The AI GTM Operating System

Most revenue organizations run on lagging metrics. I built one that runs on leading signals.

An AI operating system that converts product signals, buyer intent, and customer behavior into 90-day forward visibility across churn, expansion, and the full customer journey. Designed and proven inside a live GTM organization.

See the operating system Who I am

46%Renewal-rate lift

60%Support cost cut

90-180dRisk seen early

What you're looking at.

The AI GTM Operating System is a framework designed and built inside a live GTM organization. This site shares it completely: the architecture behind it, the results it produced, the workflows running today, and the governance model that sustains it at scale. Every section is grounded in something that was actually built. The person who built it is at the bottom.

The Prediction Loop · SOAR

Prediction is the difference between managing churn and preventing it.

AI is not a chatbot bolted onto your funnel. It is a loop that runs continuously across the customer journey: it surfaces the signal before the number moves, orchestrates the right motion, acts through people and agents, and refines from every outcome. That loop is the engine inside the operating system.

01 / SIGNAL

Signal

Leading-indicator health and engagement scoring surfaces adoption risk and expansion opportunity long before the renewal conversation.

02 / ORCHESTRATE

Orchestrate

The signal routes work to the right layer: a human, an agent, or a digital experience, matched to account complexity and value.

03 / ACT

Act

Intervention happens before the customer goes quiet. Automated where it should be, human where it matters.

04 / REFINE

Refine

Every outcome sharpens the model. The system gets more predictive with each cycle, compounding instead of resetting.

Layer 01

Humans

Senior judgment aimed where it changes the outcome: complex accounts, executive relationships, the moments a model can flag but not close.

Layer 02

Agents

Agentic workflows across onboarding, adoption, support, and review generation, delivering the motion at a scale headcount never could.

Layer 03

Digital

Self-serve and product-led experiences that carry the low-complexity journeys end to end, freeing the other two layers for what they do best.

Built, not theorized

I've built the working model of this operating system for a GTM organization and now I want to share it.

46%

Improvement in renewal rates by treating churn as a product, marketing, and success problem. Not a revenue one.

$2.7M

Operating savings from an AI support and guidance layer that automated triage, routing, and resolution.

$4M

New services revenue stood up from zero, with a 95% renewal rate for participating customers.

0→8

A predictive customer-health model giving a leading-indicator view of risk 90 to 180 days ahead of renewal.

Renewal rate · quarterly+46% lift

The predictor

The Pulse Score

A score without underlying data is just a guess. Building a truly predictive model requires investing in the signal architecture beneath it first: product engagement telemetry, customer interaction data, buyer demand signals, and renewal-context metadata, normalized and fused into a composite model. This is why agent building and data infrastructure are the same project. The score is the output. The architecture is the point.

Signal architecture · three lenses

Each lens reads High or Low. Eight combinations. One named diagnosis per account.

Layer 1 · Product-Led

Are they using it?

Triggers the flag · primary

Logins / visitor count
Review generation rate
Ads subscription
Profile completeness
CTA activation
Performance Analytics
Integration depth
Feature discovery breadth

Layer 2 · CSM-Led

Is the relationship alive?

Triggers the flag · secondary

Conversation counts
Meeting acceptance rate
Multi-threading depth
Executive engagement
NPS trend
Renewal-call sentiment
Overall call sentiment

Layer 3 · ROI & Context

Are buyers responding?

Context only · no independent flag

Leads generated from G2
Pipeline influenced / attributed
Review velocity trend
Category share of voice
AEO visibility & presence
G2 profile traffic
Referral traffic to customer site

Interpretation matrix · diagnosis, not just a flag

ProductCSMROIRenewalProfile / action

HighHighHigh93%Healthy: monitor

HighLowHigh86%Self-sufficient: light touch

HighHighLow81%Engaged, no ROI: investigate value

LowHighHigh77%CSM-carried + ROI: coach adoption

LowHighLow75%CSM-carried, fragile: escalate

HighLowLow75%Silent self-serve: monitor

LowLowHigh67%Passive value: likely safe

LowLowLow48%True at-risk: top priority

Renewal falls cleanly from 93% to 48% across the eight profiles. V1's single flag becomes eight actionable diagnoses.

83%

Churn caught, recall unchanged

1 in 4

False alarms removed

~$8M

ACV no longer chased unnecessarily

0 → 8

Same scale, eight named diagnoses

Health interpretation

A score that explains itself.

The 0-to-8 scale stays. What changes is the diagnosis. An account at 3/8 used to mean at risk, act. Now it carries a named profile: is product engagement low while no buyer demand exists, or is the CSM relationship the only thread holding? Those are different conversations, different urgencies, different next steps.

0 · CHURN RISKALERT AT 58 · HEALTHY

The missing signal

Customer words predict what usage data cannot.

A customer who logs in every day but tells their team the product is not working is a churn risk. A customer who barely logs in but calls their CSM a strategic partner will renew. Behavioral signals alone cannot distinguish those two accounts.

Gong call transcripts, NPS verbatims, support ticket language, and G2 review text carry the actual words customers use to describe their experience. Combined with product and ROI signals, they turn a number into a diagnosis: not just whether an account is at risk, but why, and what the conversation needs to be.

This is where the roadmap goes next: renewal-call sentiment, meeting acceptance trends, and multi-threading depth wired directly into the score, so prediction sharpens with every customer interaction.

GTM Engineering · intake to deployment

Ideas from the field, engineered for production.

Every agent here went through the same door: a structured intake, a priority triage, an engineering build, a maintained deployment. This is what professional AI operations looks like. Not individuals running scripts from their own computers.

Intake pipeline

Submitted

In review

Approved

In progress

Live

Declined

Where agents deploy · customer journey

Prospecting

agents

Qualification

Building

Discovery

agents

Demo & Proposal

Building

Onboarding

agents

Adoption

Building

Expansion

Building

Renewal

agent

21%

Submission-to-production rate

14 of 67 ideas became maintained, production-grade agents. The rest were triaged, deprioritized, or declined. Governance is the filter.

P0 to P3

Priority scoring on every idea

Every submission is scored by impact, complexity, and risk before engineering touches it. Ideas from the field. Accountability from the team.

~46 hrs

Returned to sellers per week

Across 6 roles with named champions. Time freed from manual work, redirected to the moments that need human judgment.

Submissions across 6 roles · anyone can submit, not everyone ships

AEAMBDRCSMSCIC

⚡

BDR Weekly Pipeline

ProspectingBDR

Weekly scheduled pipeline that runs for every active BDR. Pulls each rep's Salesforce book of business, scores accounts with a composite model combining six SF engagement signals and G2 buyer intent, enriches top-ranked accounts with contacts via ZoomInfo (AMER) or Cognism (EMEA/APAC), assembles account intel from ZoomInfo scoops and G2 profiles, drafts personalized outreach sequences using Claude Haiku, and writes net-new contacts back to Salesforce.

How it works

Triggered by

ScheduledRuns every Monday AM on schedule. One full pipeline run per active BDR. Also triggerable manually.

Processing

3 stages1. Find and Score: pulls Salesforce accounts, scores with six SF engagement signals plus G2 intent, ranks into High/Medium/Low buckets. 2. Enrich Contacts: gap-fills with ZoomInfo (AMER) or Cognism (EMEA/APAC), fetches scoops and G2 profiles. 3. Build Outreach: feeds account intel into Claude Haiku to draft personalized sequences per contact.

Outputs

CSV + CRM writeRanked accounts with enriched contact data emailed to the BDR. Net-new contacts written back to Salesforce with AccountId, BDR as owner, LinkedIn URL, and ZoomInfo metadata.

🔁

Churn-Risk Alert & Audit Loop

RenewalAMCSM

Identifies at-risk customer accounts (pulse score 0 to 3), enriches them with Salesforce and ZoomInfo context, posts actionable churn-risk alerts to the relevant team channel, and audits for acknowledgment or non-response over a 4-day window. Bridges the gap until Catalyst v2 ships so CSMs and AMs can act on churn risk immediately.

How it works

Triggered by

ScheduledRecurring batch in n8n, likely daily. Pulls active accounts, filters to pulse score 0 to 3, fires the enrichment and alert sequence. Also triggerable manually.

Processing

3 stages1. Pull and Filter: queries Salesforce for active accounts at pulse 0 to 3, enriches with ZoomInfo for multi-threading contacts. 2. Alert: posts churn-risk alert to team channel with account name, risk level, renewal window, AM/CSM ownership, and 5 recommended contacts. 3. Audit Loop: checks acknowledgment after 4 days, re-nudges or closes based on response.

Outputs

Alert + audit trailChurn-risk alert in team channel with contact quality assessment and multi-threading recommendations. 4-day audit check creates an acknowledgment record. AM/CSM confirms via reply or emoji.

🎯

MEDDPICC Research & Salesforce Updater

DiscoveryAEAM

Automates MEDDPICC field research and Salesforce updates for Account Managers by mining Gong call transcripts, Salesforce history, and account context. Eliminates the manual CRM hygiene work that follows every discovery call.

How it works

Triggered by

On-demand or scheduledAM messages the agent with account or deal details. Also fires weekly on a schedule to scan recent Gong calls across the book.

Processing

Signal extractionMines Gong transcripts for MEDDPICC signals, cross-references Salesforce opportunity history, and uses Claude to structure findings into each MEDDPICC field.

Outputs

Updated Salesforce fieldsMEDDPICC fields written back to the Salesforce opportunity. Summary posted to Slack so the rep sees exactly what changed and why.

Time saved per week · by role~46 hrs total

14 hrs

10 hrs

BDR

8 hrs

CSM

7 hrs

5 hrs

2 hrs

🤝

SC to ICon Handoff Generator

OnboardingSCIC

Automatically pulls presales context from Salesforce, Gong, and the SC Request object for newly-closed deals and assembles a structured handoff brief for the implementation consultant. No manual knowledge transfer, no lost context at the moment it matters most.

How it works

Triggered by

Scheduled dailyScans Salesforce Onboarding records created in the prior 24 hours for newly-closed deals that need a handoff brief.

Processing

Context assemblyPulls SC Request object, Gong call summaries, and Salesforce deal history. Structures into a handoff format covering technical requirements, stakeholders, and agreed success criteria.

Outputs

Handoff brief in SlackStructured brief delivered to the IC in Slack at deal close. Presales context preserved without a single manual handoff meeting.

Governance

The system that governs the system.

Scaling AI across a GTM org does not happen through enthusiasm. It happens through a repeatable intake model: one front door, one priority framework, one engineering team that builds and maintains what the field submits. Three phases make it work.

01 / ACCESS

Open the door.

Give every role a structured way to submit ideas. Without a front door, AI work happens in shadow corners: undocumented, single-owner, invisible to leadership until something breaks.

Access is not just permissions. It is a cultural signal that the org is building this together, and that the field is the best source of what to build next.

02 / INVENTORY

Catalog what exists.

Before you scale, you need to know what has already been built. Most organizations have more automation than they think, and most of it is fragile: one person owns it, no one else can maintain it, and it was never designed to scale.

The inventory surfaces shadow IT, informal automations, and one-off experiments before they become technical debt. It is how you prevent the same thing from being built six times by six different people.

03 / GOVERN

Rebuild it to last.

Not every submission should be built. Not every informal automation should survive review. The governance layer triages by priority and risk tier, assigns ownership, and rebuilds approved workflows in a maintained, QA'd, champion-supported form.

The output is a catalog the whole org can trust: tested in the field, documented, and handed to a champion who owns it going forward.

The revenue architect.

VP, Global Solutions & Customer Success · G2

I started as a software engineer. Twenty years later, I'm still the most technical voice in the executive room, fluent enough to partner directly with product and engineering, commercial enough to own the number.

That combination is the whole point. I build and re-architect post-sales organizations around adoption, outcomes, and AI, moving teams up the customer lifecycle so retention, expansion, and new revenue compound instead of getting chased downstream.

At G2 I designed an AI operating system across the customer journey, human-led and agent-delivered. It rebuilt a broken retention motion, lifted renewal rates 46%, cut support operating costs 60%, and stood up new services revenue from zero. Before that I built functions from the ground up at Slack, through the Salesforce $27B acquisition, and scaled a Workday practice from a $10M acquisition toward a $120M global business.

2024 to NOWG2 · VP, Global Solutions & Customer SuccessArchitected the AI GTM Operating System; built Professional Services from inception; co-launched G2's commercial MCP strategy.

2022 to 2024Upwork · VP, Enterprise SolutionsBuilt the enterprise solution and post-sales model; contributed to 37% enterprise growth.

2020 to 2022Slack (Salesforce) · Professional Services LeaderBuilt services and technical consulting from the ground up; named Slack Leader of the Year, 2022.

2012 to 2020Alight / Strada · VP, Professional ServicesScaled a Workday HCM practice from ~$10M toward ~$120M under private-equity ownership.

2007 to 2012IBM · Managing Consultant, Global Business ServicesLed global HR transformation and HCM technology deployments across enterprise clients. Built technical depth in integration strategy, ETL design, and large-scale system implementation.

LinkedIn ↗GitHub ↗

Let's build it

If your GTM finds out too late, that's a system problem.

I work with enterprise B2B SaaS teams putting AI at the center of go-to-market as a predictive system, not a pile of point tools.

Start a conversation Connect on LinkedIn