💬 What this is: We analysed the GTM stacks and go-to-market behaviour of the 100 fastest-growing private B2B companies. We pulled from The Signal, Forbes Cloud 100 2025, Forbes AI 50 2025, CNBC Disruptor 50, and Enterprise Tech 30. We were looking for patterns that don't show up in standard GTM playbooks. Three appeared across every segment of the list.
01The meta-finding before the plays
The default GTM stack is now standardised across the top 100.
Salesforce (100%). Gong (84%). dbt (84%). Snowflake (79%). Outreach (71%). Clay (65%).
These companies are not competing on tools anymore. They are competing on what they build on top of them.
The teams that win have one thing the others don't. A GTM Engineer. Someone who thinks like a developer about revenue problems. Who builds instead of buys. Who treats pipeline as a systems problem.
📊 54% of the fastest-growing private B2B companies have a GTM Engineer on staff. Only 3 to 7% of all B2B SaaS companies do.
That gap is where all three plays live.
Play 1
Build Your AI SDR In-House. Then Know When to Kill It.
🔍 The pattern: Zero of the 100 companies in this dataset use an off-the-shelf AI SDR (11x, Artisan, Regie, AiSDR). The companies with engineering resources build their own with Clay + Claude + Outreach. The most advanced ones are now doing something even more interesting with it.
Why off-the-shelf AI SDR fails
When you buy a vendor AI SDR, you hand your ICP definition, your messaging, and your sequencing logic to a black box tuned for the median customer. Not yours.
The result is generic personalisation at scale. Your prospects have seen it. Reply rates tank. The tool gets blamed. The real problem is the approach.
What the fastest-growing companies build instead
They build a custom Clay workflow connected to their own data. Here is the actual stack:
- Clay: pulls from their warehouse, enriches with 75+ data sources, builds the lead list
- Claude: receives context about each prospect (tech stack, hiring signals, recent news, ICP fit score) and writes a message specific to that person, not a template with variables swapped in
- Outreach or Smartlead: executes delivery, tracks replies, feeds response data back into the warehouse
The key difference: the message is generated from context, not filled into a template. Claude reasons about what to say based on what it knows about the prospect.
The named examples
Ramp built OAT (Outbound Automation Team). At peak, OAT was generating ~30% of Ramp's total pipeline. Co-founder Gene Lee shut it down in October 2025, not because it stopped working, but because Ramp moved upmarket. When deal size crosses a threshold, fully automated outbound starts to damage the exact buyer relationships you need. Ramp distributed OAT's capabilities to AEs, CSMs, and RevOps instead.
Vercel collapsed a 10-person SDR team to one GTM Engineer in 6 weeks. One person built workflows that replicated the output of ten. (Source: Jeanne DeWitt Grosser, Vercel CRO, Lenny's Podcast 2026)
Verkada built an MDR (Marketing Development Rep) model. Automation handles cold outreach end-to-end. The MDR only engages once a prospect responds. Result: ~120 meetings per rep per month vs ~20 for a traditional SDR. 4x the efficiency.
How to build this yourself
Step 1: Build your signal-enriched lead list in Clay
In Clay, create a table that pulls from your ICP criteria (industry, headcount, tech stack, funding stage). Layer in at least 3 enrichment sources: ZoomInfo or Apollo for firmographics, LinkedIn for org signals, Clearbit or BuiltWith for technographics.
Add a column that calculates an ICP fit score (1 to 10) based on how many criteria are met. Only pass leads with a score of 7+ to the next step.
Step 2: Write a Claude prompt that generates context-aware messaging
In your Clay table, add a Claude column. The prompt should include:
- The prospect's role, company, tech stack, and recent signals
- Your ICP definition and what problem you solve
- 2 to 3 examples of messages that have worked historically
- Instructions to write one short, specific opening line (not a template)
The output should be a unique first line per prospect, not a personalised template.
Step 3: Set a deal-size threshold before deployment
Below $15K ACV: automated outbound can handle the full sequence.
$15K to $50K ACV: automated first touch, human follow-up after reply.
Above $50K ACV: human-written first touch, automation handles research and scheduling only.
Ramp's lesson is the lesson. Automation works until it doesn't fit the relationship your deal size requires.
⚡ The shortcut most teams miss: Connect your Gong or Chorus data to Clay. When a rep logs a loss reason, pull that signal back into the Clay table to exclude similar prospects or adjust the messaging for that segment. Vercel's GTM engineer built this. Their AI agents now review Gong calls to find actual loss reasons vs the rep-reported reason. The gap is often the missing personalisation signal.
Play 2
Run GTM on the Warehouse, Not the CRM.
🔍 The pattern: dbt appears in 84% of the fastest-growing B2B stacks. Snowflake in 79%. Both are as common as Gong. Neither is what most people think of as a sales tool.
Why CRM-as-database fails at scale
When Salesforce or HubSpot is your single source of truth, your GTM data is:
- Siloed from product usage data
- Siloed from billing and expansion signals
- Updated manually by reps (which means it's wrong ~40% of the time)
- Missing the full picture of who is engaging and how
You cannot build signal-based outbound (Play 3) on top of a CRM-as-database. The signals live outside the CRM. You need a place to unify them first.
What the warehouse model looks like
The fastest-growing companies build a unified data layer where everything converges:
- Product data (usage, feature adoption, time-to-value) via Segment or Rudderstack
- Billing data (ARR, expansion signals, churn risk) via direct database connection
- CRM data (activity, pipeline, deal stage) via Fivetran or Airbyte
- Third-party signals (intent, hiring, news) via Clay or Unify
- Website behaviour (pages visited, content consumed) via Clearbit Reveal or RB2B
All of this lands in the warehouse (Snowflake or BigQuery). dbt transforms it into clean, enriched models. Then reverse-ETL (Census or Hightouch) pushes the enriched view back into Salesforce so reps see the full picture without leaving their CRM.
📊 Rippling runs dbt, Snowflake, Fivetran, Airbyte, Census, Hightouch, and Segment simultaneously. Seven data tools in parallel. The warehouse is not the nice-to-have. It is the foundation.
The orchestration layer is where the real advantage lives
Top-performing companies run 3 to 5 automation platforms at the same time:
- Zapier: 54% adoption across the top 100
- Workato: 41%
- Make: 21%
- n8n: 19%
- Tray.io: 16%
Vercel runs all five. This is not tool bloat. Different workflows run on different orchestration layers based on data type and execution speed. Zapier handles simple triggers. Workato handles complex enterprise workflows. n8n and Make handle custom automation. Tray.io handles high-volume API orchestration.
The companies at the bottom of the list run one orchestration tool (usually Zapier) or none.
How to build this yourself
⚠️ You do not need to build this all at once. This is a phased build. Kyle Norton (CRO, Owner.com) calls data foundations Phase 1 of his AI GTM Roadmap: before automation, before AI, before any of the other plays.
Phase 1: Connect your data sources (Week 1 to 2)
Start with Fivetran or Airbyte (both have free tiers). Connect:
- Your CRM (Salesforce or HubSpot)
- Your product database (if applicable)
- Your billing system (Stripe, Chargebee, etc.)
Point all three at a Snowflake or BigQuery warehouse. This is your unified data layer.
Phase 2: Transform with dbt (Week 3 to 4)
Install dbt Core (free) or dbt Cloud ($50/seat/month). Write your first models:
dim_accounts: one row per account, all enrichment fields unifieddim_contacts: one row per contact, engagement history includedfct_signals: one row per signal event (hire, funding, product action, intent)
This is the foundation that makes Play 3 possible.
Phase 3: Activate via reverse-ETL (Week 5 to 6)
Install Census or Hightouch (both have free tiers). Build a sync that pushes your enriched dim_accounts model back into Salesforce as a custom field: ICP Fit Score, Last Signal Date, Signal Type.
Now every rep sees the enriched picture in the CRM they already use.
Phase 4: Add the orchestration layer (Ongoing)
Start with one orchestration tool. Make or n8n are the best starting point for GTM use cases. Build your first workflow: when fct_signals gets a new row with signal_type = 'new_hire_VP_Sales', push that account to a Clay table and trigger Play 3.
Play 3
Signal-Triggered Warm Outbound. Not ICP-List Cold Outbound.
🔍 The pattern: The fastest-growing companies aren't sending the same sequence to 5,000 ICP-fit companies. They're sending a precise message to the 50 accounts that just entered a specific moment of need. The list is not the filter. The trigger is.
Why ICP-list cold outbound is losing its edge
The problem is not volume. The problem is timing.
When you pull 5,000 ICP-fit companies and sequence them all, most of those companies are not in an active buying window. They don't have the budget approved. The champion hasn't been hired yet. The pain isn't live enough to act on.
Your open rates drop because the message is right but the timing is wrong.
Signal-based selling solves the timing problem. You wait for the specific event that tells you a buying window just opened, then hit it within 24 hours.
The 5 signals that drive the highest performance
Based on what the companies in this dataset have documented publicly:
1. New hire in a buyer role
When a company hires a new VP Sales, Head of Revenue Operations, or GTM Lead, there is a 60 to 90 day window where they are evaluating tools and not locked into existing vendor relationships.
Rippling triggers Clay outreach when a company hires a new HR Manager or IT Admin. That sequence runs at a 60% open rate and 10% reply rate. (Source: Clay case study, Noah Adelstein, Manager International Growth)
2. Job change: existing champion moves companies
When a contact who knows your product moves to a new company, they are your warmest possible prospect. They have already sold themselves. UserGems and Common Room both track this signal.
Apollo's analysis: job-change triggers produced 4x more meetings than standard ICP sequencing.
3. Funding event
Series A or B funding means the company just got budget to spend and is actively building out their GTM stack. The window is 30 to 60 days post-announcement.
4. Intent signal: pricing page visit or competitor research
When a prospect visits your pricing page or is researching your competitor category on G2 or Capterra, they are in an active evaluation. This is the hottest signal available.
Combined with a job-change signal, Apollo found this produced the highest meeting rate in the dataset.
5. Product or behavioural trigger
For PLG companies: when a free user hits a usage threshold, invites a third team member, or uses a feature that correlates with paid conversion, that is a trigger.
How to build signal-triggered outbound in Clay
Step 1: Pick your 2 to 3 highest-intent signals
Do not try to run all five at once. Start with the signal that historically correlates most closely with your closed-won accounts.
Pull your last 50 closed-won accounts. Ask: what happened at the account in the 60 days before they first engaged with us? That is your signal.
Step 2: Set up signal capture
For hiring signals: connect LinkedIn Sales Navigator or a hiring data provider (Theirstack, Rocketreach) to Clay via API.
For job-change signals: connect UserGems or Common Room to Clay. Both have Clay integrations.
For intent signals: connect 6sense, Bombora, or G2 Buyer Intent to Clay or to your warehouse.
For funding signals: Crunchbase API or Harmonic both push funding event data.
Step 3: Build the trigger workflow
In Clay: create a table that receives a new row every time a signal fires. Each row should contain:
- Company name and domain
- Signal type and date
- Contact name and role (if available)
- ICP fit score from Play 2's warehouse model
Add a filter: only pass rows where ICP fit score is 7+ AND signal fired within the last 7 days.
Step 4: Write signal-specific messaging
Each signal type needs its own message. A new hire message reads differently from a funding message.
New hire message structure:
- Line 1: Reference the hire specifically ("Saw [Company] just brought on [Name] as [Role]")
- Line 2: The specific problem that role usually faces in their first 90 days
- Line 3: One-line proof that you've solved it for a similar company
- CTA: One question, not a pitch
The message is 4 lines. Not a paragraph. Not a template with [FIRST_NAME] and [COMPANY_NAME] swapped.
⚡ The Verkada play worth copying: Their Clay workflow cross-references a prospect's LinkedIn city against the company HQ city to calculate commute distance, then sends direct mail to whichever office the prospect actually commutes to. This replaced a manual address-validation team. You don't need to do direct mail to copy the logic: use the same geo-signal approach to personalise email subject lines by office location or local market reference.
05The through-line
All three plays point to the same conclusion.
GTM is becoming an engineering function.
The gap between the top 20% and the bottom 80% in this dataset is not tools. It is not budget. It is not brand.
It is the ability to build systems that find the right buyer at the right moment with the right message, at a scale and precision no human-only team can match.
🔗 The plays are a stack, not separate tactics:
- Play 2 (warehouse) is the foundation
- Play 3 (signal-based outbound) runs on top of that foundation
- Play 1 (in-house AI SDR) is the outreach layer that activates when the signal fires
You don't need to build all three at once. But every company running Play 3 at a high level has already built Play 2 underneath it.
06Work with Kinetyca
We build these systems for founders and GTM teams moving from volume-based outbound to signal-based pipeline.
If this framework maps to what you are trying to build, the next step is a 30-minute discovery call with David.
Questions people ask
Which signals show an account is ready to buy?
The guide sets out five: a new hire in a buyer role, a champion moving to a new company, a funding round, intent such as a pricing page visit, and a product usage trigger. To find yours, look at what happened in the 60 days before your last 50 closed-won accounts first engaged.
How much of the outreach should be automated?
It depends on deal size. Below $15K ACV, automation can run the full sequence. Between $15K and $50K it handles the first touch, with a person following up after a reply. Above $50K, a person writes the first touch and automation handles research and scheduling.
How does signal data get back into the CRM?
Through reverse-ETL. The guide models the data in the warehouse with dbt, then uses Census or Hightouch to push fields such as ICP fit score, last signal date and signal type back into Salesforce, so reps see them where they already work.
Software, accelerator and startup background. Runs the GTM architecture side: how the system is designed, what it is built on, and where AI actually earns its place.
LinkedIn ↗YouTube ↗