All articles
The SRE Approach to AI: Why Reliability Engineering Is the Missing Piece in Business Automation
The AI industry has a delivery problem, not a capability problem. Models can do incredible things, but most AI projects fail because nobody built the operational scaffolding to keep them working. SRE is that scaffolding.
Cost Anomaly Detection: Catching AWS Spend Before It Spirals
One client’s AWS bill spiked from $800 to $8,400 in 48 hours before anyone was watching for it. The three-layer defense I build into every infrastructure to catch spend anomalies before they become a crisis.
The Client Onboarding Playbook for AI Automation Projects
The structured onboarding process for AI projects: discovery call, data audit, architecture proposal, SOW, kickoff — and how to avoid the miscommunication disasters that kill them.
AI for Real Estate: Automating Property Analysis and Comps
How real estate teams use Claude to extract listing data, pull comps, and generate market analysis reports, cutting 4-6 hours of work down to 30 minutes.
Dead Letter Queues: The Safety Net Your Pipeline Is Missing
Why every AI pipeline needs an SQS dead letter queue: setup, monitoring, investigation, and replay patterns with Terraform and Lambda examples.
When to Build vs. When to Buy AI Solutions
A five-factor framework for deciding whether to build custom AI automation or buy a SaaS platform, with real cost comparisons and red flags for each path.
Multi-Tenant AI Architectures for Consulting Firms
How to serve multiple clients from one codebase: DynamoDB partition-key isolation, per-tenant cost attribution, failure isolation, and Terraform modules for onboarding new tenants.
Token Economics: What You’re Actually Paying For With AI APIs
Prompt caching cuts input costs 90%, batch processing 50%. How token pricing actually works, with real per-task cost math and six optimization checks.
The Business Case for Infrastructure Ownership
SaaS at $33k over 3 years vs $14.5k owning it on AWS. The TCO math, the maintenance reality, and when renting still wins.
Your AI System Needs a Disaster Recovery Plan
DR planning for AI systems: what happens when your provider goes down, fallback paths that keep you operating in degraded mode, and the one-page runbook format I’ve used after three client outages.
AI for Legal Operations: Contract Review and Clause Extraction
Legal ops teams spend thousands of hours reading contracts. An S3-to-DynamoDB pipeline that extracts clause terms and flags non-standard language with Claude — the architecture, the code, the compliance considerations, and the cost math.
Structured Outputs Changed How I Build AI Systems
Regex parsing breaks the moment Claude’s phrasing shifts. Structured outputs and tool use guarantee the format instead — the before-and-after, a real invoice-extraction example, and a migration path off regex.
The 30-Day AI Pilot Framework for Small Business
A structured 4-week framework for running an AI pilot that actually produces a go/no-go decision — what to measure, what it costs, and the mistakes that kill most pilots before they finish.
Why Most AI Demos Never Make It to Production
A working demo and a production system are not the same thing. Here’s what’s actually missing — error handling, validation, monitoring, deployment pipelines, documentation — and what it costs to skip it.
Document Processing Pipelines: PDF to Structured Data on AWS
An architecture for extracting structured data from PDFs using S3, Lambda, Claude, and DynamoDB — plus a cost comparison against Textract for invoices, contracts, and claims at scale.
How DNS Misconfigurations Quietly Break AI Integrations
A real production war story: how a CNAME chain and an aggressive DNS TTL caused intermittent Lambda timeouts calling the Claude API, and the Route53 fixes that made it stop.
The Vendor Evaluation Checklist I Give Every Client
A practical checklist for evaluating AI vendors and consultants — what to ask, what red flags to look for, and how to avoid getting locked in.
The State of AI Automation for Small Business in 2026
Where AI automation actually stands in 2026: what works, what doesn’t, what’s overhyped, and what small businesses should build right now.
What I Learned Building 10 AI Automations for Small Businesses
Ten AI automation projects in eight months. The lessons that held up: most projects are data pipelines, reliability beats magic, and clients care more about speed than accuracy.
From Spreadsheet to System: Migrating Manual Processes to Automation
A 6-step framework for replacing a spreadsheet-based business process with reliable automation — without disrupting operations or losing data.
Building AI Systems That Survive Model Updates
Version pinning, output schema contracts, and regression testing: the three patterns that keep your AI automation running when your model provider ships a new version.
How I Build AI Automation That Survives Production
The demo is the easy part. Here’s the SRE-driven workflow I use to build small-business AI automation that runs unattended, survives model updates, and stays yours — fixed-price, fully owned.
The 3 AWS Services Every Small Business Should Know
Lambda, S3, and DynamoDB cover 80% of small business automation needs. Here’s what each does in plain English, with real costs and a complete workflow example.
Pricing AI Consulting: Why I Use Fixed-Price and You Should Too
Hourly billing rewards inefficiency. Here’s why fixed-price models work better for AI consulting, with real examples, a scope document template, and pricing tiers that scale.
How I Use Claude's Tool Use to Build Self-Correcting Pipelines
How to use Claude's tool_use feature to build self-correcting pipelines that validate their own output, catch errors before they reach your database, and automatically retry with feedback.
The 15-Minute Security Audit for Your AI Automation
A practical 15-minute checklist to secure your AWS AI automation before it ships. Covers API keys, IAM least-privilege, encryption, input validation, and dependency scanning.
AI Agents Are Coming. Your Business Isn’t Ready.
The three guardrails - human-in-the-loop, audit trails, output validation - that keep autonomous AI agents from making expensive mistakes in your business.
Event-Driven Architecture for Small Business: S3 + Lambda + DynamoDB
Small businesses don’t need Kafka. S3 + Lambda + DynamoDB makes resilient, near-zero-cost event-driven automation pipelines you can build in an afternoon.
Why I Don’t Use LangChain (And What I Use Instead)
Why I avoid LangChain for production LLM systems. Abstraction leakage, debugging hell, and the explicit-Python alternative I use with the Anthropic SDK.
AI for Healthcare Admin: Automating Prior Authorizations and Intake Forms
Automate prior authorizations and intake forms with AI. HIPAA-compliant document extraction, validation, and routing on AWS that cuts 40% of manual processing time.
The Solo Consultant’s Guide to Production-Grade Deployments
Build a production-grade deployment system as a solo consultant using GitHub Actions, Terraform, and AWS OIDC. No stored credentials, automated rollbacks, environment promotion — one person, production-grade.
How to Calculate ROI on an Automation Project Before You Build It
A practical framework for calculating the ROI of automation projects before you build them. Includes formulas, real numbers, and a reusable template for presenting to your CFO.
Stop Building AI Features. Start Automating Workflows.
AI features impress in demos. AI-automated workflows deliver ROI. Here’s how to identify, scope, and prioritize the workflows worth automating in your business.
Serverless vs. Containers for AI Workloads: When to Use Which
A practical decision framework for choosing between AWS Lambda and ECS/Fargate for AI workloads, with real cost comparisons and architecture patterns.
The Runbook: Your AI System’s Most Important Document
A runbook is the difference between “we hope someone knows what to do” and “we have a plan.” Here’s how to build one for your AI system before 2am hits.
AI for Retail: Demand Forecasting Without a Data Science Team
You have 18 months of sales history and no data scientist. Here’s how to build a practical demand forecasting system on AWS using Claude, Lambda, and DynamoDB for about $30/month.
Building Idempotent Data Pipelines on AWS
Idempotency is the difference between a data pipeline and a nightmare. Here’s how to build pipelines on AWS that handle retries, duplicate events, and network failures without corrupting your data.
Prompt Engineering Is Not Software Engineering (And That’s the Problem)
Prompt engineering is instruction writing. Software engineering is architecture. Here’s how structured outputs, validation layers, and schema contracts bridge the gap for production AI systems.
How I Structure Client Projects for Zero-Downtime Handoff
Most AI consulting projects fail at handoff. Here’s the six-phase structure — repo layout, documentation, monitoring, CI/CD, training, and warranty — that lets clients own their system from day one.
Monitoring AI Systems: What to Measure and Why
Your AI system is up but silently degrading. Here are the five metrics — token cost, validation rate, latency percentiles, model drift, and end-to-end success — that catch problems before users do.
The Small Business Guide to AWS Costs (It’s Cheaper Than You Think)
AWS will bankrupt us — said every SMB owner before seeing the real numbers. Here’s what serverless automation actually costs, with a free tier that covers most small businesses entirely.
Why Your AI Chatbot Needs a Kill Switch
Customer-facing AI will have an off day. Here’s the real-time confidence threshold architecture — validation rules, escalation routing, and CloudWatch alarms — that keeps you out of PR trouble.
AI for Professional Services: 3 Automations Every Accounting Firm Needs
Three practical AI automations for accounting firms — client document intake, reconciliation validation, and report generation — with real costs, real ROI, and production architecture on AWS.
API Rate Limits Will Wreck Your AI Project (Here's How to Handle Them)
Rate limits will break your AI automation if you're not prepared. Here's the production architecture — exponential backoff, token buckets, and queue patterns — that keeps your system running.
The $5,000 Automation That Replaced a $60,000 Process
A real case study: how a $5,000 AWS automation replaced 700 hours/year of manual invoice processing — with an ROI of 1,190% in year one.
Terraform for Non-Engineers: Why Your AI System Needs Infrastructure-as-Code
Your AI automation isn't just code — it's infrastructure. Here's why Terraform matters and how Infrastructure-as-Code keeps your AI systems running.
How to Evaluate an AI Consultant (Without Being Technical)
Seven questions every business owner should ask before hiring an AI consultant — from success metrics and pricing models to handoff plans and red flags.
The Real Cost of "Free" AI Tools for Your Business
No-code AI platforms sound great until you hit rate limits, vendor lock-in, and zero observability. Here's what "free" actually costs — and what to build instead.
My Claude-Assisted Development Workflow
How a solo consultant delivers production-grade automations in days instead of weeks — using AI-assisted coding backed by a decade of SRE discipline.
5 Processes Every Small Business Should Automate in 2026
Invoice processing, support triage, report generation, onboarding, and inventory alerts — five high-ROI automation targets with real architecture patterns and expected savings.
Why Your AI Automation Breaks at 2am (And How to Fix It)
Most AI automations are built like demos: they work perfectly during the presentation, then silently fail in production. The difference isn't intelligence — it's operational discipline.