Mark Kosoglow

Chief Revenue Officer (CRO)

Are your customers your growth engine? They should be.

As a teenager, watching 14 videos in the back storeroom on a 7" black and white TV to learn how to sell shoes at the mall was a great foundation. Running a small business with 200+ employees showed me how to be organized. Creating a highly profitable sales territory from one dead for a decade was hard work. Managing 12 salespeople across 9 states cemented my sales philosophy. Getting one of the fastest growing SaaS companies in history to $250M+ in revenue taught me how to balance scaling and delivering a world class customer experience.

Working with a customer-centric team with, by far, the best, smartest, hardest working people I've ever worked with....well, that's an honor and privilege I get to enjoy everyday.

CFO - Chief Fun Officer

Doing dad stuff, exploring interesting and fun opportunities...cookin' up something SPICY!

We are hiring a Services Sales Manager at Docebo. 
This is a critical position that we need to fill quickly.
If you are interested, please DM me!

If you clicked "See More", what you should know:

🤩 You will be leading a team of 3 awesome professionals

🤝🏼 You will work closely with our amazing AEs and AMs

🫶🏼 You have a host of dedicated partners

😍 You will report into amazing PS and CX leadership

💵 Competitive pay and great benefits

🌴 This role has potential to go to President's Club

🧠 Our GTM org is smart, hungry, and building cool stuff

🤖 We have an AI assembly line so you'll get to learn and flex those skills

🎁 An amazing product that just got better with recent acquisition

⚡ Have some fun in a great culture that knows how to have fun

Everyone has AI. Humans are the differentiators.

My newest demo technique is what I call bursts.

3-5 sections that are 3-5 minutes long.

Instead of putting on a 30-minute show.

Prepare 3-5 TikTok clips.



Our brains are rewiring.

People think movies are too long.

But those same folks will watch a whole movie if it comes in a series of 3-5 min TikToks.

Data backs this up. Consensus 2026 Buyer Behavior Report says most buyers only watch 5 mins of 15 min long demo videos.

The feeling of progress is real.

Watching a TikTok and moving to the next feels like progress.

You can fight it.

Or, you can use it.



Here’s an example with Docebo.

We just acquired 365Talents.

Our demo could easily go from 30 to 60 mins.

Uh oh.

We would be out of TikTok AND even TV episode range into feature films.

Engagement could wane and so would results.

So what if we did 4 clips like this (each less than 5 mins)?

1️⃣ How our AI creates a complete taxonomy of skills by role using job descriptions, best practices, and internal documents and data.

*Pause, converse, and assess*

 2️⃣ How we use multiple sources (including humans) to assess the skills of an employee.

*Pause, converse, and assess*

 3️⃣ How we use skill gaps to serve up individualized learning served up in the way the leaner likes to receive information the most.

*Pause, converse, and assess*

 4️⃣ How we evaluate skill mastery by using systems of record to provide data of the outcomes the skill should result in.

*Pause, converse, and assess*



That can be a 30 minute “marathon.”

Or it can be 4 “bursts” each 5 mins long.

And you ask questions between bursts to make sure the buyers are following you and converting the information into ways to apply your product to their problems.

That works.

Better yet.

Demo platforms like Consensus help you:
- create these bursts
- use them in autonomous demos
- get data on the demand of each burst
- see where you are being confusing
- do amazing followup

We are in a faster world.

One where attention is so bombarded with gratification that 4 minutes of boredom causes us to throw out 26 minutes of greatness.

Reduce that risk.

Make your demo pop with some bursts!

P.S. I got this idea organizing my 2 sons' Lego collection over the last few months. Each burst is a brick that you can add to an awesome build!

AI GTM tech is in T-R-O-U-B-L-E. 99% is just a prompt built on top of an LLM I can use for ~1% of the cost with a prompt I wrote that is ~80% as good.

10 yrs ago, would you sell Google search queries templates to get better Google search results?

No, even though few people use?
- Exclude Words: Use “-” before a word to exclude it from the search results
- Site-Specific Search: Use “site: {{URL}}” to restrict results to that specific source
- Related Websites: Use “related: {{URL}}” to find sites similar to it

Nobody paid for or even created a tool that injected those parameters into your search bar despite their making Google work better.

Yet, most AI GTM tech is a version of this.

They are a middleware that relies on a prompt to add value to a workflow that usually can be created using your current systems.

What' gonna happen in 1 year?

I think AI-natives who saw insane growth in less than 12 month will see insane churn.

Crippling churn.

Example: Jasper (based on my research)
- from 2020-2023, got to $120M in ARR
- left 2024, at $55M in ARR
- almost 100 employees today


Now, I’m a failed AI GTM tech founder so take this as you will.

Winning AI GTM tech builds SLMs - Small Language Models.

They own them. They iterate on them. They master them.

They use a HUGE set of data about a NARROW topic to train an SLM.


Examples:

👂🏼 Conversation Intelligence → what if WINN.AI takes all that valuable conversation data and build a specific SLM just for sales conversations? 

📨 Sales Engagement → what if Outreach takes all the email content data and builds a specific SLM for prospecting emails?

🗣️ Marketing Automation → what if HubSpot takes all their inbound content data and builds a specific SLM for nurture?

🌐 Visitor Deanonymization → what if RB2B takes all their website visitor data and builds a specific SLM for better identification and conversion?

👤 Contact Data → what if ZoomInfo takes all their contact data and builds a specific SLM for understanding buyer committees?


The current landscape can’t continue. 

If you are a AI GTM tech company, I’d be investigating how to turn my data set into an SLM that nobody can copy, use, steal, or compete with.

Jonathan Moss

Turning GTM into a revenue factory through recurring impact & helping GTM leaders do the same | GTM Executive leveraging AI ✨ | Keynote Speaker & Advisor | Podcast 🎙️ Host |

I created the Revenue Nervous System™ a framework for building intelligent, self-improving growth engines that learn, adapt, and scale. The IMPACT+ Framework™ is how you implement it: a structured approach for embedding AI across the revenue lifecycle and turning strategy into execution.

I help Founders, Business Executives and Go-to-Market leaders transition to AI-native GTM and architect Agentic Growth Systems that eliminate friction and turn AI into measurable performance outcomes across their business.

At Experity, I serve as EVP & GM for Patient Engagement and Care Agent, the industry's first AI-native clinical intelligence platform. We're transforming how urgent care connects with patients through a multi-agent system that automates communication, personalizes engagement, and unifies clinical + commercial workflows across 3,000+ clinics.

Think of it as the Orchestration and Routing Layer for Healthcare: where data, context, and intelligence finally work together.

Beyond Experity, I work with founders, executive teams, and AI-native startups on growth architecture, GTM execution, and multi-agent system design.

Quick professional stats:
→ 87% median growth rate across all of my roles
→ $1B+ capital raised across portfolios
→ Scaled revenue $0 → $15B
→ Backed by Sequoia, Lightspeed, GTCR, Warburg Pincus, SoftBank

Core expertise: Revenue Nervous System™ | IMPACT+ Framework™ | Agentic AI Growth Strategy | AI-Native GTM | Patient Experience & Healthcare AI

Executive Member

Sprouts.ai named some amazing people and somehow I was included in the top 20 GTM Leaders to Watch in 2026.
 
Grateful. Also a little on the hook now.
 
Because if there's one thing I'm convinced of heading into this year, it's that the GTM leaders worth watching won't be the ones with the most tools, the most initiatives, or the loudest voices.
 
They'll be the ones who figured out how to build systems that actually compound.
 
Here's what I keep seeing: teams stall out not because they lack effort, but because they're running on architecture that was never designed for how buying actually works today.
 
Buyers are more informed. Cycles are less predictable. AI is raising the output bar for everyone.
 
In that environment, staying busy is easy. Knowing what matters is hard.
 
What I've found is that most GTM breakdowns trace back to the same root cause: the system doesn't remember, doesn't learn, and doesn't coordinate in real time.
 
You fix ICP in January. By March, the signal has shifted and nobody updated the model.
 
You run a feedback loop between sales and marketing. But nothing gets captured, so the same mistakes repeat six months later.
 
You add an AI tool. But it's bolted onto a broken process, so you get faster noise instead of better decisions.
 
The teams winning in 2026 aren't just using AI differently. 

They're building differently, around layers of intelligence that compound over time instead of resetting every quarter.
 
Clear context. Institutional memory. Coordinated execution. AI that amplifies judgment instead of replacing it.
 
That's the work. And honestly, it's still early enough that the teams who start now will have a gap that's very hard to close later.
 
Grateful to be in a community of operators who are helping make sense of the GTM foundations, what is working on the "field" and how to manage this transition into the AI-era. 

Some of my good friends Elizabeth McCalley.eth 🦋, Koen Stam 🐰 Jen Igartua, Stephanie Valenti, Steph Reck ⚓, Dale Zwizinski, Sophie Buonassisi ,Sarah Allen-Short, Kiva Kolstein 
 
Strong company to be in.
 
If you're building GTM for 2026, make sure you to follow these great leaders and tell us what is on your mind.

EVP @ Experity | AI-Native GTM Architect | Revenue Nervous System™ | Care Agent | Healthcare AI

Anthropic published new research last week that should change how every GTM leader thinks about their headcount plan.
 
Not "AI might affect jobs someday." Right now. Today.
 
The paper introduces something called "Observed Exposure" 

Combining what LLMs can actually do with how Claude is being used in real work, weighted toward tasks that can be automated. 

Then mapping that against what specific jobs require people to do all day.
 
The GTM numbers:
 
→ Customer service reps: 70.1% AI exposure 
→ Market research analysts: 64.8% 
→ Sales reps: 62.8%
 
That's not theoretical capability. 

That's real-world deployment, mapped against what sales reps actually do all day. Today.
 
But here's the part nobody's talking about.
 
For computer and math roles, AI theoretically can handle 94% of tasks. 

Real-world observed deployment? 33%. That's a 61-point gap sitting there, untapped.
 
That 61-point gap belongs to someone. 

The question is whether you're the one closing it or the one getting closed on.
 
There's another finding buried in the BLS data overlay: for every 10 percentage point increase in AI exposure, employment in that occupation is projected to grow 0.6 points less through 2034.
 
Sounds small. Run it across your entire revenue org and it isn't.
 
But here's what actually surprised me: companies aren't doing layoffs. 

They're doing something quieter. Not backfilling roles. Not approving the next SDR class. 

Running the same RevOps team for three years instead of growing it. 

Letting attrition do the work that announcements would make public.
 
That's how structural labor market shifts actually happen. Not with a headline. With a headcount plan that never gets approved.
  
The GTM leaders getting ahead of this aren't asking "can AI help my team be more productive?" 

They're asking: "What does this role need to look like if AI handles [X] at scale?"
 
Different question. Different answer.

Most GTM awards celebrate the big number.
 
The quarter you 3x'd pipeline. The campaign that went viral. The deal that saved the year.
 
But here's the thing nobody talks about — big numbers built on shaky systems don't compound. They collapse.
 
I've watched it happen more times than I can count across 21+ years of scaling companies. A team absolutely crushes a quarter, leadership throws a party, and six months later they're trying to figure out why the machine stopped working. 

Turns out there was no machine. Just a bunch of really talented people grinding really hard.
 
That's exactly why I'm proud to be on the judging panel for the Winning by Design Growth Awards at Impact Summit 2026.
 
Because this isn't that kind of award.
 
Three categories. Each one matters.
 
→ 𝗕𝗼𝘄𝘁𝗶𝗲 𝗳𝗼𝗿 𝗚𝗿𝗼𝘄𝘁𝗵 𝗠𝗼𝗺𝗲𝗻𝘁𝘂𝗺 — recognizes leaders who've built across the full customer journey. Acquisition, onboarding, retention, expansion. Not optimized one piece while the rest leaks.
 
→ 𝗦𝗣𝗜𝗖𝗘𝗗 𝗳𝗼𝗿 𝗚𝗿𝗼𝘄𝘁𝗵 𝗔𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗶𝗼𝗻 — for operators who've made their teams faster 𝘸𝘪𝘵𝘩𝘰𝘶𝘵 making them sloppier. Velocity from alignment, not from pressure.
 
→ 𝗔𝗜 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗼𝗿 — this one's close to my heart. Not for teams that "use AI." For leaders who've embedded it into the operating logic of how they grow. AI with intent, not AI as a checkbox.
 
The judges? Jacco van der Kooij. Kyle Norton. Sydney Sloan. Jared Brickman. And me somehow.

Leaders who've actually built and scaled systems at this level.
 
We're not looking for a good story. We want evidence. Before-and-after benchmarks. Real metrics. What actually changed and why it held.
 
𝗡𝗼𝗺𝗶𝗻𝗮𝘁𝗶𝗼𝗻𝘀 𝗰𝗹𝗼𝘀𝗲 𝗠𝗮𝗿𝗰𝗵 𝟭𝟴𝘁𝗵.
 
If you're a GTM leader who's done the hard, quiet, unsexy work of building growth infrastructure that compounds — or you know someone who has — this is for you.
 
Not for the headline. For the system behind it.
 
👉 Link in the comments to nominate.
 
Who in your network has built growth the 𝘳𝘪𝘨𝘩𝘵 way this year? Tag them below — they deserve the recognition.

Google's VP of Startups just said the quiet part out loud.

LLM wrappers and AI aggregators have their "check engine light" on.

I've been saying this for months. But hearing Darren Mowry from Google Cloud put it that bluntly? That hits different.

Here's what he means: if your entire AI startup is a nice UI sitting on top of GPT or Claude, you're toast. 

The model providers are coming for your lunch. They always were.

The startups that survive? They have vertical depth. 

Cursor doesn't just "use" AI—it understands how developers actually write code. 

Harvey AI doesn't just "chat"—it knows how law firms think about case research.

I've spent the last year building AI-native GTM systems. Not wrappers. Not aggregators. Systems where AI agents run actual operations—prospecting, content, competitive intelligence, pipeline management.

And the difference is night and day.

A wrapper adds a step. An operations-native system removes twelve.

The moat was never the model. It was never the prompt engineering. It's the workflow. 

It's understanding—really understanding—how a specific type of work actually gets done, and then building intelligence into that process.

After 21+ years of scaling operations, I can tell you: this is the same pattern every technology wave follows. The shiny demo phase dies. The "actually runs my business" phase lives.

We're entering that second phase right now.

If you're building on AI, ask yourself: would your product survive if the underlying model got 10x better and the provider launched their own version of your feature?

If the answer is no, that check engine light is for you.

IYKYK - you just need a microphone and this new keyboard.

A common question I get: "what is the difference between a Skill and an Agent?" 

I have built 80+ AI agents for Business across GTM, Product, Engineering, etc. 

I built a guide to help - comment and I will send it to you.. ok back to the program..
 
Agents are like employees in departments: Sales, Marketing, CS. 

Vertical. Domain-specific. They own a function and make decisions.
 
Skills are like apps those employees use: horizontal, reusable across departments, any agent can pick them up.
 
Tools let agents act. 

Skills provide the knowledge of how and when to act: including your company-specific context.
 
That last part is the whole game.
 
Because raw capability is not competence.
 
Your AI can call APIs, analyze data, generate content, and orchestrate workflows. 

That's capability. 

But does it know your team's specific 8-step deal qualification process? 

Does it know anything over $50K hits Deal Desk? 

Does it know the naming conventions, the approval chains, the compliance requirements?
 
That's competence. And that's what skills provide.
 
The architecture is elegant:
 
• Agents carry full context constantly (expensive, bloated)
• Skills load progressively — metadata at startup (~100 tokens), full instructions when activated, resources only when needed
• One skill serves many agents
• Agents call skills. Skills can invoke agents. Bidirectional.
 
No single agent carries all that procedural knowledge. They just know what skills exist and call them when needed.
 
If you're building simple bots, tools are enough. 

If you're building digital employees, you need skills. 

The bottleneck is never whether the agent can perform an action. 

It's whether the agent knows YOUR specific process for performing it correctly.
 
Agents are the who. Skills are the how.
 
Stop building capable agents. Start building competent ones.
 
What's the biggest gap between what your AI agents can do and what they actually do well? Drop it below.

The opportunity is massive for the builders of agents in back office, GTM and may other places in business. 

Gonna be a fun next couple of years! 

Let’s cook! 👨‍🍳

A CEO told me last week he'd never tried Claude Code.
 
"That's for technical people, right?"
 
He uses AI every day. And he's been scrolling past the most powerful tool in his stack because of how it was positioned.
 
He's not alone. I hear this from revenue leaders, ops executives, and founders constantly.
 
I put an operator's guide together - let me know if you want it. 

Let me tackle this directly. 
 
Claude Code is not a code editor. It's an AI operating system.
 
The terminal is just where you type. That's it. 

Just like you type a text message to your friends or loved ones. 
 
Here's the mental model I use to explain the difference between Code and Cowork:
 
𝗖𝗼𝘄𝗼𝗿𝗸 is a smart contractor you hire for a day. They do great work. They leave. Next session, you re-explain everything from scratch.
 
𝗖𝗼𝗱𝗲 is a full-time operator who sits at your desk every day. They remember everything. They build systems. They get better over time. Every session starts where the last one ended.
 
Same model underneath. Completely different relationship.
 
Four things make Code different. None of them require writing code.
 
1. CLAUDE.md = Your Business Operating System
  
In Cowork, every session resets. You re-explain. You re-upload. You re-establish context.
 
In Code, it's architecturally guaranteed to load. The compound effect over months is significant.
 
2. Named Agents = Your AI Team
 
Cowork sub-agents are ephemeral — one task, then gone.
 
In Code, you build agent specs. Your CMO agent knows your positioning. Your Competitive Intel agent knows your rivals and preferred format. You build it once. You refine it over time.
 
I run 90+ agents. When I need a competitive battlecard, the agent already knows everything. What used to take an hour takes 7 minutes.
 
3. Your File System Is Already There
 
Cowork works in a sandbox. Code works in your actual world.
 
"Read the pipeline export in Downloads and tell me which deals are at risk this quarter."
 
You type that. It reads it. No upload prep. No copy-paste. No re-explaining where things live.
 
This sounds minor until you've done the upload dance three times in one week.
 
4. Automation That Actually Runs
 
Cowork's scheduled tasks are beta and unreliable right now. They work sometimes. You can't build a business process on "sometimes."
 
Code supports hooks, cron jobs, and full automation pipelines. 

Things I have running in production today:
 
→ Daily competitive monitoring that flags pricing changes 
→ Automated content pipeline: research → draft → review 
→ Weekly research digests from 10+ sources synthesized automatically 
→ Pipeline health alerts triggered by CRM events
 
None of those required writing code. They required describing what I wanted.
 
"But I'm not technical"
 
This isn't either/or. 

I use Cowork for quick documents and one-off tasks. 

I use Code as the operating system underneath — context, agents, automation, memory.

What Cassie Young said! 

When she talks you should 👂! 👇

Partner at Primary Venture Partners, Seasoned Enterprise CRO/CCO

🌶️🌶️ CROs, there's a GTM reckoning coming for the board room. Your board members are done rewarding your AI experimentation efforts.🌶️🌶️

The efficiency gains of the last two years were a half-truth. Sure, you trimmed the fat, but the GTM engine has been atrophying for most companies.
Here's the cold hard truth: GTM inefficiencies used to hide behind 80% SaaS margins but in the age of AI, there is nowhere left to hide. As AI compute costs eat into gross margins, you can no longer afford higher operating expenses.

In 2026 Boards will stop rewarding AI experimentation and start demanding P&L impact through the PRIME framework:
⚙️ Productivity: Is ARR/revenue per GTM head actually climbing?
📈 Retention: Are AI investments moving the needle on GRR/NRR?
💲 Investment Efficiency: Why is CAC payback still stuck at ~16 months?
🪴 Momentum: Is growth predictable and coming from a healthy mix of new and existing?
⬇️ Expense Reduction: Have you actually eliminated labor or software costs?

Most CROs use AI for "marginal gains"—better meeting summaries or slightly faster cycle times. That won't save your P&L. The winners are thinking transformationally:

🟢 Jeanne DeWitt Grosser's team at Vercel replaced a 10-person inbound SDR team with one "Lead Agent"
🟢 Kyle Norton's team at Owner.com 2x'ed productivity per rep with their AI operating system
🟢 Matt Braley’s team at EliseAI had three AEs close $50M in new ARR by supercharging their talent with AI infrastructure
🟢 Amanda Kahlow and the 1mind team use the same AI Superhumans they sell to fuel their own growth, comfortably surpassing the age-old T2D3 growth thresholds with a <6-month CAC payback

PRIME has become our non-negotiable for GTM investing, and it needs to be yours for GTM operating. Are you ready to be held accountable for the P&L? 👀👀

Spent an hour watching 🚀 Victor Adefuye run 3 crazy workflows in Claude Cowork and its freakin rad. 

Want to see it? Comment below.

Victor has been in GTM enablement for over a decade. 5 years as an MD at Winning by Design. 

Teaching reps how to write emails, run discovery, execute MEDDPICC. 

The guy knows what "good" looks like better than almost anyone.Then Cowork dropped. 

And he turned 10 years of enablement IP into AI agents.

Here's what he showed my man Jonathan M Kvarfordt, MBA live on the GTM AI Podcast .

He pointed Cowork at 360 conference leads and typed 3 sentences. Cowork spun up 15 parallel sub-agents. 

Each one independently researched a company, scored it against his ICP, and wrote personalized nurture emails that pulled from his 110-page content library. Matched by pain point.

He fed 7 sales calls into a folder. A custom MEDDPICC skill scored every single one in parallel. Individual scorecards with direct quotes and timestamps. 

Then a synthesis agent rolled up a team-wide gap analysis showing exactly where reps were weak: economic buyer, champion development, paper process.

And then came the one that broke my brain.

Victor is NOT a developer. Never has been. 

But in 2 weeks he built 7 𝗠𝗮𝗸𝗲.𝗰𝗼𝗺 automations by having a conversation with Cowork. 

It opened Make in his browser. Selected the modules. Configured connections. Wrote the prompts. One automation finds newly hired CROs every week and builds drip campaigns off the trigger. Built through chat.

He said something I can't stop thinking about: "The same training I used to give to humans and hope they would remember, I now give to AI. It's more reliable.

"That's the whole game right there. Your enablement playbook IS your AI moat. The people who spent years defining what good looks like are now the most dangerous builders in the room.

Stop prompting. Start packaging your expertise.

All 3 workflows are broken down step-by-step in my latest newsletter. Full playbook. Skill templates. MEDDPICC scoring rubrics. 

Everything you need to build these yourself this week.

Want to see it in action and get the full playbook?

-Like this post--Comment "Cowork Playbook"---Repost if you know a GTM leader who needs this

Or get it here: 𝗵𝘁𝘁𝗽𝘀://𝗹𝗻𝗸𝗱.𝗶𝗻/𝗴𝗚𝗰𝗷𝗤𝗸𝗝𝗱

Enjoy!

Enterprise AI adoption just jumped 7x in a single quarter.
 
Not 7%. Seven times.
 
JPMorgan Chase is producing investment banking decks with AI agents in 30 seconds. 

BMW is running multi-agent systems across manufacturing plants—not as a pilot. In production.
 
The chatbot era is over. Most people just haven't noticed yet.
 
Multi-agent setups handling competitive intel, content, pipeline research, and prospect analysis—coordinated, not siloed. 

And here's the thing nobody talks about in the breathless coverage:
 
The hardest part has never been the technology.
 
It's always the humans.
 
A YC startup called Trace just raised $3M on exactly this insight. Their data shows 2-3x higher agent utilization when teams design the human-AI handoff intentionally.
 
I've seen this same movie. CRM. Marketing automation. Cloud. Analytics. 

The tech always works. The org is always the bottleneck.
 
Three things I've learned the hard way building these systems:
 
→ Start with the workflow, not the tool. Map the actual process step-by-step. Find where a human adds the least value. That's where your agent goes.
 
→ Graduated autonomy is mandatory. Draft before publish. Recommend before decide. Flag before act. The leash gets longer as trust builds. Skip this step and you break things. (Ask me how I know.)
 
→ The real metric isn't task completion. It's time-to-decision. My agent system doesn't replace strategic analysis—it eliminates the 4-6 hours of research that used to happen before the analysis. The human still decides. Just faster, with better inputs.
 
Gartner says 40% of enterprise apps will have embedded AI agents by end of 2026. 

In September 2025, that number was 5%.
 
The companies that spent 2025 experimenting are now operationalizing. 

The companies that spent 2025 watching are now scrambling.
 
That gap compounds every quarter.
 
What's your first workflow you'd hand to an agent?

What would you attempt to do if you knew you would not fail?

Important question to ask yourself especially with all the opportunity right now. 

State your goal, go achieve it. 

Regardless of the outcome you have succeeded. 

The outcome is always learning, experience, connections, and many things you would have never accomplished without going for it.

AI has changed from a chat back and forth to a coworker.. catch up on everything that happened this week 👇

Open article: The Week AI Became Your Coworker by open.substack.com

View newsletter: GTM AI Podcast & Newsletter

Open article: 3/6/26: This Founder Tried Every AI SDR. They All Failed. What He Built Instead Converts at 90% Open Rates. by Jonathan M Kvarfordt, MBA • 4 min read

3/6/26: This Founder Tried Every AI SDR. They All Failed. What He Built Instead Converts at 90% Open Rates.

They said ideas are cheap and execution is king. AI has flipped that for the most part.

OpenAI is a $730 billion company doing what every $730 billion company eventually does.
 
Abandoning the product that made it famous to protect the product that actually makes money.
 
ChatGPT's web traffic share dropped from 86.7% to 64.5% in twelve months. 

Google Gemini went from 5.7% to 21.5% over the same period. 

DeepSeek is matching GPT-5 performance at 1/30th the cost.
 
When the commodity version of your product is 97% as good and nearly free, you don't have a product problem. 

You have a business model problem.
 
Meanwhile, the numbers on the enterprise side are going in the opposite direction:
 
→ Codex crossed 1.5M weekly active users with 20x growth since August 2025 
→ Enterprise seats grew 9x year over year 
→ 92% of Fortune 500 companies are using ChatGPT products
 
Revenue per enterprise user dwarfs consumer subscriptions. 

Switching costs are orders of magnitude higher. 

And the customer who has Codex embedded in their engineering workflow doesn't cancel because a cheaper model exists.
 
Fidji Simo ran Instacart. She's been inside brutal unit economics, where every order can lose money if you're not obsessive about the math. She knows exactly what "the consumer unit economics are breaking" looks like from the inside.
 
So she told the all-hands: stop chasing side quests. Go win where it matters.
 
That's the right call. But it comes with a sentence nobody's saying:
 
OpenAI's $280 billion revenue target by 2030 requires consumer subscriptions at $20/month to become the foundation of an enterprise infrastructure play. 

Those are different companies. 

The mission was AGI for humanity. 

The strategy is Codex for Fortune 500 engineering teams.
 
You can run both narratives for a while. Eventually the spreadsheet picks one.
 
The interesting question isn't whether enterprise AI is the better business (it obviously is).
 
It's whether OpenAI gets there before Anthropic locks up enough Fortune 500 contracts to make the race irrelevant against a projected $143 billion in cumulative losses and a market that's compressing faster than anyone expected.
 
That's not a rhetorical question. It's the actual race.

taplio

The Battle

Mark Kosoglow

Jonathan Moss

Followers

Posts

Engagements

Likes

Performance ScoreMark Kosoglow's LinkedIn Performance

Performance ScoreJonathan Moss's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Mark Kosoglow

Jonathan Moss

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Mark Kosoglow's LinkedIn Performance

Performance ScoreJonathan Moss's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreMark Kosoglow's LinkedIn Performance