AI projects that make it to production.
Most AI initiatives die between the demo and the deploy—because nobody in the room is accountable for production. I've been the CTO. I scope the opportunity, ship the pilot in your infrastructure, and stay accountable for what it costs and how it behaves at scale.
50+
Products shipped
25+
Years leading engineering teams
72%
LLM cost cut in one engagement
6 wks
PoC to production, most recent AI build
The industry has a pilot problem.
The pattern is everywhere: a strategy firm delivers a deck, an agency builds a demo, the demo impresses everyone, and then it quietly dies. It never had evals, nobody modeled the cost at scale, security wasn't consulted, and no one owned getting it into the systems your business actually runs on.
That's not an AI problem. It's an accountability problem. The fix is one person who owns the recommendation and the production system it becomes—which is the job I've done for 25 years.
How this practice is different
- Fixed-fee, fixed-scope entry points—no open-ended discovery
- Pilots ship in your infrastructure, not a sandbox
- Evals, cost models, and guardrails are deliverables, not add-ons
- One accountable practitioner, not a rotating bench
Three ways in. Each one de-risks the next.
Start small and prove value, or step straight into ongoing AI leadership—every engagement is scoped so you know exactly what you're buying.
AI Opportunity Audit
1–2 weeks · Fixed fee
A ranked map of where AI actually pays off in your business—and where it doesn't. I review your workflows, data, and systems, then hand you a board-ready report.
- Ranked AI use cases with ROI estimates
- Build vs. buy vs. wrap call for each
- Data readiness and integration assessment
- A 90-day roadmap you can execute with or without me
AI Pilot Sprint
4–6 weeks · Fixed scope
One use case from the audit, shipped as a working system in your infrastructure—not a demo in mine. Evals, cost model, and guardrails included from day one.
- A production-grade pilot in your environment
- Eval suite that gates every change
- Per-request cost model at your projected scale
- Security review and human-in-the-loop guardrails
Fractional AI Leadership
Ongoing · 1–2 days/week
A fractional Chief AI Officer seat. I own the AI roadmap, model and vendor decisions, team enablement, and governance—embedded in your leadership meetings.
- AI roadmap owned and defended to the board
- Model selection and vendor negotiation
- Your engineers trained on AI-native development
- Governance, evals, and cost kept under control
For investors & boards: AI Due Diligence
Buying or backing a company that claims AI? I assess whether the AI is real—model dependencies, eval maturity, unit economics of LLM features, and the risks hiding in the context window. Fixed fee per target, built on my technical due diligence practice.
What I build
The same systems I write about, shipped for clients—agents, RAG, voice, and AI-native products.
Agentic Systems
Multi-step AI agents that take real actions—booking meetings, filing tickets, running analyses across your tools. Orchestration, error handling, and human-in-the-loop guardrails.
RAG & Knowledge Systems
Internal search over your docs, Slack, codebase, and databases. Hybrid vector + keyword retrieval with reranking—real answers, not hallucinated ones.
Voice & Conversational AI
Real-time voice agents for support, sales qualification, and scheduling. Latency, telephony integration, and human handoff architecture handled.
AI-Native Products
Products where AI is the core value—not a feature bolted on. Build a moat, not a thin API wrapper.
Computer Vision & Multimodal
Document extraction, quality inspection, video understanding. Vision models replacing legacy OCR and manual review.
AI-Augmented Development
Coding agents, automated testing, and PR review integrated into your CI/CD pipeline. Your team ships faster with AI in the workflow.
Evals Before Vibes
Every AI feature ships behind an eval suite. If we can't measure it, we don't launch it—and the board gets numbers, not demos.
Cost as Architecture
A feature that costs $0.02 per request at 100 users costs $60K/month at scale. Caching, prompt optimization, and model routing are designed in, not bolted on.
Security & Compliance
Prompt injection, data leakage, PII in context windows, SOC 2 and HIPAA implications. Solvable—but only with deliberate design.
Receipts, not promises.
The best vetting question for any AI consultant: “show me what you shipped recently.” Here's mine.
PoC to production in six weeks
How a stalled AI proof-of-concept became a production system with evals, guardrails, and a cost model that held at scale.
Read the case study →Case studyCutting an LLM bill by 72%
Model routing, caching, and prompt engineering that turned a runaway AI cost curve into a sustainable unit economic.
Read the case study →Plus 25+ published deep dives on agents, RAG, evals, and AI cost architecture in the AI Strategy library →
Questions buyers actually ask
Do you advise or actually build?
Both—and that's the point. Strategy-only AI consulting is why most pilots die. I write the plan, then I'm accountable for shipping it in your infrastructure with your team.
We don't have a data team. Can we still do this?
Usually, yes. Most high-ROI AI use cases in 2026 run on the data you already have in your SaaS tools and databases. The audit tells you honestly if you're not ready—before you spend pilot money.
How do you price?
The audit and pilot are fixed-fee with a defined scope, so there's no meter running. Ongoing leadership is a monthly retainer scaled to 1–2 days per week. We'll put exact numbers on the intro call once I understand your scope.
What stack do you use?
Whatever earns its place: frontier models via API (Claude, GPT, Gemini), open-weight models where cost or privacy demands it, and boring, proven infrastructure underneath. I'll tell you plainly what I'm proposing and why—no black boxes.
What about AI due diligence?
For investors, acquirers, and boards: I run technical due diligence with an AI-specific lens—model dependencies, eval maturity, unit economics of LLM features, and whether the "AI" is real. Fixed fee per target.
Ready to get AI out of pilot purgatory?
Grab 30 minutes. We'll figure out whether an audit, a pilot, or ongoing AI leadership is the right first step—or whether you should wait.
Book a free intro call →