The LeanScale Podcast · Episode 29

GTM Product Demos: Exploring Ocean.io

Ocean.io founder Michael Heiberg on vector-based lookalike targeting, micro-targeting over mass outreach, and the two generations of GTM AI

Michael Heiberg · Founder & CEO, Ocean.io · Ocean.io Hosted by Anthony Enrico
Published Updated 00:27:04 20 min read 4,094 words
Outbound & Sales Development AI in GTM GTM Strategy Demand Generation Revenue Operations
▸ Listen on Spotify Watch on YouTube Apple Podcasts
SummaryTakeawaysFrameworksQuotesAdviceAIEntitiesMetricsFAQTimelineTranscript
Executive Summary

The one-paragraph brief, extended

Why this conversation matters — and who should spend the hour.

In this GTM Product Demos episode — recorded in the brand-new LeanScale Studio — Anthony Enrico sits down with Michael Heiberg, founder and CEO of Ocean.io, one of the more interesting AI companies in go-to-market tech. Ocean is built on a simple but hard-won idea: clone your best customers. Michael traces the company back to 2018–2019, when the team tried to bring the B2C 'lookalike audience' concept to B2B. Keywords weren't precise enough, so they moved onto the leading edge of natural-language processing and vector databases — 'it required rocket science' — and the technology only became robust in late 2022/early 2023. That's when Ocean took off.

The growth unlock wasn't better explainer copy; it was product-led. After years of failing to describe how Ocean was different, the team had an epiphany: put the technology on the landing page and let prospects key in their best customer and see the exact lookalikes themselves. The 'wow' hit in three to five seconds, and Ocean drove that time-to-value — its 'moments to wow' metric — from more than two minutes down to 30 seconds, fueling a self-serve free-trial motion with up to a thousand onboardings a week.

The heart of the episode is a live demo of Ocean's newer capability. On top of 65 million company lookalikes, Ocean has vectorized 230 million LinkedIn profiles, letting you drop in an individual's LinkedIn handle and find lookalike people — by role, skills, and context, not title — inside the lookalike companies of a target account. You can stack several example people (a head of growth, then a CMO) to build a persona, refine with traditional filters, preview the whole list without spending a single credit, and then export straight into Clay for enrichment. Michael's discipline: validate targeting first, enrich second — 'why enrich something that is not on target?'

The conversation then widens to where GTM AI is going. Michael frames the market as moving from a first generation of 'an LLM wrapper around an analog database' — a pretty face on a non-normalized dataset with broad, imperfect targeting — to a second generation of agentic tools that model the actual GTM process, with 40+ companies now embedding Ocean's logic via API (a Copy.ai agent that fires a lookalike search off a closed-won in Salesforce; RB2B-style inbound resolution scored against your best customer). His throughline is a strong point of view on micro-targeting over mass outreach: automation's real superpower is precision, and a contextually personalized, micro-targeted send converts at 5–10% and opens above 20%, not 0.1%.

Who should listen: RevOps and demand-gen operators building target lists and outbound motions, sales leaders trying to stand out when everyone has the same tools, and founders weighing how to evaluate GTM AI. The most memorable moment is Michael's own inbox — 20 to 30 cold emails a day, almost all deleted, and the one he read because it had clearly been generated off his LinkedIn posts and nailed who he is in the first six lines. Spray-and-pray is dead; contextual relevance is what gets you read.

Key Takeaways

10 things worth stealing

The load-bearing ideas, each with the business implication and who should care.

01

Put the product on the landing page — let buyers hit the 'wow' themselves

Ocean spent its early years failing to explain how it was different in landing-page copy and content. The unlock was letting prospects key in their best customer and see the exact lookalikes for themselves — a wow moment in three to five seconds — then flow into a self-serve free trial on their own data.

Why it matters: For any product whose value has to be seen to be understood, a product-led motion that demonstrates the aha in seconds beats any amount of explainer copy or a gated demo.

FoundersMarketing Leaders
02

Vector-based lookalikes beat keyword and title targeting

Keywords were never precise enough; the breakthrough was NLP and vector embeddings that match companies — and people — by contextual meaning. Ocean vectorized 65M companies and 230M LinkedIn profiles so a team can clone its best customers and find the lookalike buyers inside those companies.

Why it matters: List-building on firmographics plus title strings leaves quality on the table. Contextual matching surfaces the real analogs that keyword filters miss.

RevOps LeadersMarketing Leaders
03

Combine company and people lookalikes in a single search

Ocean's newer capability lets you drop in a person's LinkedIn handle and find lookalike people — by role, skills, and context — inside the lookalike companies of a target account. You can stack several example people (a head of growth, then a CMO) to build a persona, or 'origins,' from just their handles.

Why it matters: Instead of separately building an account list and a persona list, you can express your entire target audience as a few example people and let the system infer both.

RevOps LeadersSales Leaders
04

Target by context and role, not by title

Titles drift with company size — a large company has a CMO, a small one a 'head of growth' or 'VP of marketing' for the same job. Vectoring ties targeting to what a person actually does for that specific company, so you reach the real decision-maker regardless of the title on the profile.

Why it matters: Title-string targeting systematically misses the right buyer at both ends of your size range. Context-based matching normalizes for company stage.

RevOps LeadersSales LeadersMarketing Leaders
05

Validate targeting before you spend credits — start in Ocean, hand off to Clay

Michael refines an entire target list in Ocean — filtering, adding personas, previewing — without spending a single credit, then exports to Clay only once the list is validated. Because targeting is where quality is won or lost, the smart workflow is target and preview first (free), enrich second (expensive). 'Why enrich something that is not on target?'

Why it matters: Sequence your GTM data stack as target → validate → enrich → activate. Enriching an unvalidated list burns credits and pollutes your data.

RevOps LeadersMarketing Leaders
06

GTM AI is moving from Gen 1 (LLM wrappers) to Gen 2 (agentic process automation)

Michael frames first-generation tools as 'an LLM wrapper around an analog database' — a pretty face on a non-normalized dataset with broad, imperfect targeting. The second generation understands the actual GTM process and automates the whole flow, e.g., an agent that pulls a closed-won out of the CRM and returns a perfect lookalike audience.

Why it matters: Evaluate GTM AI on whether it models the underlying data and process, not on the polish of its chat layer. The differentiation is the dataset and the workflow, not the LLM on top.

FoundersRevOps LeadersRevenue Executives
07

Micro-targeting is what AI does best — 5–10% conversion, not 0.1%

The right use of automation isn't mass outreach but micro-targeting: overlay third-party data and intent on a tightly defined audience so what you send is highly relevant. Done well, open rates exceed 20% and conversion lands in the 5–10% range instead of a fraction of a percent.

Why it matters: Point AI at precision, not volume. The same tools that could flood the market are most valuable when used to shrink the audience and raise relevance.

Sales LeadersMarketing LeadersRevOps Leaders
08

Spray-and-pray is dead; contextual relevance gets you read

Michael gets 20 to 30 cold emails a day and deletes almost all of them — but read one that had clearly been generated off his LinkedIn posts and bio. The first six lines nailed who he is and why they were reaching out ('borderline flattering'), and that contextual precision earned the open.

Why it matters: As everyone gains the same automation, generic volume stops working and often lands in spam. The winning play is per-prospect contextual personalization at micro-targeted scale.

Sales LeadersMarketing Leaders
09

Keep a human in the loop — full automation only works for very narrow audiences

Michael doesn't believe you can yet fully automate targeting, copy, sequencing, and delivery end-to-end without human interference — except for very specific audiences. The near-term model is human interaction validating each automated step before moving on to the next.

Why it matters: Design agentic outbound as human-validated stages, not lights-out automation. The finesse — right message, right sequence, right inbox — still needs people until it's dialed in.

RevOps LeadersSales LeadersFounders
10

Build for agents via API — GTM data is becoming an ingredient inside other apps

More than 40 companies embed Ocean's logic inside their own applications through its API. Examples include Copy.ai agents that trigger a lookalike search off a CRM record and inbound-resolution tools (RB2B-style) that score resolved website visitors against a best-customer profile.

Why it matters: A data platform's value increasingly comes from being callable by other systems and agents, not just its own UI. GTM builders should both expose and consume these capabilities via API.

FoundersRevOps Leaders
Frameworks Discussed

6 named models

Every framework Jimmy names, defined and time-stamped.

Moments to Wow

05:41

The time it takes a new user to 'get it' after logging in — a core PLG success metric Ocean actively drives down by putting the product's aha moment directly on the landing page.

Ocean cut its moments-to-wow from more than two minutes to 30 seconds by letting prospects test the technology themselves rather than reading about it, which converted a hard-to-explain product into a self-serve free-trial engine.

Company + People Lookalikes (Vectoring)

06:35

Vectorize both companies (65M) and LinkedIn profiles (230M), then combine them in one search: input an example person's LinkedIn handle and find lookalike people, by role and context, inside the lookalike companies of a target account.

It lets you express an entire target audience as a few example accounts and people rather than a static filter set — 'clone your best customers' extended from the company level down to the individual buyer.

Contextual Targeting vs. Title Targeting

16:48

Target by a contextual understanding of what an individual actually does for a company, not by their title — because titles vary with company size (CMO vs. head of growth vs. VP marketing) for the same real role.

Vectoring finds the true decision-maker regardless of title, removing the manual work of hand-building title lists per company stage.

Preview Before You Pay (Ocean → Clay)

09:56

Build, filter, and preview the target list in Ocean without spending a single credit; export to Clay for enrichment only once the list is validated.

Sequencing targeting before enrichment keeps credit spend tied to on-target lists — 'why enrich something that is not on target?' — and keeps data clean.

The Two Generations of GTM AI

18:10

Gen 1 is an LLM wrapper around an analog/non-normalized database — a pretty face on messy data with broad, imperfect targeting. Gen 2 models the actual GTM process and automates the flow end-to-end, with human validation between steps.

Michael sees a massive wave into second-generation agents that pull from the CRM (Copy.ai) or inbound traffic (RB2B) and act, rather than just returning a nicely worded answer.

Micro-Targeting Over Mass Targeting

20:47

Automation's real strength is micro-targeting: overlay intent and third-party data on a tightly defined audience so every message is highly relevant, producing 5–10% conversion instead of 0.1%.

Doing micro-targeting 'on scale' with an agent — not blasting 5,000 generic emails — is where Michael believes outbound is heading; relevance clears the spam filter and earns 20%+ open rates.

Best Quotes

14 lines worth clipping

Pulled verbatim. Copy or share any of them.

“Why don't we put our technology on the landing page, just bring it up there for people to test it out themselves? We didn't have to explain anything. We had the wow moment within three to five seconds because they had never seen anything like that before.”
Michael Heiberg 03:28
“We push all the companies that we work with to implement a PLG motion as much as possible. Sometimes there's so much complexity in what's going on — it's tough to explain it. You just have to see it.”
Anthony Enrico 03:28
“We talk about moments to wow. How long does it take for someone logging onto Ocean before they get it? We went from more than two minutes down to 30 seconds.”
Michael Heiberg 05:41
“By the way, I haven't used any credits yet — for all the Clay users out there. I haven't used a single credit yet, because I'm refining my search.”
Michael Heiberg 09:56
“At the end of the day, you have to find the people that you can relate to and sell to — the people that feel the biggest pain. To be able to combine that into one easy search is very, very impressive.”
Anthony Enrico 12:19
“If you put a mic in front of anyone, they'll know exactly who they sell to. And now it's basically just dump them into Ocean and you have the lookalikes.”
Michael Heiberg 12:59
“Start in Ocean, kick it to Clay, and then keep the enrichment process going.”
Anthony Enrico 14:41
“Why do you want to enrich something that is not on target? You don't want to do that. You don't want to spend credits on that.”
Michael Heiberg 15:29
“It doesn't tie you into a title. It ties you into a contextual understanding of what that individual does for this company. That's the big difference, huge difference.”
Michael Heiberg 16:48
“First generation was, let me put it in popular terms, basically an LLM wrapper around an analog database.”
Michael Heiberg 18:10
“What automation and AI really does well is micro-targeting. Overlay all this information with third-party data and intent, and then you have a perfect audience for who will actually convert — not in the 0.1%, but actually in the 5 to 10%.”
Michael Heiberg 20:47
“The first five, six lines described who I am and why they wanted to talk to me. That contextual understanding of who I am — it was borderline flattering. That's why I read it.”
Michael Heiberg 23:36
“To all the people out there spamming and trying to get around the hundreds of emails per day — we're seeing that does not work anymore.”
Anthony Enrico 24:23
“The more precise you can do your targeting, the more precise you can do your messaging, the more precise you'll understand what a founder does for each of these companies — the more precise you can write that copy.”
Michael Heiberg 25:51
Practical Advice

What should you actually do?

The playbook, split by the seat you sit in.

Founders

  • If your product's value has to be seen to be understood, put it on the landing page and let prospects hit the wow moment themselves, then flow them into a self-serve trial — Ocean's growth turned on this, not on better explainer copy.
  • Judge GTM AI by its data and process model, not its LLM polish; a 'pretty face on a non-normalized database' produces broad, low-quality targeting.
  • Expose your capabilities via API so agents and other apps can call them — that's where second-generation GTM value compounds.

RevOps Leaders

  • Sequence the data stack: target and validate in a matching engine first (free), then enrich in Clay (expensive). Never spend enrichment credits on an unvalidated list.
  • Move list-building from title strings to contextual/vector matching so you reach the real decision-maker regardless of company size.
  • Express your ICP as a few example accounts and people (lookalikes) rather than a static filter set, and refine it against the live preview before you activate.

Sales Leaders

  • Kill spray-and-pray. Micro-target a small, high-relevance audience and personalize off real context (LinkedIn posts, role, company) — it clears spam filters and converts at 5–10%, not 0.1%.
  • Keep reps in the loop on agentic outbound; validate each automated step (targeting, copy, sequence) rather than running it lights-out.

Marketing Leaders

  • Demonstrate value, don't describe it: a three-to-five-second interactive 'wow' on the landing page beats gated demos for hard-to-explain products.
  • Build campaigns around micro-segments overlaid with intent and third-party data; relevance, not volume, is what earns opens and replies now.
AI Takeaways

How AI actually changes GTM

LeanScale's signature read on the AI-in-GTM question this episode wrestles with.

The thesis

The real edge AI gives go-to-market is precision, not volume. Ocean's bet is that vector-based micro-targeting — matching companies and people by context rather than keywords or titles — plus per-prospect contextual personalization is what actually converts, while the market moves from first-generation 'LLM wrapper on a messy database' tools to second-generation agents that automate the GTM process with a human validating each step.

Two generations of GTM AI

Gen 1 put an LLM 'pretty face' on a non-normalized, analog database — broad, imprecise targeting. Gen 2 models the actual GTM process and automates the flow (e.g., a CRM-triggered agent that returns a perfect lookalike audience).

Micro-targeting is AI's superpower

Automation is best used to shrink the audience and raise relevance — overlay intent and third-party data to reach a 5–10% converting segment, not blast 5,000 generic emails.

Contextual personalization beats first-name-swap

A cold email generated off Michael's LinkedIn posts and bio nailed who he is in the first six lines and earned the open — the same contextual technology Ocean is built on, applied to messaging.

Human-in-the-loop still required

Fully automated targeting-copy-sequence-delivery isn't reliable yet except for very narrow audiences; the near-term pattern is human validation between automated steps.

Validate before you pay

Refine and preview targeting for free, then spend enrichment credits only on a validated list — garbage targeting in, wasted credits and polluted data out.

Agent & automation ideas

  • CRM-triggered lookalike agent: on a closed-won in Salesforce, automatically run a company + people lookalike and route a fresh, validated target list to the owning rep (the Copy.ai pattern).
  • Inbound-resolution agent: resolve anonymous website visitors (RB2B-style) and score each against your best-performing-customer vector before routing to sales.
  • Micro-targeting campaign agent: build and preview a tight micro-segment in Ocean, enrich in Clay, and draft per-prospect copy off each contact's LinkedIn — with a human approving before send.
Operations Takeaways

By function

The same conversation, filtered for RevOps, pipeline/marketing ops, and customer ops.

Revenue Operations

  • Target, then enrich. Validate the audience in a matching engine before spending enrichment credits; ordering the stack this way keeps spend tied to on-target lists and data clean.
  • Context over titles. Vector matching reaches the real decision-maker regardless of title, which drifts with company size (CMO vs. head of growth).
  • ICP as lookalikes. Express the ICP as example accounts and people, not a static filter set, and refine it against a live preview.
  • Data quality is the moat. The differentiation in GTM AI is a normalized dataset and a real process model — not the LLM layer on top.

Pipeline & Marketing Ops

  • Micro over mass. Precision beats volume: a tightly targeted, contextually personalized send converts at 5–10% and opens above 20%, versus ~0.1% for spray-and-pray.
  • Spray-and-pray is dead. As everyone gets the same tools, generic volume stops working and often lands in spam; relevance is what earns the open.
  • Human-validated agentic outbound. Run outbound as human-checked automated stages (targeting → copy → sequence → delivery), not lights-out automation.
  • PLG as a pipeline engine. A self-serve, 'wow in seconds' landing page created up to a thousand free-trial onboardings a week that a demo-gated motion never could.
Metrics Mentioned

The numbers, with context

65M companies
Company universe

The size of Ocean's vectorized company database used to find company lookalikes.

230M
LinkedIn profiles vectorized

Ocean ran embeddings on 230M LinkedIn profiles to enable people lookalikes combined with company lookalikes.

30 sec (from 2+ min)
Time to 'wow'

Ocean's 'moments to wow' metric — how long a trial user takes to 'get it' — driven down from more than two minutes.

3–5 seconds
Landing-page wow

How fast prospects hit the aha moment once Ocean put the product itself on the landing page.

40+
API customers embedding Ocean

More than 40 companies use Ocean's logic inside their own applications via API.

5–10% (vs ~0.1%)
Micro-targeted conversion

Conversion Michael attributes to relevant, micro-targeted outreach versus mass, untargeted volume.

20%+
Micro-targeted open rate

Open rate on tightly targeted, contextually personalized emails.

20–30
Cold emails received per day

Michael's own inbox — almost all deleted; the one contextual, LLM-personalized email got read.

late 2022 / early 2023
Product robust since

When Ocean's vector technology became reliable enough that the company took off.

Entities

Companies, people & tools mentioned

Auto-extracted and linked into the knowledge graph.

Companies

Ocean.ioGTM Data / Sales Intelligence

The subject of the demo — a go-to-market data platform built on vector search. Michael Heiberg is founder and CEO; Ocean pioneered company lookalikes (65M companies) and now overlays 230M vectorized LinkedIn profiles to find people lookalikes, runs a PLG/self-serve motion, and exposes its logic via API.

00:00 · 06:35 · 13:50Company →
DocuBeeE-Signature / Document Workflow

Used as the concrete demo example: Ocean runs company lookalikes for DocuBee (an e-signature company), then finds lookalike people to a DocuBee employee among those lookalike companies.

07:21 · 08:22Company →
Copy.aiGTM AI / Sales Automation

Cited as a strong example of second-generation agentic GTM — building an agent on top of the CRM that picks a closed-won out of Salesforce, fires an Ocean lookalike search, and returns the perfect target audience.

13:50 · 18:10Company →
RB2BWebsite Visitor Identification

Named as an inbound use case for Ocean's API: resolving anonymous website traffic to people/companies, then scoring those resolved visitors against a customer's best-performing-customer profile.

13:50Company →
11XAI Sales / AI SDR

Referenced by Anthony ('quite a few headlines on 11X') as a high-profile AI SDR company, prompting Michael's framing of the first vs. second generation of GTM AI.

17:23Company →

People

Michael Heibergguest

Founder & CEO of Ocean.io, the Copenhagen-based GTM data platform that pioneered vector-based company and people 'lookalikes.'

Founder & CEO, Ocean.ioProfile →
Anthony Enricohost

Co-founder of LeanScale and host of The LeanScale Podcast.

Co-FounderProfile →

Tools & software

ClayGTM Data / Enrichment

The enrichment and automation destination Ocean exports into (transcribed throughout as 'play'). Michael refines and previews the list in Ocean without spending credits, then loads the validated audience straight into Clay.

Website ↗
LinkedInSocial Platform

The data Ocean vectorizes (230M profiles) and the input for people lookalikes — you drop in an individual's LinkedIn handle to find lookalike buyers by role and context.

Website ↗
SalesforceCRM

The CRM a second-generation agent (e.g., Copy.ai) pulls a closed-won record from before initiating an Ocean lookalike search.

Website ↗
ChatGPTAI Assistant

The kind of LLM the cold email that got Michael's attention had clearly been run through — but pointed at his LinkedIn posts and bio, producing a contextually precise, personalized message.

Website ↗
Related LeanScale Services

Where this shows up in the work

AI for GTM

Agents and workflows that multiply operator throughput — the 'make humans superhuman' thesis, operationalized.

The episode is a working example of AI-for-GTM operating leverage — vector-based targeting, agentic outbound, and human-in-the-loop personalization LeanScale operationalizes for customers.

Fractional RevOps

Embedded, field-facing RevOps leadership and execution — the operator model Jimmy describes, delivered as a service.

Targeting, ICP definition, and the target → validate → enrich → activate workflow are core field-facing RevOps work LeanScale embeds.

Frequently Asked Questions

Straight answers

Generated from the conversation, marked up for search and AI extraction.

What is Ocean.io?

Ocean.io is a go-to-market data platform built on vector search that helps teams 'clone their best customers.' It vectorizes 65 million companies to find company lookalikes and, more recently, 230 million LinkedIn profiles to find people lookalikes — matching by context and role rather than keywords or titles. It runs a self-serve, product-led motion where you can test it on the landing page, and exposes its logic via an API used inside 40+ other applications. Michael Heiberg is its founder and CEO.

What is vector-based (lookalike) targeting, and why is it better than title targeting?

Vector-based targeting uses NLP embeddings to match companies and people by a contextual understanding of what they do, rather than by keywords or exact title strings. It's better than title targeting because titles vary with company size — a large company has a CMO while a small one has a 'head of growth' or 'VP of marketing' for the same job. Vectoring finds the real decision-maker regardless of the title on the profile, removing much of the manual, per-company list-building work.

What is Ocean.io's company + people lookalike feature?

It lets you combine two searches into one: start from a target company's lookalikes, then drop in an example person's LinkedIn handle to find lookalike people — by role, skills, and context — inside those lookalike companies. You can stack several example people (for instance a head of growth plus a CMO) to build a persona, apply traditional filters like location and company size, and preview the full audience before exporting it.

Should you build your target list in Ocean.io or in Clay?

Build and validate it in Ocean first, then hand off to Clay. Because targeting is where quality is won or lost, the smart workflow is to filter, add personas, and preview the list in Ocean without spending any credits, and only export the validated audience into Clay for enrichment and downstream automation. As Michael puts it, there's no reason to spend credits enriching a list that isn't on target.

What did Michael Heiberg mean by the 'two generations' of GTM AI?

First-generation GTM AI, in his words, was 'an LLM wrapper around an analog database' — a polished chat interface on top of a non-normalized dataset, which produced broad, imperfect targeting. Second-generation tools understand the actual go-to-market process and automate the whole flow — for example, an agent that pulls a closed-won record from the CRM, runs a lookalike search, and returns a ready-to-use target audience — with a human validating each step.

Is mass outreach (spray-and-pray) still effective, and what is micro-targeting?

No — mass, untargeted outreach largely stops working as everyone gains access to the same automation, and it tends to land in spam. Micro-targeting is the alternative: define a small, tightly relevant audience, overlay intent and third-party data, and personalize each message off real context. Done well it produces open rates above 20% and conversion in the 5–10% range, versus roughly 0.1% for generic blasting.

Does AI-powered outbound remove the human?

Not yet, according to Michael. Outside of very narrow, specific audiences, you can't reliably automate targeting, copy, sequencing, and delivery end-to-end without human interference. The near-term pattern is human interaction validating each automated step before the next one runs — the finesse of getting the right message into the right inbox still needs a person in the loop.

Timeline

Jump through the conversation

Click any chapter to open it in the transcript.

00:00Welcome from the new LeanScale Studio 00:41The origin of Ocean: from lookalikes to vector search 03:28The landing-page epiphany: put the product in front of them 05:41'Moments to wow': two minutes down to 30 seconds 06:35Demo: company + people lookalikes in one search 09:56Preview, then export to Clay — no credits spent 13:50Building on the API: agents, CRM, and inbound 16:09Why vectoring beats title-based targeting 17:2311X and the two generations of GTM AI 20:47Micro-targeting over mass targeting 22:36Standing out: the cold email that got read 25:11Where to find Ocean and what's next
Full Transcript

The whole conversation

Broken into chapters, searchable, verbatim from the audio. Speakers inferred (not diarized).

Transcript as markdown ↓ Whole episode as markdown ↓
00:00Welcome from the new LeanScale Studio

0:00 (logo whooshes) - Michael, we're so excited to have you here today. We have the founder and CEO of Ocean.io, one of the most exciting AI companies in go-to-market tech. Michael, thank you for taking the time to be here. We're also in the brand new LeanScale Studio, and I think it's an amazing way to kick it off. So appreciate it. I know there's so much going on in this space, but the way you've approached go-to-market and leveraging some really powerful AI techniques is really, really exciting for us. I wanna know, and I think our audience is always interested in hearing, what kicked off this idea, and how did you get the idea to found Ocean?

00:41The origin of Ocean: from lookalikes to vector search

0:41 - Yeah, sir, it has to take us back to around 2018, 2019, and we were experimented with the ability to do localize for companies, and what inspired that was basically the whole look-alike concept out of the BTC space.

1:01 We wanted to be able to clone your best performing customers, and as you know, the traditional way of doing that didn't really work out. So we experimented around with keywords, it's not precise enough, and then 2019, 2020, we stumbled into the whole natural language processing and contextual understanding of massive tech bodies, and that got us on a trajectory where we are today, but it was long and hard work because we were on the leading edge of technologies in terms of vectoring databases and what is translation and multilingual abilities and everything else, and the first two or three versions of this the whole way up to 2022 was not really workable.

1:52 It required rocket science to basically do it, and so it's not until late '22, early '23, it became really robust, and that's where the company as such started to take off. - That makes a lot of sense, and you said you took some of the inspiration from B2C or B2B companies. How was that initial adoption for those teams? I mean, when you got into the market, what was this just general concept? How did people take to this type of idea of finding lookalikes for their customers to go target those people? - Our biggest problem and our biggest challenge in the early days was actually explaining why we are different and how we are different,

2:41 and we tried several iterations on our landing page. Here's what we do and explain and then everything else, and it didn't really work out for us, and we were only used by very dedicated companies like your IOTs and things like that that was looking very niche-oriented customer segments. Then we had an epiphany, and that was basically, why don't we put our technology on the landing page? Just bring it up there, and then for people to test it out themselves, and that was pretty much the part where the whole thing started to move. We didn't have to explain anything. People keyed it in. We had the wow moment within three to five seconds

03:28The landing-page epiphany: put the product in front of them

3:28 because they had never seen anything like that before. I just wish we had done it earlier, but that was really what made this whole thing take off. And then obviously our self-sign ops or PLG, the ability to, you don't have to get a demo on this. You can just demo yourself on the landing page, and from there you go into a free trial and test it out on your own data, and that is really what got Ocean going big time. - Yeah, I think we push all the companies that we work with to implement a PLG motion as much as possible, and that's why sometimes there's so much complexity in what's going on. It's tough to explain it. You just have to see it.

4:14 - Yeah, I can only advocate for it. That made the difference for us. Explaining it on a landing page or in all kinds of content, it doesn't really make any sense until you key in your best performing customer and find the exact local likes, and among that local like list, you can see four, five, or six companies you're actually trying to target because you know it's the same thing. That validates the quality of what we do. And it was like no discussion beyond that point. It's spot on. - Yeah, that's really interesting 'cause we talk a lot about go-to-market ops as a way to increase distribution before big tech or other technology companies

4:54 build that technology. So for you, you got a chance to basically use this motion to get it into more companies with less friction, and that's what opened it up. You just got them to use the technology out the gate. - Yeah, exactly. We have, I think we have something like five to 1,000 onboardings a week and on a free trial. We would never reach that by demoing it and everything else. And it puts a lot of things on the line for you as a company because people has to get, and we talk about moments to wow. How long does it take for someone logging onto Ocean and before they get it? And we can actually monitor that based on the searches they do.

05:41'Moments to wow': two minutes down to 30 seconds

5:41 And we went from I think more than two minutes down to 30 seconds. That means within 30 seconds, on average, our trial users, which is to the June or several thousand on a monthly basis, spend around less than 30 seconds or above that to get to that. Whoa, we get it now. And that's very powerful. That is one of our key metrics for success in Ocean is to make it so easy, intuitive that you can get to that. And that's a rippling effect throughout the company. Once you put your whole thing off in front of someone who has no other commitment, then there are time spent on your application. - Yeah, that's amazing. That's amazing.

06:35Demo: company + people lookalikes in one search

6:35 And I think in the spirit of moments to wow, I know Joe and I are really excited to see what you prepared today. And we can't wait to share it with our audience. Can't wait to share it with our customers and using it for ourselves as well. - Thank you very much, Anthony. This is actually the newest version of Ocean. And what that does is it takes what everyone knows Ocean for the company Lookalikes, which is 65 million companies with that has been pictured in a way that you can actually find the perfect Lookalikes. On top of that, we have built a complete victory of 230 million people's LinkedIn profile. That means we have run embedding and victory

7:21 on everyone's in this rooms and everyone pretty much on this calls LinkedIn profile. That allows us to combine two things into one search. That means if I key in the argument for one individual, which is their LinkedIn profile, then I will be able to find the Lookalike origins based on that individual's title and skill sets and everything else in the Lookalike companies that individual look work for. And let me show you an example of that. If I go to Crystal Hergar from DocuP, DocuP is in their e-signature space. And if I go back to Ocean and kick off a company Lookalike for DocuP,

8:22 this is what you've seen before. This is Ocean finding the exact Lookalikes to DocuP. That means all the companies that is the exact Lookalikes to DocuP. If I then go into the new aspect of Ocean, that is basically taking the LinkedIn handle of Crystal

8:48 and do a...

8:55 You see what I just did? Well, based on that LinkedIn handle, I'm now able to find the Lookalikes to Crystal among the Lookalike companies of DocuP. See, I then have the ability to limit it to, let's say, I only wanna see three individuals. I can then include a CMO into this from also a document company. And the moment I can then build my origins into this, I basically start to expand it into in one argument, which is the LinkedIn handles of these individuals, I now build my entire target origins. Obviously, traditional filters like location, like size of companies and keywords and everything else you know is in the system.

09:56Preview, then export to Clay — no credits spent

9:56 I can target it and refine it and by the way, I haven't used any credits yet for all the clay users out there. I haven't used a single credit yet because I'm refining my search. Well, now comes the trick here. I can do many things with this. And I can actually, let's say I wanna save the first 50 just for the sake of, I have more than 10,000, but just for the sake of argument for this demo, it takes a little bit to stop a large file. I now have the ability to, I can basically save it and I can also enrich it or I can actually export it. And here's where I can export it to anything you wanna export it to. And again, I have not used any credit yet.

10:49 The moment I press play, I know this entire target origins directly into play. And from there, you basically, as you can see in the bottom right, you can see it's being loaded over to play. Now I take that entire thing. This is like finding your target origins, honing it in, doing all the filtering, preview the whole thing. Once you're ready, you load it into play and now it's over in play. You see, that is taking this whole thing to a very different level. And we all know that Ocean works inside play, but this, where you actually have the ability to really go into an origins builder capability. And if you go to any sales or any marketing,

11:41 they know exactly who they sell to. Drop in their LinkedIn handle and you get these results in. So guys, what do you say? - That's incredible, that's incredible. It is such a difficult problem to solve that you're able to do in literally minutes and then completely integrated in all the systems that you're using today. It's really, really intentional. And I love that about it. And this is something, this is the type of work that we do with our customers every day and the type of work that we do ourselves. You're looking for the right companies. You know there's gonna be some firmographic components that are relevant to what you're offering.

12:19 But then at the end of the day, you have to find the people that you can relate to and sell to and the people that feel the biggest pain. And to be able to combine that into one easy search, bring in your best customers to find the local likes and then put that into your systems, assign them out, start your outreach, start your way to getting in front of them. It's very impressive, very, very impressive. - If you put a mic in front of anyone, they will know exactly who they sell to. They'll know, yeah, I sell to John, I sell to, you know, they know them. And now it's basically just dump them into Ocean and you have the local likes.

12:59 And for any agency out there, they can adapt it to their target origins per client. And it's extremely powerful and you can save all the searches. You can fine tune the searches and you can actually then load it into whatever, you know, CRM or whatever clay you wanna load it into. And we see that as a game changer in the market. You know, obviously we are first mover again, we were first mover on the company lookalike and now we overlay that with the entire victory of LinkedIn on top of that. So we think this is gonna be a game changing for the entire go to market. And by the way, on our API side, this has been available for some months.

13:50Building on the API: agents, CRM, and inbound

13:50 You saw Freggle actually released this a couple of months ago. And can you imagine building an agent on top of your CRM system, which is what copy AI is doing, using this ability to on a close one, you can basically pick it directly up from the CRM from Salesforce and then initiate a lookalike search on the individual or on the company or both. And what you get back into that agent is the perfect target audience. You could do it on your landing page. Guys like RB2B, they resolve all the inbound traffic, we can put it into context of your best performing customer. All these things is what is going out on the our API side. But that's to give you a little bit of

14:41 for what we're doing at Ocean right now. - Yep, and I think you also, for all the clay users out there, it sounds like starting in Ocean now could be a better workflow if you're trying to do this. So I think that's a really big takeaway as well as you don't have to go and put this into an individual column or so. Start in Ocean, kick it to clay and then keep the enrichment process going. - I would lead it up to the individual users who have seen this, if they believe this is a smarter, more smoother way of doing it, easier way of doing it. And everyone knows it's the preview that will show you whether you're on the right track.

15:29 And those preview, you need to be able to validate. And once they are validated, then you know your list is correct. And then you can do the enrichment. Why do you want to enrich something that is not on target? You don't want to do that. You don't want to spend credits on that. So I think your spot on. - If it were me, I would start the workflow in Ocean to get the preview. Because I do think there's some experimenting that you might want to do. Like you added a head of growth, then a CMO, then maybe DocuBee, but maybe it's not just DocuBee. Maybe there's some other ones that you want to kind of relate to start to build a different persona or profile.

16:09Why vectoring beats title-based targeting

16:09 And then once you feel like, hey, yeah, these results are starting to feel like what I'm looking for. Then pump it into Clay and start using your credits on other things. - Exactly. And as you can see, anyone who has used title searches and everything else, it gives you really limited because it goes only by the title. It goes by the picture of those individuals. That means you get a much more, or, you know, origins that is on the level you actually want. Because in some companies, on the bigger companies, they are cheap marketing officers. In smaller companies, they have VP of marketing or head of growth. That is what Victoring does.

16:48 It doesn't tie you into a title. It ties you into a contextual understanding of what that individual does for this company. That's the big difference, huge difference. And you can look at the titles if you go through it, it relates to the size of the company. - No, and that's something, normally we're doing this manually. We're saying, okay, I need to do a profile of, if it's the size of the size, I'm looking for these titles 'cause they're typically running sales. They're typically running marketing. And then if it's at this stage, I'm actually looking maybe for like a director level 'cause maybe they have the budget for this thing that we're selling.

17:2311X and the two generations of GTM AI

17:23 So I think this takes so much manual work and investigation out of that process. - If we look to Think, which is a very small company, they don't have the VP. They have a head of marketing. But that is their decision maker within the marketing sphere for that particular company. That's what Victorin does for you. - So Michael, I know something that's, it's been in the news a lot right now, quite a few headlines on 11X and what they're doing and how they've approached their business. I think from your perspective, how do companies like 11X relate to what you're doing and relate to the future of AI and where AI is going?

18:10 'Cause I think you planted a lot of seeds during that demo of some of the agentic capabilities that you're gonna be building the foundation for. From your perspective, where do you see this all going? - I think we're moving fast into second generation right now. I think first generation was, let me put it in popular terms, was basically an LLM wrapper round analog database. Why is that significant in terms of using the processes, the go-to-market processes? That means you have a non-normalized, and now I speak, data science speak, you have a non-normalized data set where you put a pretty face on it, which is what the LLM produced for you,

19:01 but the processes and the targeting of origins is not perfected. It's very broad and it's based on that analog database. And I see, and it's just, we have more than 40 donors using Ocean inside their application. That means they're using the Ocean logic inside their application. And we see a massive wave going into those second generation agents. And that is where there's a much better understanding of the actual process that we actually have to support. And a process automating the whole flow when copy AI is a good example of it, where you can actually take straight out of a close one and do magic with it. I think this demo illustrated that.

19:57 If it's an API supporting it, I see those processes now being starting to be interesting for people, because it's the first wow effect of, I'm speaking to a robot or it's a robot doing the whole sequence. I think we're moving into much more human interaction with an automated flow validating and moving on to the next automated flow. I don't believe we are there and for the foreseeable future, where unless it's a very, very target audience, I don't think we are there where we can actually fully automate it without any human interference, write the right copy, embed it into the right sequence, perfect target audience,

20:47Micro-targeting over mass targeting

20:47 and make sure it reaches the right individuals in the inbox and in their LinkedIn profile with the right message. I think that needs finessing, but once it's finessed, it runs. And that's where I'm a big advocate for micro-targeting instead of mass targeting. What automation and AI really does well is micro-targeting. Overlay all this information with third-party data, intent everything else, and then you have a perfect audience for who will actually convert, not in the 0.1%, but actually in the 5 to 10%. Because what you send to them is highly relevant. And doing micro-targeting on scale without anything but an agent, I think that's where we are moving.

21:49 Does it make sense? - It does make sense. And I had a follow-up question because I think this is something we worry a little bit about in the go-to-market space. I think these capabilities are being available to the masses now. And we're seeing just the sheer volume, even when it's micro-targeted, the sheer volume of communication and being targeted, really being amplified. When everybody has access to these capabilities and tools, and I think what you mentioned, keeping the human in the loop can help finesse this a little bit. But how do you stand out when the playing field gets to the point where everybody can kind of do this at scale,

22:36Standing out: the cold email that got read

22:36 how do you stand out to still be competitive in that new environment? - I got an email the other day, and I think I get 20 to 30 cold emails a day. And most of them end up in my spam filter, but I actually got one I read. And it was very obvious that the script was run through any of the LLMs, chat GPT or whatever. But it was, that script had been run on my post on LinkedIn.

23:07 And it was run on my text on description on LinkedIn

23:14 because it was spot on. It was spot on what I stand for, it's spot on what ocean stands for. So the first five, six lines, which is unheard of, describe who I am and why they wanted to talk to me. You see, that contextual understanding of who I am

23:36 and that communication to me with spot on, it was borderline flattering, but it's, anyway, exactly, of course. - That never hurts, I don't mind. - That's why I read it. Yeah, that's why I read it. No shame in saying that, no shame in saying that. But it was like, you know, that was interesting. That was an interesting, wow. Because intellectually I was stimulated, how did they do that? How did they get it so precise? And I decided to figure out the algorithm behind that, which is interesting. But those kind of, that's because that's really what we do with ocean. But it's, I think, when it gets to that level

24:23 of contextual understanding of who you communicate with and what business I'm in, then it becomes relevant. And that's back to the micro-targeting aspect of this. You do not do this on 5,000 emails. You do this 5,000 emails. But the open rate they get, I don't know how the underlying data bike, I tell you it's more than 20%. - To all the people out there, spamming and trying to get around the hundreds of emails per day. I think, yeah, we're seeing that does not work anymore. So, you know, hopefully we'll save you some pain trying to get around that spam filter. But I completely agree with you and appreciate you showing us how people

25:11Where to find Ocean and what's next

25:11 can actually do this in practice. I think you're allowing this to happen with ocean. And yeah, I mean, before we sign off here, if people are interested, reach out to you, reach out to the team. What's the best way to get in touch with you to learn more? - Yeah, you know, test us out. This is the best thing. You know, people, there's some people who prefer to communicate on a demo, others actually just wanna see how it works for them. There's no barrier here. There's no, you know, you don't have to, you know, just go in, test it out. And I actually meet a lot of people who have said, oh yeah, I went to your landing page, this is amazing.

25:51 They might not be in the market now, but when they are in the market, they come back. That's what we see. Yeah, then obviously you get the, you know, LinkedIn operation, I just close to new meetings and all these kind of things. Thank you very much. I appreciate that. I obviously appreciate that, but it's down to targeting. The more precise you can do your targeting, the more precise you can do your messaging, the more precise you'll understand what a founder does for each of these company. If that is your target audience, the more precise you can write that copy. - Amazing, amazing. Well, sounds like easiest way. You can see it in Michael's background.

26:26 You can see it on our big board back here. Ocean.io, go test out the product. Take a look for yourself. You've made it so intuitive and easy. And really just appreciate you being here with us, Michael. And can't wait to see what you guys do next because you've been pushing the limits so much since the beginning. I know there's more in the future too. - Yeah, there's a couple of things up to sleep. I tell you, there's a couple of things up to sleep, but there, that is for, you know, we can meet again in a half year. Then I have a couple of things to lay on top of this. - We'd love to have you back. - Thank you very much.

27:00 I really appreciate this and thanks for the opportunity. Amazing. Thank you, Michael.