Overview

Senior transformation leaders discussed why AI pilots in finance (and adjacent enterprise functions) often look successful in a controlled environment but break when pushed beyond their original scope. One core theme prevailed in that scaling AI is less about “the model” and more about the operating system underneath it: data, process design, governance, and adoption. Therefore, success hinges not on the technology itself but on the factors around it. 

This roundtable was held on January 15th, 2026.

Roundtable Participants

Led by Puneet Thakkar, Google - Finance Process and Systems Transformation

• Adriana Garay, IBM - Executive Director

• Ani Ziyalyan, First American - Vice President, Transformation & Organizational Change Management

• Brian Hricik, Sherwin-Williams - Change Management Lead

• Kevin Chanashing, Bank of America - Executive SVP, Transformation Strategy

• Christy Didion, Crawford & Co - Organization Change Management and Transformation Lead

• Chuck Podgurski, Hertz - Global Implementation Change Management Lead

• David Cuenca, IBM - Vice President of Process Transformation, CIO

• David Stein, VML - SVP, Performance Marketing Operations

• Emeline Forbes, AstraZeneca - Change Management Lead

• Grégoire de Chevron Villette, RONA - Former Head of Transformation

• Jackie Cazar, Moody's  - SVP Process Excellence

• Jesse Taylor, Border States - Director of Transformation

• Jessica Cormier, BPM - Transformation Strategist

• Justin Marchi, City National Bank - Head of Enterprise Change Management

• Kaili Eperjesi, TriNet - Divisional Vice President, Enterprise Transformation

• Kathy Rokni, Netflix  - Senior Director, Globalization

• Madhuri Kumar, OXEA - Global Head of Talent

• Matt Twitchell, Stryker - Sr Director, Manufacturing Operations & Business Excellence 

• Miena Joseph, Ryan - Senior Director, Transformation & Strategy

• Rachel Pavsner, Senior Communications Advisor, Employee Engagement Strategist

• Sara Hyde, PenFed Credit Union - VP Finance Transformation

• Sara Jetta, Purolator Inc. - Vice President of Strategic Enablement

  
  

Part I: Why AI pilots succeed… then collapse at scale

The “pilot vs. reality” gap

Participants described a common pattern: pilots are designed to prove feasibility, but they often don’t include the very conditions that determine whether something can scale (exceptions, edge cases, localization, regulatory variance, and messy data).

“Pilots are often built like a straight road where all you are doing is going from point A to point B, like a highway with no exits. But when you factor in real-world complexity (like border crossings between countries, tolls, several exits), what used to be a point A to B journey is no longer that. That’s one of the main reasons pilots fail.”

Jackie Cazar, Moody's  - SVP Process Excellence

Fragmentation creates “local wins” that don’t compound

AI experimentation and early success stories often happen team-by-team or function-by-function. Even when leadership is excited, pilots can become disconnected efforts with different goals, different assumptions, and no shared definition of success, making enterprise rollout feel like stitching together unrelated prototypes. 

Scaling breaks where the enterprise is most inconsistent

Throughout the discussion, several breaking points came up:

• Inconsistent definitions and hierarchies across regions/business units

• Local market requirements that weren’t designed into the global workflow

• Model performance that varies by language/market context

• Underestimated cost and complexity when moving from one environment to many more

Part II: Tool-first vs. problem-first 

The “AI for AI” trap

A key tension surfaced: organizations are often pulled toward AI because it’s timely and ‘a new shiny object’, then they reverse-engineer a use case to justify its usage. That tends to produce impressive demos that don’t map cleanly to measurable business outcomes.

Problem-first reframes the work

Participants advocated for a different starting point:

• Define the business problem clearly (and confirm it’s the real problem, not a symptom)

• Decide whether AI is the best lever (vs. elimination, simplification, or standard automation)

• Only then design pilots and rollout plans that reflect real operating conditions

Part III: Process design — avoiding the “faster mess”

AI doesn’t fix broken workflows; it accelerates them

Participants aligned on a practical reality: if the underlying workflow is outdated, inconsistent, or full of legacy steps, AI will speed up the chaos rather than solve for it.

“A lot of organizations have a lift and shift approach to transformation. Let’s take this workflow as is and transfer it onto a new platform or a new way of doing things. But what happens when 30 to 40% of that workflow is just legacy habits? When you’re automating a bad process, you end up institutionalizing inefficiency. That’s why, before making the shift, you want to re-engineer the process.”

Puneet Thakkar, Google - Finance Process and Systems Transformation

Two approaches emerged (and they’re not mutually exclusive)

1) Re-engineer before you automate

This approach focuses on:

• Mapping the current state honestly

• Removing legacy habits and redundant steps

• Aligning stakeholders on a future-state reference model

• Pressure-testing the end-to-end flow with cross-functional teams before implementation

2) Reimagine the operating model (not just the steps)

For certain workflows, AI changes how the work should be done, so redesigning from the old process can be limiting. Instead, leaders advocated rethinking:

• Roles and decision rights

• Where “human-in-the-loop” belongs (and why)

• What should be standardized vs. where flexibility is required

A practical design principle: build for exceptions

A repeated critique of pilots: pilots validate the happy path but don’t prove how exceptions will be handled. The moment real-world complexity appears (special cases, regulatory differences), the workflow breaks.

Part IV: Governance as the missing layer between pilots and scale

Central coordination prevents chaos

Leaders described the need for a portfolio-level mechanism to:

• Review and prioritize use cases

• Prevent duplication across teams

• Align pilots to enterprise outcomes and guardrails

• Decide what scales globally vs. what stays local

Ownership keeps the model relevant as the goalposts move

Participants highlighted that transformation roadmaps and reference models can become stale quickly as business conditions change. A clear owner for process and documentation updates helps prevent “we designed it, then reality changed” failures. 

Part V: Defining success when the future keeps shifting

Success isn’t one metric

Leaders discussed multiple success lenses that matter at senior levels:

• ROI/value creation (including whether the initiative should exist at all)

• Cycle time and throughput improvements

• Risk reduction and control improvements

• Adoption and sustained usage

• Readiness and maturity (as leading indicators of where to start)

Sometimes the best move is simplification, not AI

“Everyone wants to use AI now, like everyone wanted to use agile a few years back. Neither is the answer for everything and, before using AI to improve a workflow, it’s important to consider exactly what the ROI will be. Maybe AI isn’t the answer; maybe the answer is pure elimination and simplification. The answer might be: let’s save money by just no longer doing this.” 

David Cuenca, IBM - Vice President of Process Transformation, CIO

Part VI: Data harmonization determines outcomes

A single source of truth as the scaling constraint

Participants repeatedly returned to data as the core limiter. If definitions, hierarchies, and master data vary across systems or regions, AI will amplify inconsistency rather than solve for it.

Governance before intelligence

Key practices discussed:

• Treat data as a business asset with clear accountability

• Build governance structures that prevent duplicates and conflicting definitions

• Create a unified data layer where possible before expecting AI to perform reliably

Reality check: this can be multi-year work

In complex legacy environments, data lineage, harmonization, and governance were described as long-haul efforts. Plan for it, fund it, and don’t fall into the trap of thinking this will be a quick prerequisite.

Part VII: Adoption and transformation

Adoption is not optional

Leaders emphasized that adoption has to be embedded from the start because AI introduces:

• Anxiety about job impact

• New ways of working with “digital coworkers”

• Trust and control questions (especially in risk-sensitive processes)

The organization needs a path, not a gatekeeper

Readiness standards shouldn’t become a reason to block progress. Instead, teams can use readiness criteria as a roadmap to determine what needs to be true (process, data, volume, governance) for a use case to scale responsibly.

Key Takeaways

• AI pilots fail at scale less because of the AI and more because of process variance, exception handling, and inconsistent data foundations.

• Avoid “AI for AI.” Start with problem clarity, confirm root causes, then decide whether AI is the right lever.

• If you apply AI to broken workflows, you often get a faster version of the same dysfunction.

• Design pilots to reflect enterprise reality: exceptions, edge cases, localization, regulatory constraints, and real data.

• Scaling requires an operating model: portfolio governance, clear ownership, and mechanisms to keep workflows current as conditions change.

• Define success across multiple lenses: ROI/value, cycle time, risk reduction, adoption, and readiness/maturity. Not just “the pilot worked.”

• Data harmonization and master data governance are frequently the true critical path and can be multi-year projects in legacy environments.

• Adoption is a clear senior leadership issue: teams need psychological safety, clarity, and practical enablement to work with AI-enabled processes.

The Executive Council for Leading Change

The Executive Council for Leading Change (ECLC) is a global organization that brings executives together to redefine the landscape of organizational change and transformation. Our council aims to advance strategic leadership expertise in the realm of corporate change by connecting visionary leaders. It's a place where leaders responsible for significant change initiatives can collaborate, plan, and create practical solutions for intricate challenges in leading large organizations through major shifts.

In a world where change is constant, we recognize its crucial role in driving business success. ECLC’s mission is to create a community where leaders can excel in guiding their organizations through these dynamic times.