
Head of Studio @clay i have a pasta biz on the side Ex-HubSpot (Views and opinions are my own)
New York, NY
Joined May 2017
- Tweets489
- Following470
- Followers285
- Likes6.8K
Jake retweeted
"But like.. what does @clay even do!?" -- Here's my take.
Explain Clay to:
1. My Grandma: Sales AI platform
2. A first Date: AI assistant for sales & marketing. Research potential customers, draft emails, handle repetitive work so people have more time for real conversations (and life).
3. My art school friends: Businesses buy contact data to figure out who to prioritize. Clay is like Netflix for those customer insights + Agents that help make sense of it to automate the busywork
4. My ex-finance friends: Excel with 220+ plugins like FactSet or CapIQ with concierge account Agents. Enrich prospects with custom firmographic meta-data. Agents route prospects based on context and tier + execute build, launch, and iterate campaigns. Lower CAC. Raise deal velocity.
5. To CROs: Clay is a GTM orchestration and enrichment platform, combining your 1P data with 220+ data providers to give you the maximum coverage and accuracy. Unify your sales & marketing team with this always-updating source of truth. Build inbound plays, outbound automation, rep productivity, and ABM campaigns to turn your revenue systems into a self-learning GTM motion.
⇒ See how you can replicate the same systems build by Anthropic, OpenAI, SpaceXAI: dub.sh/uDlnkKw
6. To GTMEs & RevOps: Data hygiene is solved. Now, orchestrate your GTM data and build workflows straight from your agent of choice. Get the benefit of enterprise-grade infra (scale, audibility, security, credit budgets, etc) along with the "idea-to-execution" agility of your best agentic plays.
What'd I miss? (Shout out to grandma CROs everywhere).
13 REASONS TO BECOME A CLAY CAMPUS AMBASSADOR:
↳ Enrich your crush and get their number
↳ Automate your internship hunt
↳ Throw campus events on our dime
↳ Escape the banking-consulting-pre-med pipeline
↳ Hunt down elite cheese curds
↳ Build a secondhand college-merch empire on Depop
↳ Outsource your dance-class calendar to Clay
↳ Flex free merch that clears your Greek life merch
↳ Learn Clay and GTM engineering with other students (massive upside)
↳ Meet our talent team if the pop-star thing flops (also massive upside)
↳ Tell AI to write the code you might've written as a SWE
↳ Build with free Clay credits
↳ Get a shot at SCULPT
Apply to be a Clay Campus Ambassador by Dec. 3: clay.link/sGjjKcj
iykyk: 🕷️📚🕷️📚🕷️📚🕷️📚🕷️
Jake retweeted
Clay is building a rep tool.
I’m trying to understand what it would take to make a BDR/SDR or AE confident enough to switch from their legacy tool.
@gtm_engineer told me I need to validate this with 100 sales people.
Who should I be talking to?
Jake retweeted
Anthropic announced Fable 5.2 Monday on X but the announcement email hit my inbox at 8:13 am on Tuesday. That's why X is the go to B2B social channel now.
I've been getting so much value from X content that I even deleted Instagram off my phone.
People are sharing such great tactics and use cases for etc that it's going to be a better B2B marketing platform than LinkedIn next year.
The big AI companies are optimizing their launches for X which in turn brings all of the B2B buyers to X so they never miss an announcement.
Dragging them away from LinkedIn and on to a platform that mixes content they can genuinely use to grow their business, news, and any other topics they find interesting. I get random weightlifting and classical guitar videos.
Also, the reach seems to be easier to get a post past 10k views than on LinkedIn.
Mix this with Grok Bot and it's social listening and it's already a great B2B channel but I think in 2027, it's going to be a must have for every B2B company.
Jake retweeted
Well that was fun.
Had no idea GTM Engineering existed.
Spent the last ten hours doing no work, only watching 10 hours of GTM Engineering / Clay YouTube videos.
Jake retweeted
The number one thing I think sales teams should be doing but aren't is (compliantly) downloading all of their connections and pooling the intelligence together.
You're going to forget almost everyone you're connected with. And even if you've never spoken to them, it's still a little bit of a connection.
Here's the whole play:
1. Every rep (and anyone else willing) goes to Settings > Data privacy > Get a copy of your data > Connections.
2. LinkedIn emails you a CSV a little later.
3. Send every CSV to one place and upload to a Clay audience.
4. Then merge it into one list and match it against your TAM.
5. Anyone who fits your ICP gets a direct email/call/outreach from the person connected.
The opener we use when you have nothing better to say (and please, think about something better to say but this works too)
"We've been connected since {{date}}. I've never introduced myself."
That's the best generic opener we've found. When I said this on stage, someone even implemented it the next day and booked meetings.
I'll keep repeating that the best customization to a message is unique to each company/contact you're reaching out to, but that's one I don't see anyone using when they message me and it crushed when I was running Growth Engine X.
Jake retweeted
"If you have an idea that you genuinely think is good, don't let some idiot talk you out of it."
Stan Lee said that, and I think about it a lot, cause when you lead a team, you can very easily become that idiot by telling your team not to do something.
Here's the thing about ideas. When someone pitches you something new, you might think it's stupid. I quite often do, actually. But ultimately I don't have enough datapoints to know if I’m right or they're right. Nobody does, until we actually try it.
So unless there's an obvious flaw in the thinking, I try never to just shut something down. Cause if you shut your team down two or three times, they stop bringing you ideas at all. They fall into this apathy where they stop seeing the point of trying.
Our livestream series How Clay Uses Clay is the example I keep coming back to. I pitched the idea to marketing before my own team even existed, and it sat there for six months cause nobody wanted to own it. I stopped waiting, hired Nicole Goot and we just launched it together as a Growth program. We’re 8 months post-launch, and 65% of our opportunities with over $100k pipeline have at least one person in the buying center with a livestream touchpoint.
If you're leading a team, especially in a function where creativity is essential for success, don't be the guy who tries to kill Spider-Man. Let your team cook.
Jake retweeted
Many people in GTM said that Claude Code “killed” Clay.
But in 2026, Clay has been shipping like crazy.
And the platform has completely changed.
10 new features you should check out:
Jake retweeted
155,833 leads built through Clay in one week, across three client builds.
135,879 for a financial advisory client.
12,000 for a consumer-brand agency client.
7,954 for a security-services client.
None of them started with a fresh list. Every one started by counting what was already in that client's Clay Audiences.
Jake retweeted
People overthink how to invest in brand awareness motions. Brand growth just means having the right people consuming your content.
“Why have I been seeing @clay everywhere for the past 2 years?” I get this a lot.
Anyone can do it and the main thing you need is a deep understanding of your customer and very strong channel conviction. Content consumption and engagement on a channel is the indicator you should look at in terms of brand awareness.
Basically, are you growing your presence within the channel your customers are on? Are you creating content that is relevant to them? That’s enough. In other words:
Brand awareness is relevant content creation in the right channel.
For Clay, this was and still is LinkedIn, at Figma the channel is X, Owner focuses a lot on traditional social channels like Facebook and Instagram.
The last point is that you can distribute this through your own brand (company and execs) and through other people in this channel. You should map the people your ICP follows and work closely with them to leverage this distribution motion. You should aim for all of them.
Later on, you map additional touch-points like events, out-of-home, etc. The goal is for relevant people in the market to feel that “everyone is talking about you.”
Coming from a growth background, investing in this area at Clay ~3 years ago was very scary to me. I wanted control over every single input. However, investing here early is what actually compounded over time.
Happy to answer any questions in the comments!
Jake retweeted
A year ago, X (y'all) roasted companies for hiring $270k “storytellers".
I’m one of those cringe storyteller dudes (though I don't make $270). I lead social at @Clay. TBH I think the title is a whole a** meme -- but alas, a consensus seems to have emerged:
> Distribution's importance (vs product) is increasing
> Taste, literary agility, and new media fluency generate alpha. Legacy relationships & paid media budgets don't
We've been focused on LI for the the last few years. It's where our ICP is. We coined the term "GTM Engineer". Got on everyone's radar. Became synonymous with the 'cutting edge' in our space.
Here are 5 decisions behind how we built our brand, presence, and narrative dominance to get the market to speak our language. 📺 ⤵️
See you 'round the feed :)
Jake retweeted
The @Clay sales team just had its biggest quarter ever.
- We hit 157% of company target in Q3
- We broke our all-time largest deal record... then broke it again
- 5 sellers each closed >$1M in new ARR, and one crossed $2M (!!!)
- 80% of sellers finished over 100% of target.
- 36 new Claymates joined us in sales, and 9 people got promoted
....and we did all of this while navigating 13 new product releases over the summer and throwing Sculpt, our biggest user conference yet w/ 2,000+ people in SF. The team also hit 30+ events in 12 cities this quarter, including Dreamforce, HubSpot’s Unbound, and an NFL game in London.
This is broad-based attainment where everyone can win and many can win BIG. Come join us. We're building the future of AI GTM and the best sellers want a front row seat.
clay.com/careers
Jake retweeted
The team @clay is executing. It’s a great time to join.
The @Clay sales team just had its biggest quarter ever.
- We hit 157% of company target in Q3
- We broke our all-time largest deal record... then broke it again
- 5 sellers each closed >$1M in new ARR, and one crossed $2M (!!!)
- 80% of sellers finished over 100% of target.
- 36 new Claymates joined us in sales, and 9 people got promoted
....and we did all of this while navigating 13 new product releases over the summer and throwing Sculpt, our biggest user conference yet w/ 2,000+ people in SF. The team also hit 30+ events in 12 cities this quarter, including Dreamforce, HubSpot’s Unbound, and an NFL game in London.
This is broad-based attainment where everyone can win and many can win BIG. Come join us. We're building the future of AI GTM and the best sellers want a front row seat.
clay.com/careers
Jake retweeted
This article is one of the first examples of well measured GTM harness I've seen - and I think it bodes well for @clay and other specialized GTM harnesses who follow.
Now... the comparison here is not really fair, which distorts the experiment a bit. But they still make a very good point that's worth considering.
They're comparing a general prompt fed to likely Claude/ChatGPT with basic @salesforce and @Gong_io connectors, to a purpose built account harness.
That's like saying "we put a bodybuilder in a 1 mile race against a collegiate mile champion and the bodybuilder lost!"
Of course, that's pretty obvious.
That said... the comparison is useful and fair because for the vast majority of companies... they are not going to want to deal with the overhead of building, managing, eval'ing, and optimizing an account research skill over time.
The ability to just plug in an account agent and have it just... work at $0.006 is insane.
And considering that models will likely be far better in a year that's powerful.
Even if you COULD build and eval your own account research skill... you'd be better off just using Clay and building skills and agents on top of their agent.
The other part that is very thought provoking to me is the input tokens and tokens per step.
Clay's account agent seems extremely lean and is loading exactly what it needs, and is very efficient comparatively at reasoning out the next step.
Now, 250k tokens is pretty bloated. That is either an insanely long prompt, there are tons of MCP tools, or lots of memory. My Claude code sessions are bloated and they start around 150k tokens. So at $.006 they probably are using a worse or fine tuned model.
Clay also has to maintain the account context which would be separate cost. But you'd need to do that either way.
Still... even assuming you trim your default context down to 100k, and have a good skill... it would be VERY hard to beat Clay at scale.
And at a certain point... why bother.
You probably would still want a frontier harness to build more intense thinking that relies on details specific to your company messaging and positioning.
But maintaining account context?
Definitely just use Clay.
Model testing? No. Data orchestration testing → 6.4x fewer input tokens. Answered 2.6x faster.
We gave 2 agents the same 5 enterprise accounts and pre-call questions:
→ A general-purpose agent connected to Salesforce + Gong over MCP
→ Clay's native Account Agent, running on Clay's data layer
We hid deal risks in the accounts and ran each agent 30 times.
What we found:
1/ Duplicate records hide deal risks.
General-purpose agent caught it 18/30 runs. It checked one and moved on.
Clay merged it and caught it every time. 30 for 30.
2/ Data sync overwrote a risk rating from medium to low.
General-purpose agent missed it every time. 0/30 runs.
Clay smelled something fishy, read through additional fields, and caught it 16 of 30 runs
3/ Organized data gets agents to the answer faster.
General-purpose agent paged through 100s of unrelated Gong calls.
Clay opened an account that was already sorted. 6.4x fewer input tokens. Answered 2.6x faster.
Agents aren't a magic box. They exist within a data architecture and their performance is a function of how well their context is orchestrated.
Full report + methodology: clay.link/zaXrceT


























