An agentic content system with a human at the gate
When I joined Maven AGI, content ran on daily posts and one-off builds. I replaced that with one thesis, four themes a year, and an agentic AI system to run it: 17 specialized skills orchestrated as a single agent that researches, scores ideas, drafts, and publishes. Every piece is held to the same brand voice, claims rules, and quality bar, so the content engine scales without drifting off brand.
The agent does the legwork. I own the strategy, the edit, and the final yes. Nothing is scheduled, published, or sent without my approval.
Seven layers, one loop
One thesis goes in; published, tracked content comes out. Each layer filters what reaches the next, and results flow back up to rewrite the foundation.
L1
Strategy foundation
Thesis, theme calendar, audience and brand-voice files, SEO targets, and a claims register that every stat must clear.
L2
Signals
A news monitor with a strict 7-day window, a Slack scanner for wins and launches, trend research, and internal customer data. Signals stay only if they reinforce the current theme.
L3
Planning
A weekly
autonomous-content-agentchains research into scored briefs.idea-scorerrates fit, audience pain, differentiation, and SEO. No connection, no slot.L4
Quality gates
Connection, claims clearance, voice rules, and my approval. Nothing ships until it clears all four.
L5
Creation
Skills like
blog-writerdraft what the editorial calendar calls for: deep blog posts, white papers, ebooks, guides, LinkedIn posts, the biweekly newsletter, sales enablement, PR pitches, executive bylines, and more.L6
Distribution
Webflow CMS, Typefully for LinkedIn scheduling, earned media and analyst briefings, with Airtable as the system of record.
L7
Learning
performance-reviewerreads results and rewrites the foundation files with what worked. North-star metrics: share of voice, inbound press and analyst interest, branded search growth.
↺ and round again: L7 feeds L1
The weekly loop
Every week runs the same ten steps, in order. The agent handles seven. The three that need judgment are mine. Hover or tap a step for detail.
One week, ten steps
Seven run on the agent. Three need my judgment: the edit, the approval, and the last mile.
Guardrails I designed
- ConnectionEvery piece ties to the thesis, the active theme, a keyword cluster, or category vocabulary.
- ClearanceEvery stat is checked against a four-state claims register. Accurate and cleared are separate checks, and anonymizing a customer never substitutes for clearance.
- VoiceConfident and enterprise-credible, never hype. A banned-word list and a list of AI-writing tells to cut.
- ApprovalReading, research, and drafting run freely. Publishing, sending, and scheduling wait for my explicit yes.
Plugged into the team's tools
Most marketing tools weren't built for AI. Through MCP connectors, the agent works inside them directly: publishing posts with SEO metadata, cutting video by transcript, scheduling social. The approved version is the published version, with no copy-paste between tabs.
I built it as a non-engineer, writing and pushing code within eight weeks of marketing getting GitHub access. It cut production overhead and became the company's reference example for agentic marketing. Leadership highlighted it to the full organization, noting I'd been more productive here than in a prior role with a seven-person team.
The writing
Samples from three roles. The newest, at Maven AGI, were made with the help of the system above. Everything from Workhuman and Cognex came before it, and shows the craft the system is built on.
the craft came first ↓Maven AGI
Content Marketing Lead → Corporate Marketing
Made with the systemWriting for an enterprise AI company that's still defining its category. Each piece has a job: earn a point of view buyers remember, carry the CEO's voice, or launch a product in plain language. Most of it moves through the content system above.
Services-as-Software: “Don't buy seats, buy outcomes.”
The company's thesis, turned into a full campaign. Software is moving from a tool that helps with the work to a service that does the work, and buyers need new criteria to judge it. I built every piece, from the CEO's Fast Company byline that introduced the idea to the gated guide that turns it into a buying checklist, plus a six-week cross-channel plan to carry it.
See the landing page ↗- Fast Company byline“Your CRM Was Built Backwards,” CEO Jonathan Corbin's essay for the Fast Company Impact Council, published October 2. Read it ↗
- Landing pageThe argument in five moves: pricing, memory, trust, and what to ask vendors.
- The Agentic AI Enterprise Buyer's GuideThe gated second edition, adapted from the CEO's Services-as-Software manifesto: questions, tests, and scorecards for buying the work, not the tool.
- Gate and email campaignThe capture form and the nurture emails that follow a download.
Two Years After Klarna's AI Reversal
Rereads the most-cited AI customer service story. The AI did what it was measured on, and containment was the wrong measure. The piece argues for outcomes over volume.
Read on mavenagi.com ↗ Executive byline · Fast CompanyYour CRM Was Built Backwards
Written for CEO Jonathan Corbin, drawing on his years leading customer success at HubSpot. It argues that the CRM funnel mirrors the org chart, not the customer, and that AI agents can finally hold the whole relationship.
Read in Fast Company ↗ Search-led editorial78% Have AI Support Pilots. Only 14% Reach Production.
Takes a fresh survey stat and explains the 64-point gap. Pilots are built for the demo, and production demands a different architecture.
Read on mavenagi.com ↗ Trust and positioningWhy AI Governance Is Becoming a Competitive Advantage
Turns a compliance topic into a buying argument. It opens on a timely cultural moment and lands on the questions enterprise buyers now ask every AI vendor.
Read on mavenagi.com ↗ Product launchResolution Is Only the Starting Point
Launches Charters by reframing the category's benchmark. Resolution is the floor, and the support conversation is where upsell and retention are won.
Read on mavenagi.com ↗ Product launchIntroducing Intelligent Fields
Opens on the buyer's pain: the transcript exists but the data doesn't. It explains why manual tagging fails at scale before introducing the feature.
Read on mavenagi.com ↗Use case pages, from template to launch
I designed one page template, wrote the copy for all nine agents, and got every page live on mavenagi.com. Each page follows the same structure, so buyers can compare use cases at a glance:
- The problem, stated in the buyer's terms
- What the agent handles
- Where each skill works
- What changes when it's live
- The call to action
- Awareness & EngagementThat turns anonymous traffic into pipeline
- Product DiscoveryThat ends the comparison stall
- Inbound SalesThat catches the signals your forms never see
- Onboarding & AdoptionThat closes the gap between signup and proven value
- Customer SupportThat can take real action, and actually resolve tickets
- Churn PreventionThat acts before the cancel button
- Upsell & Cross-SellThat hears the signal your team missed
- Advocacy & LoyaltyThat asks while they're still happy
- Agent CopilotThat makes every rep sound like your best one
- The Agentic List 2026The highest-traffic content page on the site during my first six months.
- Deflection vs. Resolution: The Metric That Decides Whether AI Customer Service WorksDefines the category's core metric and explains why automation rates hide what customers experience.
- What's Changed One Year Since Gartner's "80% AI Resolution" PredictionA one-year check on a headline prediction: what held, what didn't, and what it means for 2026 planning.
- What "Deterministic Control" Actually Means for AI AgentsPins down a governance term before it becomes marketing shorthand.
- Relevance Is Not Permission: The AI Access Problem No One Talks AboutUses a late food delivery to show why AI agents on multi-sided platforms must check eligibility before retrieval.
- How AI Agents Are Transforming Healthcare SupportAn industry piece on resolving patient and provider questions while keeping trust and HIPAA in view.
Workhuman
Director, Growth Content · Sr. Manager, Customer Communications
Before the systemContent for HR buyers at every stage of a deal: a business case an HR leader can take to the CFO, a point of view for a tough economy, a customer story for late-stage deals, and talk tracks for reps on the conference floor.
Merck: Culturally Energizing, Global Social Recognition
How Merck replaced paper, top-down awards with peer recognition launched in 85 countries on one day. The story runs from challenge to proof: 3M recognition moments, 100% reach, and a 12-point lift in engagement scores.
Open PDF ↗ White paper · 14 pages3 Ways Recognition Combats Economic Uncertainty
A business case HR leaders can take to the CFO. It pairs Gallup and Korn Ferry research with customer turnover and safety data, then models a 6.9x ROI on a $150-per-employee investment.
Open PDF ↗ Thought leadership · eBookPrioritizing DEI During Times of Economic Uncertainty
Argues against cutting DEI in a downturn, using DDI, Deloitte, and Glassdoor research and a conference voice from Akamai to show the cost in retention, innovation, and hiring.
Open PDF ↗ Demand gen · Guide3 Creative Ways to Fund Your Social Recognition Program
A short, practical guide for buyers who believe in recognition but lack budget: find hidden ad hoc spend, consolidate reward programs, and rethink the bonus pool.
Open PDF ↗ Sales enablementWorkhuman Live: Sales Cheat Sheet
Talk tracks for reps at the flagship customer conference, covering a new product, a media partnership, DEI, the mid-market offer, and how to answer economic worries.
Open PDF ↗Cognex
Sr. Manager, Head of Internal Communications and Events
Before the systemInternal communications for a global engineering company. The audience was employees, and the story was how their work connected to the business. Much of it is written in a leader's voice, from an org change to a note on productivity. Names, products, and customers are redacted.
Engineering Organization Announcement
Written for two engineering leaders. It opens with the strategy, a common software framework, and only then explains which teams move where, so the change reads as a reason and not a reshuffle.
Open document ↗ Team newsletterSome Assembly Required
A newsletter for a global hardware engineering team, themed like a product manual. It combines a leader's note, product milestones, a tooling change explained in plain terms, and team thank-yous.
Open document ↗ Executive voicePlanning for Productivity
An EVP's note to managers as summer Fridays ended: cut outdated work, run smarter meetings, and protect time for learning. Written in his voice and tied to company programs and culture.
Open document ↗ StrategyBusiness Unit Comms Strategy
A communications plan for a business unit: audiences, channels, cadence, and measures.
Open deck ↗