprise AI yields massive cumulative value through small incremental wins, scaling to 17,000 saved hours weekly across 17,000 employees if each saves just one hour per week. - [13:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=805s) Successful AI deployment requires a cross-functional 'village' comprising business units, operations, engineering, and risk partners rather than an isolated data science team. - [13:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=805s) Navigating executive buy-in across a 200+ idea backlog requires patience, as initial rejections frequently turn into approvals after industry readiness and internal awareness mature. - [15:48](https://www.youtube.com/watch?v=4J00q-G3DFo&t=948s) KeyBank protects customer privacy by ensuring raw PII like Social Security numbers never reaches AI modeling environments, utilizing tokenized and obfuscated data layers in the enterprise data lake. - [15:48](https://www.youtube.com/watch?v=4J00q-G3DFo&t=948s) Overcoming internal adoption friction relies on solving acute frontline pain points and delivering immediate utility rather than imposing top-down mandates.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4J00q-G3DFo&t=4s) Defining Banking Value and Core Functions

Keenan Moukarzel opens his presentation by introducing his background, noting that while he has worked in artificial intelligence for about eight years, he has spent the last four years at KeyBank—the past ten months of which have been dedicated specifically to directing the consumer bank's AI strategy and execution. Setting an informal and engaging tone, he jokes with the audience that while many industry events rely on high-level panel discussions, his goal is to deliver actionable content, quipping that he will be monitoring the room and charging a fifty-dollar fee for anyone taking photos of the slides. He explains that after days of deeply technical presentations, his focus is grounded squarely in the mechanics of generating tangible business value from AI within the banking sector, particularly in consumer banking.

To frame how AI can deliver economic impact, Moukarzel steps back to define the core value proposition of a bank. Stripped down to its fundamentals, a bank fulfills three primary promises for its customers:

- **Safeguarding Funds:** Providing a secure place for individuals to deposit and protect their money. - **Growing Wealth:** Offering mechanisms, interest, and investment products that help customer balances appreciate over time. - **Facilitating Transactions:** Providing seamless money movement services, such as transferring funds to family members or paying vendors.

---

[01:16](https://www.youtube.com/watch?v=4J00q-G3DFo&t=76s) Finding Where AI Truly Moves the Needle

Executing these three core functions requires a complex web of internal divisions across an enterprise, including customer-facing branch bankers, back-office operations (such as loan servicing), human resources, legal, risk management, finance, and technology. For a bank executive evaluating AI investment while competing against roughly 4,500 to 5,000 other financial institutions across the United States, determining where AI yields the highest return on investment is a critical strategic decision. Moukarzel engages the audience to poll which department holds the greatest potential value, hearing suggestions ranging from customer experience and operations to risk management.

While acknowledging that risk and customer experience are strong contenders, Moukarzel humorously dismisses finance as the primary differentiator, questioning whether generating slightly better spreadsheets will meaningfully alter a bank's competitive standing. Instead, he highlights the combination of **Technology** and **Operations** as the twin engines capable of driving sustainable competitive advantage. Technology teams can leverage AI to write and refactor code substantially faster, while operations teams can automate complex back-office workflows to scale productivity.

```mermaid flowchart TD accTitle: Core Pillars of AI Value in Consumer Banking accDescription: Three pillars of AI implementation showing Data and BI, Document Management, and Fast Custom Solutions.

subgraph BI["Data & Business Intelligence"] A["Customer 360 Acceleration"] B["Next Best Action for Sales"] C["Lean Web Visualizations"] end

subgraph DOCS["Document Management"] D["Competitor Statement Analysis"] E["Trust Document Navigation"] F["Lending & Procedure Search"] end

subgraph SOLUTIONS["Fast Custom Solutions"] G["Task Queue Automation"] H["Conversational & Copilot Tools"] I["Rapid Web App Prototyping"] end

BI --> Impact["Enterprise Value Creation"] DOCS --> Impact SOLUTIONS --> Impact ```

---

[03:36](https://www.youtube.com/watch?v=4J00q-G3DFo&t=216s) Pillar 1: Data, Customer 360, and Business Intelligence

KeyBank breaks down its operational and technical AI focus into three foundational pillars: **Data and Business Intelligence**, **Document Management**, and **Fast Custom Solutions**. Within the data pillar, banks maintain an unmatched advantage: they hold vast amounts of personal transaction history and behavioral data, often exceeding what any other commercial entity possesses.

The primary prize in this pillar is the elusive "Customer 360." When polling the room, Moukarzel observes only three audience members raise their hands to confirm their organizations have successfully achieved a true Customer 360 view. While AI is not strictly required to unify customer data, it acts as a massive accelerator. KeyBank uses AI to rapidly build data stores, analyze transaction views, and track significant fund flows—such as spotting when clients begin moving capital out of the institution.

Closely coupled with Customer 360 is the development of "Next Best Action" capabilities for branch and relationship bankers. Armed with deep behavioral context, front-line staff can initiate higher-quality, tailored sales conversations and proactively address specific client needs. Moukarzel notes that his manager, Ramesh, has been passionately driving this initiative across the bank for over a year. Additionally, AI streamlines internal data engineering pipelines and replaces traditional, rigid Business Intelligence platforms like Tableau or Power BI—which can be slow, difficult to maintain, and cumbersome to configure—with lightweight, bespoke, web-based visualizations that require far less engineering effort.

---

[07:16](https://www.youtube.com/watch?v=4J00q-G3DFo&t=436s) Pillar 2: Document Intelligence and Procedure Search

Because banking is one of the most heavily regulated industries in the United States, institutions are inundated with enormous volumes of structured and unstructured paperwork, from regulatory disclosures and bank statements to internal standard operating procedures. AI transforms this documentation burden into operational leverage across several distinct areas:

- **Key Merchant Services (KMS) & Competitor Analysis:** KeyBank invites prospective clients to submit merchant statements from competing banks. AI analyzes these complex fee schedules instantly, allowing the bank to generate a competitive, optimized pricing quote faster and win deals at the point of sale. - **Trust & Wealth Management Documentation:** Trust documents in wealth management are intricate legal instruments that frequently evolve over time. AI organizes and surfaces the exact governing documents and legal clauses needed, cutting through hours of manual legal search. - **Home Lending Processing:** Given the standardized yet document-heavy nature of mortgage origination, AI extracts and cross-checks data rapidly while keeping human underwriters firmly in the loop for final decision-making. - **Branch Procedure Search:** Branch bankers often encounter unfamiliar transaction requests, such as complex ACH transfers. Rather than forcing employees to sift through ten pages of irrelevant keyword search results, AI delivers precise, contextual procedural instructions instantaneously.

---

[09:02](https://www.youtube.com/watch?v=4J00q-G3DFo&t=542s) Pillar 3: Fast Custom Solutions and the Prototyping Surge

The third major operational pillar focuses on deploying bespoke software solutions rapidly to eliminate operational bottlenecks. A major upcoming area of focus is automating administrative task queues within operational processing centers to clear work backlogs faster.

When discussing conversational AI, Moukarzel notes that while internal contact centers have adopted it extensively, deploying fully autonomous customer-facing conversational agents remains tricky. Customers frequently prefer speaking directly with human bankers, and enterprise comfort around conversational risk continues to mature.

Meanwhile, internal AI-assisted coding has created a groundswell of grassroots innovation across business units. Employees are using AI to solve day-to-day workflow friction on their own. Moukarzel shares an anecdote about a national director in wealth management who used generative AI tools to code three to five functional prototype applications on his own and immediately asked the engineering team to push them straight to production. While Moukarzel chuckles that these bespoke tools were essentially built to solve problems for only three people and highlights the ongoing challenge of establishing formal production pipelines for citizen-developed tools, he emphasizes that empowering business teams to rapidly build their own functional solutions represents an exciting shift in enterprise productivity.

[10:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=625s) AI Enablement: Readily Available Tools, Vendor Solutions, and Governance

Beyond targeted priority projects, empowering the broader organization requires a tiered strategy for AI tooling. The goal is to provide everyday employees with tools that solve routine problems without requiring dedicated engineering teams with "all these eyes and hands on keyboard." The first tier is "readily available AI"—corporate-grade, secure equivalents of consumer tools like ChatGPT or Microsoft Copilot. The enterprise value of these broad deployments compounds rapidly through incremental productivity gains. Across KeyBank’s roughly 17,000 employees, saving just a single hour per employee per week scales to 17,000 hours saved weekly across the enterprise, creating immense operational efficiency.

The second tier leverages built-in AI capabilities from existing enterprise vendors, such as Salesforce or Tableau's automated Data Stories. Enabling AI within tools that staff already use every day lowers the barrier to adoption. When neither general-purpose AI nor out-of-the-box vendor capabilities can solve a specialized business challenge, the internal AI team steps in to build custom AI solutions. These custom projects are drawn from a actively managed backlog of business ideas and prioritized based on strategic impact.

Underpinning this tiered approach is a disciplined governance and risk framework. Different AI applications carry vastly different risk profiles, and low-risk applications may require different governance tracks than high-risk models. The priority is ensuring rigorous oversight rather than pursuing business expansion at all costs.

```mermaid flowchart TD accTitle: KeyBank AI Tooling and Delivery Tiers accDescr: Three tiers of AI delivery leading to governance and deployment.

subgraph Tiers["AI Delivery Tiers"] A["Readily Available AI<br/>(Copilot, Internal LLM Tools)"] B["Vendor-Embedded AI<br/>(Salesforce, Tableau Data Stories)"] C["Custom AI Development<br/>(Prioritized Line-of-Business Backlog)"] end

subgraph Oversight["Oversight & Review"] G["Risk & Governance Framework<br/>(Differentiated by Risk Tier)"] end

subgraph Production["Production"] P["Secure Business Value & Deployment"] end

A --> G B --> G C --> G G --> P ```

- **Enterprise Scale Math**: Saving 1 hour per week across 17,000 employees generates 17,000 saved hours per week enterprise-wide. - **Vendor Native Features**: Taking advantage of native features like Tableau Data Stories brings AI to users already comfortable in their existing software workflows. - **Risk-Calibrated Governance**: Ensuring risk oversight matches the complexity and exposure of each use case rather than applying a one-size-fits-all blocker.

---

[13:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=805s) Key Takeaways from 10 Months in Production: People, Ecosystems, and Backlogs

Reflecting on lessons gathered over ten months of scaling AI within KeyBank, successful execution relies on organizational alignment rather than pure technology. AI delivery "takes a village"—the myth of a single, isolated data science team delivering end-to-end transformation is unsustainable. Sustainable AI requires tight, day-to-day collaboration among business units, core technology teams, risk partners, and operational leaders.

Engaging openly with the broader vendor ecosystem is another vital asset. Rather than dismissing vendor representatives as mere salespeople, maintaining strong partner relationships exposes internal teams to emerging market frameworks and external innovations. Crucially, engineering must always be reverse-engineered from genuine business pain points. As the speaker highlights by referencing Elon Musk, spending resources engineering the wrong solution is wasted effort. Navigating executive buy-in also requires patience; out of a backlog containing roughly 200 ideas, receiving an initial "no" is routine. Many ideas need six months to simmer before broader industry awareness and leadership readiness turn them into a "yes."

- **Cross-Functional Village**: Successful deployment demands active integration across business leaders, operations, technology, and risk partners. - **Ecosystem Learning**: Vendor relationships serve as an intelligence pipeline for new industry architectures and best practices. - **Bidirectional Buy-In**: Solutions must secure high-level executive sponsorship alongside bottom-up grassroots support from end users. - **The Backlog Pipeline**: Managing a backlog of over 200 ideas means accepting that rejected proposals often resurface and gain traction as market conditions evolve.

---

[15:48](https://www.youtube.com/watch?v=4J00q-G3DFo&t=948s) Audience Q&A: Data Architecture, Tokenization, and Adoption Strategies

The session closes with an audience Q&A, where the speaker invites his manager, Umesh, and lead architect, Chris, to chime in on the technical and structural foundations supporting their work.

An audience member asks whether KeyBank runs AI models directly against raw customer records or relies purely on metadata and obfuscated data layers. The speaker explains that KeyBank enforces strict data supply chain safety: direct personal identifiers (such as Social Security numbers) never enter the modeling pipelines. The enterprise data lake contains tokenized and obfuscated customer information, ensuring that when developers pull data into development sandboxes to build prototypes and productionalize pipelines, customer privacy remains fully protected under standard enterprise analytics protocols.

```mermaid flowchart LR accTitle: KeyBank Secure AI Data Pipeline accDescr: Data flow from raw storage through tokenization, data lake, sandbox, and production.

A["Raw Core Banking Systems"] --> B["Tokenization & Obfuscation Layer"] B --> C["Enterprise Data Lake"] C --> D["Secure Sandbox Prototyping"] D --> E["Production AI Deployment"] ```

A second audience question addresses overcoming institutional resistance and employee fear to drive AI adoption. The speaker jokes that progress often comes through "sweat and blood," but emphasizes that genuine user adoption hinges on user-centric design. Rather than forcing top-down tools onto reluctant teams, the AI group focuses on directly addressing acute frontline pain points. Delivering tangible time savings and practical utility makes end users enthusiastic partners in adopting AI technologies.

## Notable quotes > When you think about effective management strategy for operational success, you think of People, Process and Technology. The latter (Technology) is not the problem in generating value in the AI field - it's mostly solved. It's the People and Process part that is challenging in an already existing business. — Keenan Moukarzel

> It's not always the most complex things that can give your business quick wins. — Keenan Moukarzel

> [10:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=625s) all these eyes and hands on keyboard — Keenan Moukarzel

> [13:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=805s) takes a village — Keenan Moukarzel

> [03:36](https://www.youtube.com/watch?v=4J00q-G3DFo&t=216s) Customer 360 — Keenan Moukarzel

> [10:25](https://www.youtube.com/watch?v=4J00q-G3DFo&t=625s) readily available AI — Keenan Moukarzel

> [15:48](https://www.youtube.com/watch?v=4J00q-G3DFo&t=948s) sweat and blood — Keenan Moukarzel

> [03:36](https://www.youtube.com/watch?v=4J00q-G3DFo&t=216s) Next Best Action — Keenan Moukarzel

## Mentioned Keenan Moukarzel, KeyBank, Ai4, Ramesh, Umesh, Chris, Elon Musk, Key Merchant Services (KMS), ChatGPT, Microsoft Copilot, Salesforce, Tableau, Power BI, Tableau Data Stories

## Video description Keenan Moukarzel - Head of Consumer AI & Data, KeyBank

When you think about effective management strategy for operational success, you think of People, Process and Technology. The latter (Technology) is not the problem in generating value in the AI field - it's mostly solved. It's the People and Process part that is challenging in an already existing business. In this talk, we will discuss how to think about generating value for the business in the banking world, how to effectively identify AI opportunities that don't disrupt the business but add value. It's not always the most complex things that can give your business quick wins.

[Source](https://www.youtube.com/watch?v=4J00q-G3DFo)