53% of leaders are seeing little to no impact from AI. How are some GTM teams pulling ahead?

They're connecting their internal context and intelligence directly to their AI assistants. No more jumping between dashboards. No more lost context. Automatically turning insights into smart actions.

AirOps, which helps you craft content for winning in AI search, is hosting a live session showing exactly how leading GTM teams are doing this. Learn how to upgrade your strategy and reporting, building branded decks and shareable dashboards from your data. Join the free webinar on February 9th.

One of the favorite workflows for the companies I work with is building their own AI Content Assistant that helps more of their team members get active on LinkedIn.

How to use it: Help you ideate, draft, and edit content.

Step 1: Create a new Project in your ChatGPT account.

Upload key files to the project like your messaging or brand guidelines, knowledge base (all key descriptions of your company/product), and examples of your best-performing content.

Step 2: Write detailed project instructions on how to help you with content.

Define your voice, “do”s and “don’t”s.

Example: In this project, you will help me write LinkedIn content for [name] - he/she is [short bio and key fields of expertise]. Create strong hooks, avoid overused buzzwords, and don't use hashtags.

Step 3: Define the voice, if needed, or add some examples.

Example: You can use content from the attached files to refer to our services.

Use case: Build competitor comparison pages with Clay, Claude, and Webflow

Why it's been great for my clients:

Step-by-step workflow:

PS: Matteo is running his 3rd cohort of the GTM Engineer School, which includes 16 hours of live hands-on training. Growth Unhinged readers can save 15% with this link.

The use case: BattleBot is a sales co-pilot for our revenue teams. It analyzes Gong and opportunity data, then provides competitive strategy, research, case studies, landmines, and whatever else they need to win.

Why we built it: Traditional battlecards are static and aren't customized to a specific deal. Competitive intelligence is also bottlenecked by Product Marketing bandwidth to conduct research and analysis. BattleBot solves both: we scale human insight by automating pipelines of competitive intel and win-loss analysis, then make that insight accessible and dynamic through a conversational interface.

The workflow:

How I use Claude to simulate customer reactions to our marketing before it even launches

Most marketers use customer research sparingly because it takes so long. But what if you could have customer research on tap, and ask for opinions of your work before it ever launches? After reading research from PyMC Labs and Colgate-Palmolive, I realised AI can act as your customer's digital twin and give accurate feedback to help you iterate faster.

My digital twin runs on three layers: real customer data, behavioral insights, and anchor statements.

Then I load everything into a Claude Project as a persistent "digital twin" I can test any campaign against. Caveat: this system gives you the accuracy on purchase intent, but only if you feed it real customer data. Synthetic data in, synthetic insights out. The twin is only as good as the voice-of-customer you train it on.

Get the full tutorial via Kieran’s newsletter.

​After noticing a subset of deals closing faster than our average, we built a binary scoring model combining hard data (revenue, age, industry) with AI research to pinpoint high-potential prospects. Companies passing both checks receive increased marketing spend and sales focus. The goal of this algorithm is to create a perpetually running scoring model that answers in a simple true or false whether a company is within your serviceable addressable market (SAM).

Here's how you can recreate our model:

​We finally cracked outbound at scale for Inflection because of AI-based personalization. We're now booking more meetings per month WITHOUT BDRs than when we had two BDRs.

Here's how we do it:

We’ve built out some pretty wild AI-powered outbound and micro-ABM workflows using Clay, RB2B, Instantly and ChatGPT. They’re driving strong results, faster conversions, big pipeline lift and super-personalized outreach that adapts in real time to buying signals.

Crescendo's marketing and sales teams use a human-in-the-loop AI system that automates prospecting, research, and sequencing across three distinct funnels: ABM, outbound, and inbound-to-outbound. The system combines HubSpot, Clay, Instantly, Heyreach, Nooks, and OutboundSync, with Gong for discovery-call intelligence and Slack for live routing. The AI engine autonomously identifies, enriches, and engages prospects, then hands off to humans at key conversion points.

Impact metrics:

Prospect Identification

Marketing and sales upload or tag accounts based on ICP, ABM signals, or social metadata. Clay enriches each record with phone, email, LinkedIn, tech stack, and outsourcing data. The system refreshes accounts weekly, giving BDRs and marketing a new set of prioritized targets based on the most recent buying signals.

Automated Research and Sequencing

The system auto-generates email, LinkedIn, and call sequences. Sequences are sent via Instantly, HeyReach, and Nooks. Replies automatically stop the sequence.

Prospect Prioritization and BDR Enablement

HubSpot prioritizes and assigns tasks to BDRs based on enrichment data, engagement signals, and ICP fit from the previous steps. AI call scripts are generated directly on each contact record. Each BDR has a personalized Clay table, filtered by ICP fit and campaign relevance.

Engagement and Handoff

BDRs engage prospects using hyper-personalized, AI-generated messaging and Gong-recorded discovery calls. No-shows automatically re-enter Clay for resequencing. Qualified opportunities are passed to sales automatically.

Closed-Loop Feedback

Replies sync back to HubSpot and Slack via OutboundSync. Marketing logs win/loss data and attaches Gong recordings to discovery calls. Non-responses trigger resequencing through Clay.

Thanks for reading Growth Unhinged! To receive new posts and support my work, consider becoming a free or paid subscriber.

AI becomes exponentially more useful as you increase the surface area of data it can access. Before any meeting, we upload all prior meeting transcripts, LinkedIn conversations, email threads, HubSpot info, and website information into a custom GPT. The GPT shares a summary of all prior meeting history, LinkedIn overviews, and potential issues the prospect is facing. In addition, the transcripts are all query-able, so we can ask any questions on prior meetings in real-time.

Thus, in ~2 mins, we have all the context we need for every meeting. The whole process right now is orchestrated by an SDR (takes him 5 mins), though we're looking to automate the upload component as well.

It's a custom GPT inside of ChatGPT. We simply paste in the following items in order:

The prompt for the custom GPT essentially asks for these 4 outputs: (1) Attendees & Personal Details, (2) Purpose of Today’s Meeting, (3) Summary of Past Meetings, (4) Prospect Context & Opportunity.

Deals that are closed lost are much easier to re-engage than getting a brand new logo aware, educated, interested, and signed so we’ve automated our re-engagement. We use Sybill as our AI call recorder which is connected to HubSpot. We created about 12 custom fields following our sales methodology (SPICED) plus other things like tech stack, etc. and Sybill listens and fills in and updates information based on the calls.

When a deal is moved to the closed lost stage in HubSpot, it’s automatically added to a dynamic list of “Closed Lost Deals” and added to a HubSpot workflow to re-engage. There is a 90-day wait step and then they start receiving content. After the 90 day wait period, the deal is automatically sent to a Tofu campaign that ingests the fields that Sybill created and other fields like closed lost reason, notes, etc. For each deal Tofu runs an additional research agent looking for any new chances to marketing leadership, new campaigns, events, or content they’re recently produced, and any updates to their GTM motion.

The setup: Sybill (AI call recorder), HubSpot (CRM + Marketing automation), Tofu (personalized content creation), HubSpot + Amplemarket (email + LI execution), LinkedIn (ads)

The exact prompt: Have they had any new marketing leadership changes in the past 3 months? What personas and industries do they sell to? What types of marketing content and campaigns are they currently running? Have they made any major announcements, hosted or had a major presence at any events in the past 3 months? Have they made any recent changes to their GTM motion?

It gets combined with the CRM fields and Tofu creates personalized content for each contact: (1) 4 marketing emails, (2) 4 sales emails, (3) 2 LinkedIn DMs, and (4) a 1:1 landing page.

The landings are exported back to our CMS (Webflow) and dynamic links are embedded in the emails and DMs. The marketing emails containing personalization tokens for the subject lines and bodies get exported to HubSpot and added to the workflow. The sales emails and LinkedIn DM personalization tokens are added to an Amplemarket sequence (and Amplemarket is integrated into HubSpot). The account also gets added to a LinkedIn Campaign Manager audience where we start showing their marketing team ads (these are 1:many).