I help leaders make better decisions by connecting strategy to the systems, models, data, and governance underneath it.
For twenty-five years inside major U.S. financial institutions, I worked across that entire spectrum — building economic capital and stress‑testing frameworks, implementing interest‑rate risk systems, and establishing an AML data science function using machine learning. I built not only the analytics, but also the governance and decision processes needed to make them useful, durable, and defensible through regulatory, audit, and independent third‑party review.
I’m a builder at heart, and that work continues today. Alongside executive advisory in risk, financial crime, AI, and technology, I am actively developing new tools, frameworks, and ventures that turn ideas into working solutions. That combination of executive perspective, technical depth, hands‑on building, and experience operating under sustained scrutiny is what I bring to clients.
The same sequence every time, whether the subject is capital, rate risk, or financial crime.
The same sequence every time, whether the subject is capital, rate risk, or financial crime.
Five disciplines, twenty-five years. What I actually built in each.
Provided strategic direction on AML technology to an international FIU to help them chart a path through their current plans to the next generation of detection and investigation capabilities.
International FIUThe program relied on rules that had outlived the typologies behind them. We integrated supervised classification and clustering into AML models, delivering more than $5 million in operational efficiency — fewer false positives and more true positives, not a trade of one for the other.
$5M+ efficiency · false positives down, true positives upRegulators care about behavior detection coverage before they care about elegance. We moved the program off inefficient rules to full adoption of machine learning algorithms, and closed the open coverage requirements in the same work.
coverage maintained through the transitionMachine learning in a regulated detection program is only allowed if it can be examined. We built the formal modeling governance structure that made the AML models defensible — and cleared the open regulatory findings against the program.
MRAs clearedDetection has a coverage gap no single bank can close alone. Partnering with a start-up, we pioneered federated learning architectures for cross-institutional behavior detection — sharing intelligence while preserving privacy and data sovereignty.
privacy-preserving · cross-institutionalNamed inventor on two pending U.S. patent applications in AML detection technology, assigned to PNC — US 2025/0156940 A1, on efficient detection of money laundering, and US 2025/0348880 A1, on monitoring customer transactions for potential financial crimes.
RegTech · financial crime detectionFraud and money laundering increasingly share common data, behavioral signals, and investigative tools. I helped define where integration between fraud and AML made practical sense — particularly across data management, analytical tools, and shared features — and helped shape the AML team’s contribution to the broader FRAAML integration program.
FRAAMLI introduced graph analytics to map the relationships between products, typologies, and models — and to surface unusual peer-to-peer transactions that row-by-row analysis cannot see.
products × typologies × modelsAfter 2008, supervisors wanted stress testing no one had built yet. I architected and operationalized the firm's first Supervisory Capital Assessment Program — a precursor to CCAR — by developing a multi-risk stress testing framework that exceeded what the regulators asked for.
precursor to CCAR · exceeded expectationsRisk-weighted assets, economic capital, and stress-test capital each said something different about the same balance sheet. We reconciled RWA, economic capital under both Value-at-Risk and Expected Shortfall, and CCAR stress measures — turning three incompatible answers into differentiated insight that supported capital adequacy assessment and regulatory reporting.
RWA · VaR / ES · CCARAggregating risk means knowing how risks move together. We improved portfolio concentration analysis and measured inter-risk correlation, so aggregation and scenario analysis rested on measured relationships.
risk aggregation · scenario analysisPricing that ignores capital cost misprices risk. We advanced correlation models across retail, commercial, and CRE portfolios, bringing loan pricing into alignment with capital costs, risk appetite, and origination strategy.
retail · commercial · CRECapital estimation was opaque to the people making pricing decisions. I founded and chaired the Economic Capital Committee, aligning risk modeling, capital planning, and RAROC so that strategic pricing decisions rested on numbers everyone had agreed to.
RAROC · strategic pricingRegulators wanted an enterprise view of capital that the bank did not have. I led the design and implementation of an enterprise-wide economic capital modeling and regulatory capital reporting framework at National City — the first the firm had. After the merger it became the foundation PNC integrated onto.
enterprise-wide · built from nothingThe bank's liquidity process could not see far enough ahead. I designed and implemented a multi-horizon liquidity forecasting capability inside the QRM simulation model, so liquidity could be managed on the same platform as rate risk.
QRM · multi-horizonIncome volatility analysis was missing its biggest driver. I initiated development of a mortgage loan production forecast model built on macroeconomic drivers, then folded it into the QRM earnings-at-risk simulation framework to provide ALCO with a holistic view of enterprise interest rate risk exposures.
macroeconomic drivers · EaRI established an economic value-at-risk framework to assess long-term market value sensitivity, fully capturing embedded optionality across products ranging from MBS and whole‑loan mortgages to non‑maturity deposits — the features that often drive a bank’s underlying interest‑rate risk.
EVaR · Embedded optionsThe bank's reliance on mortgage banking revenue was not visible enough to manage. I founded and chaired the Mortgage Banking Earnings-at-Risk Committee, a subcommittee of ALCO, integrating repricing risk, credit spreads, and loan demand elasticity to optimize risk-adjusted profitability.
ALCO subcommitteeI delivered the Active Credit Portfolio Management toolkit at three IACPM Annual Meetings — Stamford, Miami, and Washington — then recorded it for the Online Educational Seminar in 2022 and refreshed it in 2025. It has run every year since. I previously served on a working group on evolving capital approaches.
curriculum · working group · Senior AdvisorI established the AML Data Science Solutions team at PNC and directed the design and oversight of both traditional analytics and machine-learning detection across transaction monitoring, sanctions screening, and customer risk rating — a function that did not exist before it was built.
founded the functionA consortium of more than forty North American banks had not met since 2019 — its leader retired, then COVID arrived, and it stopped. In 2023 I brought it back, then led the member discussions that produced what it never had: an agreed organizational structure and governance model for running it year over year. I served the first one-year term as President and handed it on.
40+ institutions · one-year termPost-merger, two capital frameworks had to become one. I led the integration at PNC, managing a 35-member cross-functional team spanning quantitative analytics, operations, and model governance, along with the budget and vendor relationships behind it.
35-member team · budget & vendor ownershipModeling only matters if the answer lands. Across ALCO, executive committees, the board, and regulators, my role was translating what the models said into decisions people could actually take.
ALCO · board · regulatorsI implemented a change control framework across model development so that what changed, when, and why was transparent — the difference between a model you can defend and one you can only explain.
transparency in developmentI represented the modeling teams before regulators across capital and AML, which shaped how we documented and evidenced the work long before any examination began.
OCC · FinCENSelected outcomes
I have spent much of my career building risk capabilities before they became standard practice. My career has largely been about taking on new problems as the risks facing banks changed.
At National City, I began by working on the measurement of interest‑rate risk. I led the implementation of the firm’s QRM interest‑rate risk management platform, incorporating economic value of equity into the IRRBB framework, calibrating mortgage prepayment and non‑maturity deposit behavior, and implementing earnings‑at‑risk simulation. As the work matured, I founded and chaired the Mortgage Banking Earnings‑at‑Risk Committee, creating a direct link between what the models were telling us and the people making balance‑sheet and business decisions.
From there, my focus broadened from individual risk types to the enterprise as a whole. I built National City’s first enterprise‑wide economic capital framework, establishing a common basis for allocating capital across credit, market, operational, interest‑rate, and business risk. After the financial crisis, that work evolved again as supervisory expectations changed. I architected and operationalized the firm’s first Supervisory Capital Assessment Program, the precursor to CCAR.
When PNC acquired National City, I continued that work on a much larger scale, expanding credit‑risk economic capital capabilities for the combined portfolio and strengthening governance of the enterprise framework. I founded and chaired the Economic Capital Committee to bring consistency and senior oversight to how the framework was used across the organization.
Later in my career, I moved into financial crime, where I saw another opportunity to build a capability that did not yet exist. I was an early proponent of machine learning in this space and established AML Data Science Solutions at PNC, a cross‑functional group supporting transaction monitoring, sanctions screening, customer risk rating, and operations simulation. We developed the standards and model governance needed to use machine learning in a regulated environment. The work helped resolve outstanding regulatory findings and generated more than $5 million in operational efficiencies.
That progression — from interest‑rate risk, to enterprise capital, to stress testing, and ultimately to machine learning and financial crime — is the consistent thread in my career: taking emerging or complex problems, building the analytical and governance structures needed to address them, and connecting the results to real business decisions.
Career
Advisory
My advisory practice. I work with banks and financial institutions on three things: AML and financial crime detection, capital and balance‑sheet risk, and AI and agentic risk governance. Engagements run from fractional and interim risk leadership to board and executive advisory. Every one is written to be defended in front of a board, an auditor, or an examiner.
semperclarusrisk.com — launching soon
Active projects
Three systems I’m building now. Two in financial crime, one well outside it. All three come from the same training: define the decision, constrain it honestly, make the answer reproducible and auditable.
Model teams label, review, and log every decision in one workflow, so the training data behind a detection model can be examined the same way the model is. Built in partnership with RCS Analytics.
We don’t make the models you use; we make building them easier.
Ask about a design partnership →Two methods open into 35 loss‑producing typologies and the 46 techniques that compose them, staged across a four‑step kill chain, each carrying its own detection signals and controls.
Fraud taxonomies usually blur channel with method. Separating them — vishing is a channel, pretexting is a method — is what makes coverage gaps visible.
Open the taxonomy →Set goals and a linear program returns a diet that satisfies them against multiple nutrient standards and anthropometric reference data, then evaluates the result.
The same optimization discipline I used to tune AML detection parameters, pointed somewhere unexpected. Different domain, identical method.
Request early access → rick.dietcheck@gmail.comWorking sessions with frameworks the room can take back to work on Monday — delivered to practitioners, executive audiences, and graduate programs.
Thought leadership
Most AI risk conversations collapse three very different technologies into “AI.” The difference decides how you govern them — and whether your existing model risk framework is enough.
Governed for years
Turns data into scores and predictions. This is model risk, and the discipline is mature: validate before deployment, monitor performance, control change.
e.g. logistic regression, gradient boosting, transaction monitoring scores
The recent problem
Turns context into text, code, and summaries. The risk is what the AI says. You manage it by validating outputs before anyone relies on them.
e.g. drafted SAR narratives, policy summarization, code generation
The new problem
Combine reasoning, tools, and memory to act across your systems. The risk is no longer what AI says. It is what AI does — and pre‑deployment validation does not reach it.
e.g. an agent that queries, decides, and files without a human in the loop
Agents have none of the inhibitions people do.
Enterprise risk management rests on a quiet assumption — that people self‑govern because they fear consequences. Agents execute with speed, persistence, and scale, and fear nothing. That calls for a different discipline: Agent Conduct Monitoring — supervising a digital actor at runtime, enforcing authority limits, detecting deviations, and being able to pause or shut it down.
Published
The proposed rule moves supervision from process adherence to demonstrable effectiveness, and explicitly invites experimentation with collaborative and federated machine learning. Why institutions that take up the invitation will examine better than those running isolated systems.
Co-authored with Gary M. Shiffman, co-founder of Consilient, and Shelly Liposky of BMO Capital Markets. How AI can train across many banks’ data without compromising customer privacy — and why that changes the economics of detection.
Money laundering persists because criminals have strong incentives to innovate faster than the institutions detecting them. What that arms race demands of behavior detection.
Longer-form writing on where AI, financial crime compliance, and enterprise risk meet — how regulation, data, models, and technology have to align to support sound decisions.
One idea worked all the way through each week — a short teaser Tuesday, the full framework carousel Thursday. Recent: the three forms of AI risk, and the five AI exposure surfaces.
A design partnership on the tools, an advisory engagement, a session for your team, or a longer conversation — start here.
Connect on LinkedIn