https://weswest.ai/ Recent content on Wes West Hugo -- gohugo.io en-us Tue, 14 Apr 2026 00:00:00 +0000 https://weswest.ai/writing/contracts-skeleton/ Tue, 14 Apr 2026 00:00:00 +0000 https://weswest.ai/writing/contracts-skeleton/ This is the fifth article in a series on running an eight-agent AI coding team. Across the first four articles, a single design philosophy keeps surfacing: define one structural path for the system to succeed, then make it difficult to deviate from that path. Everything else is fault tolerance. Three principles have emerged so far: Make wrong behavior structurally impossible, not just discouraged. Agents run in physically isolated worktrees so they can’t see each other’s in-progress work. https://weswest.ai/writing/self-auditing-system/ Tue, 07 Apr 2026 00:00:00 +0000 https://weswest.ai/writing/self-auditing-system/ The first three articles in this series described a system: seven domain-specialist agents and a manager, a three-layer enforcement model keeping them in bounds, and a sprint cycle that coordinates their work. What none of them addressed is what happens when an agent follows all the rules and still writes bad code. Not broken code — code that passes tests and respects every boundary. Just code that’s inconsistent, poorly documented, or quietly drifting from conventions nobody explicitly checks. https://weswest.ai/writing/sprint-cycle-walkthrough/ Tue, 31 Mar 2026 00:00:00 +0000 https://weswest.ai/writing/sprint-cycle-walkthrough/ My last two articles — It’s Just Microservices and Team Management and Policy Enforcement for Agentic Workflows — documented how I set up an eight-agent AI coding team: seven domain specialists and a manager, each operating in isolated workspaces with layered enforcement keeping them in bounds. In this article, I share how these agents coordinate through a sprint cycle — and how I spend nearly all of my time on one tiny but sacred sliver. https://weswest.ai/writing/policy-enforcement-agentic-workflows/ Tue, 24 Mar 2026 00:00:00 +0000 https://weswest.ai/writing/policy-enforcement-agentic-workflows/ Agentic AI is a management problem. Most people have never managed others’ work — never had to set someone else’s objectives, define best practices, verbalize exactly what “success” and “failure” look like, implement an organizational structure, and then quality-control the output. That’s the new skill, and most of the field is trying to skip it. I wrote about this more broadly in AI Makes Everyone a Manager. This article is the specific version: how I set up guardrails for my AI coding team, and why it looks exactly like goals-oriented management. https://weswest.ai/writing/multi-agent-microservice/ Tue, 17 Mar 2026 00:00:00 +0000 https://weswest.ai/writing/multi-agent-microservice/ I’m running an all-AI team of seven programmers and a manager. I provide guidance, strategic direction, and surgical intervention; the seven programmers handle their workloads; the manager coordinates. It works really well. Why does it work so well? The short version: I didn’t invent anything. I just noticed that an agentic coding team is both a software architecture problem and a team management problem at the same time, and applied known best practices from both. https://weswest.ai/writing/complexity-tax/ Mon, 02 Feb 2026 00:00:00 +0000 https://weswest.ai/writing/complexity-tax/ Originally published at Nomis Solutions A neural network achieves 0.82 AUC. The logistic regression gets 0.72. The gap between them is what we call the complexity tax — the accuracy you sacrifice when you choose interpretability over raw predictive power. This article explains why we routinely build models we know will never hit production, and how doing so leads to better conversations with the C-suite about what to actually deploy. https://weswest.ai/writing/voice-enabled-mortgage-assistant/ Thu, 14 Aug 2025 00:00:00 +0000 https://weswest.ai/writing/voice-enabled-mortgage-assistant/ Originally published on Nomis Solutions Sometimes the most valuable projects are the ones you don’t pursue. Our team built a voice-enabled mortgage application system in just four hours, then decided not to take it to production. The lessons about compound AI systems proved more valuable than the prototype. Key Topics: Building compound AI systems by orchestrating specialized tools Technical architecture: Whisper, GPT-3.5, voice activity detection, and text-to-speech Five critical engineering decisions for natural conversation interfaces Why we didn’t ship: customization, voice quality, error handling, and integration challenges Knowing when to prototype, learn, and move on https://weswest.ai/writing/ai-recommendation-engines/ Thu, 24 Jul 2025 00:00:00 +0000 https://weswest.ai/writing/ai-recommendation-engines/ Originally published on Nomis Solutions Most recommendation systems provide binary responses: approve or deny, raise rates or lower them. This article demonstrates how AI can provide nuanced decision support that goes far beyond simple yes/no recommendations. Key Topics: Limitations of traditional rule-based recommendation systems How AI provides arguments for and against each decision Generating creative alternative solutions beyond predetermined options Contextual insights that adapt to unique circumstances Combining quantitative models with qualitative judgment The evolution from recommendation engines to AI business analysts https://weswest.ai/writing/ai-makes-everyone-a-manager/ Fri, 11 Jul 2025 00:00:00 +0000 https://weswest.ai/writing/ai-makes-everyone-a-manager/ Originally published on Nomis Solutions Millions of knowledge workers just became managers overnight. They don’t know it yet, and that’s precisely why so many AI initiatives are failing. This article explores why AI tools require management skills, not technical skills. Key Topics: Why AI tools behave like team members, not calculators The critical management skills needed for effective AI usage Setting clear expectations and providing appropriate context Giving constructive feedback to improve AI outputs Building organizational capacity for AI adoption https://weswest.ai/writing/accelerate-data-analysis/ Fri, 27 Jun 2025 00:00:00 +0000 https://weswest.ai/writing/accelerate-data-analysis/ Originally published on Nomis Solutions The debate about AI in banking often frames it as a choice: humans or machines. This article demonstrates how combining AI’s data processing capabilities with analyst expertise creates superior outcomes. Key Topics: Moving beyond reactive chatbots to proactive AI intelligence The critical importance of domain-specific banking knowledge in AI systems How AI can scan thousands of data points to surface patterns and anomalies Practical examples of AI-augmented analysis in deposit pricing and portfolio management The irreplaceable role of human judgment in strategic decision-making Building effective human-AI collaboration workflows https://weswest.ai/writing/pragmatic-approach-to-ai/ Wed, 18 Jun 2025 00:00:00 +0000 https://weswest.ai/writing/pragmatic-approach-to-ai/ Originally published on Nomis Solutions Most AI projects fail due to overpromising and unrealistic expectations. This article outlines Nomis’s pragmatic approach to AI implementation, focusing on practical applications that create measurable value. Key Topics: Why most AI projects fail and how to avoid common pitfalls Three core AI use cases: data analysis, recommendation engines, and decision support The reality of AI implementation: augmenting rather than replacing human judgment Measuring ROI and demonstrating tangible business value https://weswest.ai/writing/fis-emerald-2025-recap/ Thu, 29 May 2025 00:00:00 +0000 https://weswest.ai/writing/fis-emerald-2025-recap/ Originally published on Nomis Solutions FIS Emerald 2025 highlighted a critical gap in banking’s AI journey: while 85% of banks believe generative AI will transform the industry, only 10% have concrete implementation roadmaps. This recap examines how to close that gap. Key Topics: The AI adoption challenge: belief vs. execution Data privacy, security, and regulatory concerns in banking AI Practical AI implementation strategies for financial institutions Building trust through transparency and explainability Creating actionable AI roadmaps for banking organizations https://weswest.ai/writing/deposit-betas-lessons/ Mon, 13 Jan 2025 00:00:00 +0000 https://weswest.ai/writing/deposit-betas-lessons/ Originally published on LinkedIn The 2022-2023 rate cycle taught the banking industry an expensive lesson about deposit betas. This analysis examines why actual deposit cost sensitivity far exceeded forecasts and provides a framework for improving future predictions. Key Topics: Understanding deposit beta fundamentals and measurement Why banks systematically underestimated deposit betas (20-30% forecasts vs. 50-60% reality) The role of digital banking, competitive dynamics, and customer sophistication Practical approaches to improving deposit beta forecasting Implications for balance sheet planning and NIM management https://weswest.ai/writing/adapting-to-rate-cuts/ Wed, 18 Sep 2024 00:00:00 +0000 https://weswest.ai/writing/adapting-to-rate-cuts/ Originally published on LinkedIn In the immediate aftermath of the Federal Reserve’s 50 basis point rate cut, financial institutions face unprecedented challenges in managing deposit expenses in a falling rate environment. This article examines how banks can navigate this transition strategically. Key Topics: Analysis of deposit beta forecasting failures during the 2022-2023 rate cycle First mover vs. fast follower strategic considerations Managing CD and promotional product maturities Identifying and differentiating rate-sensitive vs. https://weswest.ai/writing/model-factory-management/ Fri, 01 Jun 2018 00:00:00 +0000 https://weswest.ai/writing/model-factory-management/ Publication: Novantas Review (Summer 2018) Co-authors: Steve Turner, Kaushik Deka As banks struggle to manage hundreds of models requiring increasing resources, this article proposes using sophisticated technology, including machine learning, to automate the model management process. https://weswest.ai/writing/rethinking-modeling-ecosystem/ Tue, 01 Aug 2017 00:00:00 +0000 https://weswest.ai/writing/rethinking-modeling-ecosystem/ Publication: Novantas Perspective (August 2017) Co-authors: Ryan Schulz, Steve Turner For banks in many countries the current model development and validation process is broken: it takes too long, there are too many nonstandard elements, there is too much rework, and it introduces material risk when critical models fail validation. https://weswest.ai/writing/cd-share-deposit-models/ Thu, 01 Sep 2016 00:00:00 +0000 https://weswest.ai/writing/cd-share-deposit-models/ Publication: Novantas Perspective (September 2016) Co-authors: Jane Lim, Ryan Schulz To address the data shortcoming that many banks lack sufficiently long internal monthly history covering multiple rate cycles, Novantas uses the time deposit share of total interest-bearing deposit balances (“CD-Share”) as an independent industry variable in deposit balance models. https://weswest.ai/writing/forecasting-percentage-growth/ Tue, 10 May 2016 00:00:00 +0000 https://weswest.ai/writing/forecasting-percentage-growth/ Publication: Novantas Perspective (May 2016) Co-authors: Sean Wallace, Ryan Schulz One of the most common transformations of the dependent variable in PPNR Stress Testing is to predict growth in balances, revenue, and expenses from period to period. This article outlines why the preferred transformation is a logarithmic one (diff-log), as opposed to percentage change. https://weswest.ai/writing/2017-ccar-planning/ Tue, 12 Apr 2016 00:00:00 +0000 https://weswest.ai/writing/2017-ccar-planning/ Publication: Novantas Perspective (April 2016) Co-authors: Pete Gilchrist, Ryan Schulz, Sean Wallace With 2016 CCAR submitted and the cycle shortened from 15 to 12 months, this article highlights the eight most critical immediate steps toward a successful 2017 submission. https://weswest.ai/writing/deposit-rate-modeling/ Tue, 15 Mar 2016 00:00:00 +0000 https://weswest.ai/writing/deposit-rate-modeling/ Publication: Novantas Perspective (March 2016) Co-authors: Greg Muenzen, Steve Turner, Pete Gilchrist PPNR and ALM modeling teams know that the bar for deposit rate modeling has risen. This article explains the Novantas regression-based approach and compares its performance to beta and replicating portfolio techniques. https://weswest.ai/writing/spurious-correlation-failures/ Tue, 01 Mar 2016 00:00:00 +0000 https://weswest.ai/writing/spurious-correlation-failures/ Publication: Novantas Perspective (March 2016) This article describes a bug in the SAS PROC AUTOREG procedure discovered during 2015 PPNR model development: models with no serial correlation fail the Godfrey serial correlation test when also performing certain stationarity tests. https://weswest.ai/writing/ccar-ppnr-modeling/ Sat, 01 Nov 2014 00:00:00 +0000 https://weswest.ai/writing/ccar-ppnr-modeling/ Publication: Novantas Review, Vol 5, No 3 (November 2014) Co-authors: Andrew Frisbie Improved modeling for pre-provision net revenue requires strengthening the project framework, working to overcome data sterility and testing alternative approaches. This article addresses the paradox that even as banks improve their competency in PPNR modeling, many institutions are shifting toward an increasingly similar and rigid approach to satisfy statistical tests, exposing banks to business line disconnect and systemic risk. https://weswest.ai/search/ Mon, 01 Jan 0001 00:00:00 +0000 https://weswest.ai/search/ This file exists solely to respond to /search URL with the related search layout template. No content shown here is rendered, all content is based in the template layouts/page/search.html Setting a very low sitemap priority will tell search engines this is not important content. This implementation uses Fusejs, jquery and mark.js Initial setup Search depends on additional output content type of JSON in config.toml [outputs] home = ["HTML", "JSON"] Searching additional fileds To search additional fields defined in front matter, you must add it in 2 places.