AEO/GEO · Machine-Readable Web · Fractional CTO · Principal SRE
ChatGPT, Claude, Perplexity, and the agents they power are the new front door to your business. I don't just optimize for answer engines — I build the machine-readable layer they consume (llms.txt, agents.md, JSON-LD, AI catalogs, MCP), and run it in production across my own portfolio. Systems that hold.
flagship · aeo / geo
Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) decide whether AI recommends your business — or a competitor's. I don't just optimize for answer engines: I build the machine-readable layer they consume, and run it in production across my own portfolio. This page is the demo — llms.txt, agents.md, ai-catalog.json, and JSON-LD, all live on the site you're reading.
A fixed-price teardown of your answer-engine presence: how the major engines describe you today, citation and schema gaps, llms.txt / agents.md / AI-catalog coverage — ending in a scored, sequenced fix list you own.
fixed price · scored report · ~2 weeksA retainer that builds and operates your machine-readable layer: llms.txt, agents.md, JSON-LD, .well-known/ai-catalog.json, and MCP surfaces — kept current as engines, specs, and your business change.
retainer · the layer engines consumeOnce agents can find and transact with you, govern them. The Agent Governance Readiness Audit and EU AI Act Readiness Assessment extend the same machine-readable rigor to your agent fleet and your regulators.
enterprise · audits + evidence pipelineA free scanner that grades any domain's answer-engine readiness — llms.txt, agents.md, JSON-LD, AI catalog, MCP discovery — in about a minute. The audit above is the human-grade version.
why now
Enterprise AI-agent adoption jumped from 11% to 42% of organizations in two quarters — but almost nobody can govern what they deployed. The gap between an agent pilot and a production system you can trust is exactly where I work.
services
I don't recommend architectures — I operate them. Every offering below is a pattern running in my own production infrastructure before it's ever proposed to a client.
The AEO/GEO ladder starts here. Offerings built on infrastructure I already run in production — agent-readable surfaces on this very site, signed agent identity, MCP servers behind CDN and WAF, multi-agent meshes with human-in-the-loop gates. Most consultancies talk about this layer; I ship it.
The flagship ladder, delivered: a fixed-price AEO/GEO Audit, then a Machine-Readability Implementation retainer — llms.txt, agents.md, JSON-LD, AI catalogs, and MCP surfaces — so ChatGPT, Claude, Perplexity, and the agents they power find, cite, and transact with you first.
audit → retainer · this site runs itDesign, build, and migrate Model Context Protocol servers — including migration sprints to the 2026 stateless/serverless spec with OAuth hardening. I operate production MCP endpoints daily.
mcp.username.md · lambda + wafCryptographically signed, third-party-verifiable identity for humans and AI agents: did:web, verifiable credentials, RFC 9421 signed responses, DNS verification. Live in my production today.
rfc 9421 · did:web · running nowA named audit of your agent fleet: identity, cost ceilings, kill switches, human-in-the-loop gates, observability, and incident response — before an autonomous agent bankrupts a budget or walks past a control. The enterprise rung of the flagship ladder.
96% run agents · 12% can govern themFixed-scope assessment against the regulation now in force: system classification, Article 50 transparency gaps, provider-vs-deployer determination, and the evidence pipeline for the December 2027 high-risk deadline — built alongside your counsel. Details & timeline ↓
transparency rules live since aug 2026Agent kill-switch and cost-ceiling review, model-distillation defense, and AI-assistant attack-surface hardening — from a practitioner with a bug-bounty pedigree.
assume agents will be exploitedSenior technical leadership and production-grade reliability — the foundation everything else stands on.
Senior technical leadership and roadmap ownership at a fraction of an executive hire — measured in shipped milestones and avoided rewrites.
retainer · strategy → executionSLOs, incident response, observability, and capacity engineering that convert outages into single-digit-minute non-events.
production readiness · on-call sanityA principal-level teardown of your stack — integrations, data flows, failure modes — with a sequenced remediation plan you can execute without me.
fixed scope · 2–3 weeksMake AI carry real operational load — visible, governed, and paying for itself.
LLM-driven runbooks, automation, and agent-assisted incident response that cut toil hours so your engineers ship instead of babysit.
I run my company this wayTracing, evals, and priority-aware model routing so every AI dollar and every hallucination is visible — engagements in this space routinely halve LLM API spend.
langfuse-grade tracing · routingOpen-weight frontier-class models on your own hardware for privacy-bound and data-sovereign organizations — law firms, non-profits, institutions.
on-prem proven · sovereignty firstreadiness assessments
Fixed-scope, fixed-price assessments that end in a scored findings report and a sequenced remediation plan you own. Engineering readiness, not legal advice — each assessment is built to work alongside your counsel, not replace them.
The EU AI Act's transparency obligations (Article 50) and penalty powers took effect on August 2, 2026; standalone high-risk (Annex III) obligations follow on December 2, 2027 after the Digital Omnibus deferral. This assessment maps your AI systems to the Act's risk tiers, determines provider-vs-deployer roles, checks Article 50 disclosure and AI-content marking gaps, and stands up the technical documentation and logging pipeline the high-risk deadline will demand — so counsel's advice has engineering underneath it.
two clocks: aug 2026 live · dec 2027 nextThe enterprise agent-fleet audit — identity, cost ceilings, kill switches, human-in-the-loop gates, observability, incident response. The operational substrate that AI-conduct regulation assumes you already have.
96% run agents · 12% can govern themKill-switch and cost-ceiling review, model-distillation defense, and AI-assistant attack-surface hardening — the security posture regulators and insurers increasingly ask to see in writing.
assume agents will be exploitedReferences: Regulation (EU) 2024/1689 (EUR-Lex) · European Commission: AI regulatory framework · Implementation timeline (artificialintelligenceact.eu) · Digital Omnibus deadline analysis (Gibson Dunn) · AI & Agentic Web Glossary
Chris Bergeron Consulting provides technical and operational readiness services, not legal advice. Regulatory interpretations should be confirmed with qualified counsel; we routinely work alongside our clients' law firms.
proof, not slideware
My own holding company runs AI-native. These aren't case studies borrowed from a vendor deck — they're systems I operate, page for, and answer to.
Multi-agent orchestration on a NATS event bus with human-in-the-loop approval gates and PagerDuty escalation.
Production MCP server on Lambda + API Gateway + CloudFront + WAF, serving agent handshakes publicly.
RFC 9421 signed HTTP responses, did:web, verifiable credentials, and DNS-verified profiles — live on the username.md platform.
Langfuse tracing, evals, and cost telemetry across every agent and automation in the fleet.
Claude-driven runbooks, log watchers, auto-remediation drafts, and nightly audits keeping a 16-stack fleet honest.
Prototyped the world's first in-dash automotive infotainment system — Slashdot-featured, printed in technology books.
about
I'm Chris Bergeron — serial founder, Fractional CTO, Principal Site Reliability Engineer, and Systems Architect. Systems integrator for recreation. I've spent 25+ years across the whole stack of a technology career: help desk, networking, systems administration, consulting, and principal-level SRE.
I've served the International Monetary Fund, law firms, non-profits, and grocery & manufacturing organizations. I've worked inside large corporations and mom-and-pop shops, and I speak both languages: governance and risk for the boardroom, terminals and trade-offs for the engineers.
Today I run The Holding Company as a fully AI-native operation — agents in production, signed machine identity, self-hosted AI infrastructure. When I advise on the agentic era, it's from operational evidence, not analyst reports.
how we work
Fixed-scope, fixed-price teardown — AEO/GEO, architecture, reliability, agent governance, or AI security — ending in a sequenced plan you own.
2–3 weeks · report + working sessionA focused build: an MCP migration, an observability stack, a discoverability package, a compliance-documentation pipeline. Shipped, not advised.
2–6 weeks · working softwareFractional CTO or standing SRE counsel — a principal on your side of the table for roadmaps, hires, vendors, and incidents.
monthly · limited seatsOne working session. Bring your architecture, your agent pilot, or your incident history — leave with a straight assessment and a sequenced next step.
consult@chrisbergeron.com