Financial Services
Three FSI transformations that
nearly didn't happen.
Each was publicly celebrated after the fact. None of those retrospectives led with what came within days or months of ending the program.
Financial Services · Agile Transformation
ING Bank Netherlands
2015 – 2018
Restructured 57,000-person organization into agile squads, eliminating 13,000 roles, creating 2,300 new ones, and flattening six management layers into two.
What almost killed it
Middle management discovered their layer was being eliminated entirely — and they controlled the implementation timeline. Every restructuring decision required their sign-off. They used that authority to quietly delay, reschedule, and reframe the scope at every level. The program nearly stalled before leadership recognized that the people responsible for executing the restructuring were its primary obstacle. The turning point was removing sign-off authority from the layer being eliminated and routing all decisions through a dedicated transformation governance structure. Without that, the implementation would have died in committee.
What they got right
- Diagnostic baseline built before any announcement — exact role counts, spans-of-control, and redundancy mapping by function — so restructuring decisions were evidence-based rather than politically negotiated.
- Three org design scenarios modeled in parallel (including one that preserved more management layers) so the executive committee chose between options with visible tradeoffs, not a single consultant recommendation.
- Every affected employee received a direct, personalized communication — not a cascade through managers who had every incentive to distort the message. Rumor mitigation was systematic, not reactive.
Diagnostic Engine
Scenario Modeler
Stakeholder Orchestrator
Governance Layer
McKinsey Quarterly · MIT Sloan Management Review · HBR
Financial Services · AI Workforce Redesign
JPMorgan Chase — COIN
2016 – 2019
Contract Intelligence system automated 360,000 attorney hours annually, requiring workforce redesign across legal, compliance, and operations and creating new AI oversight roles.
What almost killed it
Legal and compliance staff refused to sign off on AI-generated contract reviews without an explicit, documented trail of what the AI decided and how it was reviewed. The same question came from regulators. Without governance documentation, the AI system's output was legally unusable — the 360,000 hours of projected savings were worthless. The program paused for months while the team rebuilt the governance architecture around every AI-assisted decision. That architecture became the reason the program survived regulatory review. Organizations that deploy AI without building the governance trail first end up building it in crisis mode.
What they got right
- Built audit documentation for every AI-assisted decision before scale-up — showing what the AI recommended, who reviewed it, and what the final decision was. Regulators could follow the chain. Legal staff could defend outcomes.
- Modeled reskilling scenarios for affected attorneys — which roles moved to AI oversight functions, which became redundant — before the announcement, not after the litigation threat.
- Phased execution with milestone tracking across legal, compliance, and technology simultaneously — so interdependencies were visible before they became blockers.
Governance Layer
Scenario Modeler
Execution Tracker
Bloomberg · Financial Times · JPMorgan Annual Reports
Financial Services · Digital Operating Model
DBS Bank Singapore
2014 – 2018
"22,000-person startup" transformation — full operating model redesign, digital-first delivery, workforce reskilling at scale across 18 markets — recognized as the most comprehensive bank transformation of its era.
What almost killed it
Year one produced almost no measurable results. The board lost patience. A faction of senior leaders argued publicly for abandoning the program and returning to incremental technology upgrades — a position that gained traction as competitors reported short-term gains from narrower initiatives. The transformation survived by a single board vote. It survived because the CEO could produce scenario analysis showing the five-year revenue trajectory for a digital DBS versus a traditional DBS — numbers the skeptics could not refute. Without that forward model, instinct would have beaten evidence, and the program would have been cancelled before the compounding effects became visible.
What they got right
- Scenario modeling of digital versus traditional revenue trajectories over five years — giving leadership a forward view of what abandoning the program meant, not just what continuing it cost in the near term.
- Live execution tracking across 22,000 employees in 18 markets — milestone visibility that showed the board what was moving and what was stalled, in real time, not in quarterly reports that arrived after problems had compounded.
- Systematic workforce reskilling with 1-in-4 employees retrained — managed through structured communication, not announcement-and-hope. Attrition of high performers was monitored and addressed before it became a delivery risk.
Scenario Modeler
Execution Tracker
Stakeholder Orchestrator
Harvard Business Review · World Economic Forum · Gartner
Healthcare
Three healthcare transformations
that came within one decision of failure.
Healthcare transformation fails differently than financial services. The near-kills are clinical, cultural, and regulatory — and they arrive with more public visibility than most C-suite leaders anticipate.
Healthcare · Quality & Operational Transformation
Cincinnati Children's Hospital
2003 – 2015
Lean quality transformation that cut medication errors by 80%, reduced care variation across the system, and positioned Cincinnati Children's in the top tier of U.S. pediatric hospitals for a decade.
What almost killed it
Senior physicians — who controlled the clinical culture and had credibility the administration did not — viewed Lean as a manufacturing methodology being imposed on medical judgment. When a group of high-influence physicians threatened to leave, the board was days away from dismantling the program to preserve clinical staff stability. The program survived when leadership made one structural change: physicians became co-designers of every clinical process change, not subjects of it. The governance documentation showing that every protocol decision was physician-led became the institutional defense against the accusation that administrators were overriding clinical judgment. That documentation did not exist at the start. It was built because the near-kill made it necessary.
What they got right
- Stakeholder mapping that identified which physicians were cultural leaders — not just formal department heads — and targeted them for co-design roles first, before resistance could solidify into an organized opposition block.
- Governance documentation showing every clinical process decision was physician-led — creating an audit trail that protected against the accusation of administrative overreach and provided the board with evidence when it was needed most.
- Diagnostic baseline of care pathway variation by physician and unit — data that gave clinical leaders evidence they could not dismiss because it came from their own systems, not external consultants.
Stakeholder Orchestrator
Governance Layer
Diagnostic Engine
Harvard Business Review · Institute for Healthcare Improvement · NEJM
Healthcare · Process Transformation
Virginia Mason Medical Center
2002 – 2013
Applied Toyota Production System to healthcare over 11 years, eliminating $11M in planned facility spending, reducing surgical infection rates, and establishing a replicable methodology for clinical process redesign.
What almost killed it
In year two, a patient death occurred. Critics — including physicians who had always opposed the program — publicly attributed it to the Lean standardization approach, arguing that clinical standardization had constrained the judgment that would have saved the patient. The story reached the local media. The board prepared to cancel the program. Leadership produced comprehensive documentation of every clinical process decision made under the Lean transformation — decision logs, clinical review records, sign-offs — demonstrating conclusively that the death was unrelated to any process change. That documentation had been maintained from day one, not assembled in response to the crisis. Without it, the program would have been cancelled on the basis of an accusation that leadership could not refute.
What they got right
- Complete audit trail of every clinical process decision from the program's first day — which turned out to be the program's survival mechanism when the public crisis arrived, not an administrative formality.
- Structured stakeholder engagement that included patient family advocates alongside physicians — building a coalition that withstood public criticism because it included the voices critics could not dismiss.
- Scenario modeling of care pathway redesigns against patient outcome data, not just efficiency metrics — so every change could be defended in clinical terms, not only operational ones.
Governance Layer
Stakeholder Orchestrator
Scenario Modeler
Harvard Business Review · Lean Enterprise Institute · JAMA
Healthcare · AI Clinical Integration
Intermountain Healthcare
2016 – 2022
Clinical standardization and AI-driven care pathway redesign across 25 hospitals — restructuring how clinical decisions are made, documented, and governed in a CMS-regulated environment.
What almost killed it
Utah state regulators questioned whether AI-influenced clinical decisions met documentation standards required for CMS reimbursement. Intermountain could not initially produce the decision trail showing how AI recommendations moved through clinical review and approval before being acted upon. Three months of CMS reimbursement eligibility was at risk for the affected care lines. The program paused while the organization rebuilt its governance documentation architecture to capture AI-assisted decisions with human oversight attributed, reviewed, and timestamped. The near-miss drove a governance capability that became a competitive differentiator — Intermountain could demonstrate documented AI governance to payers and regulators before most competitors understood the question was coming.
What they got right
- Rebuilt clinical governance documentation to explicitly capture AI-assisted decisions with human oversight attributed and timestamped — satisfying CMS requirements and producing the evidence payers and accreditors would eventually demand from every system.
- Involved CMO-level physicians in designing the AI governance framework — not just the technology deployment — which created clinical leadership ownership of the oversight structure rather than compliance with an IT-imposed one.
- Tracked implementation of care pathway changes across all 25 hospitals with a single source of truth — eliminating the version-control chaos that had plagued earlier protocol standardization efforts.
Governance Layer
Stakeholder Orchestrator
Execution Tracker
NEJM Catalyst · McKinsey Health Institute · CMS Documentation
Government
Three government transformations
that hit statutory walls no one had mapped.
Government transformation fails in ways that do not exist in the private sector. Statutory constraints, appropriations rules, and civil service protections are not obstacles to work around — they are the environment. The organizations that succeeded understood that before they started.
Government · Digital Services Transformation
UK Government Digital Service
2012 – 2016
Consolidated 25 major government services into a single digital platform (GOV.UK), cutting the cost of government digital delivery by hundreds of millions of pounds and establishing a model replicated across 30 countries.
What almost killed it
Seven of 25 cabinet departments refused to participate. They argued their services were too complex for a common platform — and they had independent IT budgets and statutory authority over their own digital services to back that position. The program's architects had not mapped the statutory authority question before launch: they had assumed buy-in that did not exist. The standoff consumed nine months and nearly fractured the program along departmental lines. It resolved when the Cabinet Office established unambiguous statutory authority for the migration — authority that should have been documented at the outset, not negotiated after seven departments had already said no. The lesson was replicated in every subsequent government digital transformation that succeeded.
What they got right
- After the statutory authority crisis, mapped the legal constraints and spending authorities for each of the remaining 18 departments before designing their migration path — so the solution fit the statutory environment rather than requiring the environment to change.
- Cross-department execution tracking that made delays visible to the Cabinet Office before they compounded into missed service dates — eliminating the information asymmetry that had allowed the holdout departments to delay without consequence.
- Governance documentation designed to survive administration changes — three UK governments used and extended the GDS work precisely because the decisions and rationale were recorded in a form that new ministers could understand and build on.
Government Module
Execution Tracker
Governance Layer
UK Cabinet Office · GDS Blog · Institute for Government
Government · AI & Workforce Restructuring
Singapore Smart Nation Initiative
2014 – 2022
Whole-of-government AI and digital restructuring — redesigning how 145,000 civil servants work, what technology delivers, and how public services are governed in an AI-integrated operating model.
What almost killed it
Civil service unions raised formal concerns about AI replacing public sector workers. The concern became a parliamentary question. The government could not initially quantify how many jobs were actually at risk because no workforce impact model had been built before the AI deployments were announced. The political exposure was immediate and the questions were specific: which agencies, which functions, how many people, over what timeline. The inability to answer with numbers forced a six-month pause to build the workforce impact model that should have existed before the first announcement. The delay was costly in political terms. The lesson — model workforce impact before you announce, not in response to the announcement — became the organizing principle for every subsequent phase of the initiative.
What they got right
- After the parliamentary crisis, modeled workforce impact scenarios for each agency before any further AI deployment was announced — so the government could answer specific questions with specific numbers rather than managing a public information vacuum.
- Structured stakeholder engagement with union leadership through direct dialogue rather than press releases — addressing the substantive workforce concerns in the room where they originated, not in official statements that amplified the political visibility of the issue.
- Comprehensive audit documentation for every AI deployment decision, enabling public accountability and parliamentary scrutiny — a governance capability that became a policy export to other governments.
Scenario Modeler
Stakeholder Orchestrator
Governance Layer
Singapore PMO · World Economic Forum · OECD Public Governance
Government · Federal Agency Restructuring
US Digital Service
2014 – 2021
Agency-by-agency restructuring of federal digital delivery — rescuing HealthCare.gov, redesigning VA digital services, and establishing a replicable model for technical transformation inside federal statutory constraints.
What almost killed it
On multiple early agency engagements, USDS teams hit appropriations constraints and civil service hiring rules that no one had mapped before the work began. The HealthCare.gov rescue succeeded partly because the team moved under emergency authority before statutory constraints were enforced. Later engagements at the VA and elsewhere hit legal walls that cost 6–12 months each — delays that in a government context meant budget cycles, administration priorities, and political windows closing. The constraint-mapping that should have preceded the work was instead done reactively, in response to each wall that appeared. The USDS teams that worked most effectively were the ones that treated statutory constraint mapping as phase zero — before scoping, before staffing, before commitments were made to agency leadership.
What they got right
- Developed a statutory constraint mapping methodology for each agency engagement — identifying what could be changed under existing authority, what required congressional action, and what could be addressed through administrative workarounds — before scoping any work.
- Cross-agency execution tracking for shared infrastructure dependencies — identity systems, payment platforms, security certifications — that individual agency teams could not see, preventing the same blockers from appearing independently on parallel engagements.
- Governance documentation designed to survive administration transitions — playbooks, decision logs, and institutional memory that successor teams could use when the political appointees who originated the work had left.
Government Module
Execution Tracker
Governance Layer
USDS Public Reports · Brookings Institution · Government Accountability Office