The Cold LinkedIn Message That Gets Strangers to Take Your Call
Most founders run LinkedIn outbound and get ignored. Julius Körfgen used it to open his first sales conversations at Uplane, from a standing start with no produ
Julius Körfgen, Uplane
He Sold the Software Before He Built It. $1M ARR in 6 Months
HobbesSponsor
Stop losing buyers to your demo form
Visitors who aren't ready to talk to sales just close the tab. Hobbes gives them a personalized demo on the spot, 24 hours a day.
Like this episode?
Get real founder strategies for the AI era. Delivered weekly.
Free weekly newsletter · No spam
Episode Summary
Julius Körfgen built around ten thousand ads by hand as a marketing director, one at a time. So he left to build software that would do the work for him, and then sold it to his first customers before writing a line of code. Uplane reached a million dollars in ARR in about six months.
The mechanic was the same every time. A cold LinkedIn message asking to learn, a discovery call where he only asked questions, and a promise to come back in one week with a solution. Then he and his two co-founders would build whatever they had just promised.
Julius Körfgen spent years running marketing at a European startup, where he personally built around ten thousand ads. Writing headlines, testing image variants, checking which call-to-action buttons converted. He estimates that roughly ten percent of ads are ever profitable, and that companies keep running the other ninety percent because the process of finding out is slow and manual.
So he left with two co-founders to automate it. What he did not do was build first. He ran cold LinkedIn outreach to strangers, framed every message as a request to learn rather than a pitch, and used the discovery calls to find out what marketing teams actually struggled with. At the end of each call he would say the same thing: give us a week and we will come back with a solution. Then the three of them would grind out whatever they had just promised.
He also refused to run free pilots. His argument is that without a dollar attached, you cannot tell the difference between a real business case and a polite conversation.
Uplane reached a million dollars in ARR in about six months and now runs around twenty people across San Francisco and Berlin. AG1 is a customer, and a project with Deutsche Bahn is underway after roughly nine months of conversations.
In this interview, Julius breaks down the outreach that got strangers to take his calls, the one-week sprint, why he threw out per-seat pricing in favour of a fixed fee that only covers costs plus a variable share of ad spend, and the guardrails that stop an AI from putting bad ads in front of an enterprise brand's audience.
Key Insight
Uplane co-founder Julius Körfgen reached a million dollars in ARR in about six months by selling to customers before writing any code, running cold LinkedIn outreach framed as research rather than sales, then building a working demo within one week of each discovery call and charging for it rather than offering a free pilot.
Key Ideas
- Sold to first customers before writing a line of code
- Cold LinkedIn outreach framed as a request to learn, not a pitch, at roughly a 5% response rate
- One week from discovery call to a working demo, built by the three co-founders
- No free pilots. A dollar has to be attached or you cannot tell a business case from a polite conversation
- Pricing is a fixed fee covering costs plus a variable share of ad spend, so Uplane earns more only when the customer's ads perform better
Key Lessons
- 🤝 Sell before you build: Julius closed customers before writing a line of code. His discovery calls ended with a promise to return in a week with a solution, which forced both a real deadline and a real answer about demand.
- 🎯 Frame outreach as learning, not selling: His cold LinkedIn messages said he had just left his job and was exploring an idea, and asked for a few questions. People opened up about problems they would never have shared with a pitch.
- 💰 Never run a free pilot: Without a dollar attached you cannot tell a business case from a polite conversation. Julius has watched founders stay attached to an idea for months because nobody ever asked them to pay for it.
- ⚡ A week is long enough to build the thing you promised: Three founders and one week produced demos that won real customers. Scrappy was fine; fake was not, and he argues AI removes the excuse for a mock-up that does nothing.
- 💰 Align pricing with the outcome you claim: Uplane charges a fixed fee covering costs plus a variable share of ad spend. Julius says it makes the pitch easier, because he only earns more when the customer's campaigns do better.
- 🏢 Be reachable faster than an agency can be: Uplane answers customers within 120 seconds. Julius treats speed of response as the main structural advantage an early-stage company has over an incumbent agency.
- 🧠 Volume is not the constraint anymore: AI made producing ads nearly free, so the bottleneck moved to picking the roughly ten percent that perform. Companies pushing more output without connecting it to analytics are solving the wrong half.
Playbooks from This Episode
Never Give Your First Customers a Free Pilot
A free pilot feels like the safe way to start. Lower the barrier, get someone using the thing, prove the value, then charge later.
Why He Threw Out Per-Seat Pricing and Charges for Results
Charge per seat. It's the SaaS default.
Sell the Software Before You Build It
Most founders build first, then go looking for someone to buy it. Julius Körfgen ran it backwards.
Watch the Episode
Episode Q&A
How did Julius Körfgen get Uplane to a million dollars in ARR in about six months without building the product first?
He sold before building. Cold LinkedIn outreach got him discovery calls, he ended each call by promising a solution within a week, then he and his two co-founders built exactly what they had promised and charged for it.
What did Julius Körfgen say in his cold LinkedIn outreach to get strangers to take his calls?
He introduced himself as a founder who had just left his job and was exploring an idea in their field, and asked for a few questions rather than a meeting. He describes the tone as humble and grateful, with a response rate around five percent.
Why does Julius Körfgen refuse to give first customers a free pilot?
Without a dollar sign attached you cannot tell whether you have a business case. He has seen founders stay attached to an idea for too long because nobody ever asked them to pay, and the budget conversation is what exposes it.
What is the one-week sprint Uplane used to win its first customers?
At the end of a discovery call he would propose reconvening in a week with a solution. He and his co-founders would then build a scrappy but working version of what the customer had described, and demo it.
Why did Uplane throw out per-seat pricing?
Julius argues seat-based pricing does not align incentives. Uplane charges a fixed fee that covers operating costs at little or no margin, plus a variable success fee tied to a share of ad spend, so the company earns more only when campaigns perform.
How does Uplane handle attribution on performance-based pricing?
Uplane charges a percentage of ad spend rather than trying to claim credit for revenue. The logic customers accept is that if the campaigns work, they will move more budget onto them. Clients get daily reporting on every ad and its cost per result.
How does Uplane stop AI from producing off-brand ads for enterprise customers?
Through what Julius calls atomic content. Brand and compliance guidelines, reference ads, product descriptions and ERP data are fed in as constraints, evaluations run on generated ads, and an account manager reviews output before it goes live.
Why does Julius Körfgen say most companies using AI in marketing are getting it wrong?
They use AI to produce far more content without connecting it to the analytics that show what works. The volume goes up and the relevance stays flat, so the constraint moves from producing content to selecting the small share that performs.
What is Uplane's response-time rule for customers?
One hundred and twenty seconds. Julius treats being reachable as an early-stage company's main advantage over a marketing agency that takes days to reply.
Book Recommendations
Related Episodes
Marius Meiners, Peec AI
Building an AI Search Optimization SaaS to $8.6M ARR in 14 Months
Marius Meiners is the co-founder and CEO of Peec AI, a platform that helps marketing teams track how their brands appear on AI search tools like ChatGPT, Perplexity, and Gemini. After studying economics, working in venture capital and M&A at PwC, and joining Antler's Berlin cohort, Marius found himself with no team, no idea, and four years removed from writing any code. Then in late October 2024, ChatGPT launched search. Marius saw it and decided this was going to change everything. The smartest SEO experts in the world were already obsessed with it. The signal was loud. So he turned to AI search optimization as the wedge - a category that would explode as marketers scrambled to figure out how to get cited by AI assistants. He vibe coded the first prototype with V0 in a day and a half. Eight customers signed letters of intent based on it. Antler wrote a 100K check. His CTO joined and built the real product in six weeks. Peec launched in February 2025. Then came the bet. Their biggest competitor had raised five times more money and was chasing the world's biggest brands. Marius made the opposite call. Peec priced at 85 euros while competitors charged over 500. For six months, Marius and the team ate two-euro canned food every day, wondering if the mid-market AI search optimization play would ever pay off. Today Peec has over 2,000 customers, $8.6 million in ARR, and a team of 55. All in 14 months. AI search optimization went from speculation to a live revenue channel - 20% of Peec's own conversions now come through AI search itself.
Eugene Cheah, Featherless AI
The Weekend Experiment That Killed His Own AI Model
Eugene Cheah's team had built something genuinely novel: an open source foundation model under the Linux Foundation, trained across more than 200 languages, with an architecture that made AI inference dramatically cheaper. Their seven billion parameter model beat Llama's equivalent. The problem was that almost nobody was asking for it. Along the way they solved a constraint of their own making. Customers were fine-tuning thousands of models on their platform, the industry norm was one GPU per model, and they could not afford a thousand GPUs. So they built a system that swaps models on and off GPUs on demand, bringing a cold model online in about five seconds. Most inference providers keep a fixed list of under a hundred models standing by, because loading one can take thirty minutes on hardware costing eighty dollars an hour. Then someone on the team asked whether the same technology would work for Llama and Mistral. They shipped it as an experiment. Over the launch weekend it earned more than the platform they had spent two years on. Eugene renamed the company and went all in. In this interview, Eugene explains why he priced a flat monthly rate while the rest of the AI industry charged per token, how stripping the technical explanation off the homepage kept improving conversion until they removed their own research from the top of the page, and why competing for the long tail of open source AI models beats fighting a hundred providers over the top hundred.
Farzad Rashidi, Respona
He Demoted His SaaS to Sell a Service and 4x'd Revenue
Farzad Rashidi is the co-founder of Respona, a company that helps brands get cited in AI answers across ChatGPT, Perplexity, and Google AI Overviews. He first came on the show back in episode 323, when Respona was a self-serve outreach tool doing a few hundred thousand in ARR. Then the classic bootstrapped trap set in. Churn caught up with new business, and every customer they won was offset by one they lost. For years Farzad tried to fix it the way most founders do, by adding more features to make the product stickier. Nothing moved. The real reason customers left was not missing features. It was that they never had the time to do the work the tool required. The turning point came in early 2025. A marketing agency CEO haggled over an $800-a-month license, then offered to pay per result instead. That one conversation nudged Farzad toward a service-as-software model: do the work for the customer, charge for the outcome, and use the software in the back end. That first customer now spends around $65K to $70K a month, and in twelve months the company 4x'd the revenue it had spent six years building. What makes this a service-as-software story rather than a slide back into agency work is what came next. Farzad demoted the self-serve SaaS on the homepage, productized the service into fixed tiers with no negotiation, and rebuilt a software layer (a client portal, a publisher network, and a brain in the middle) on top of the manual delivery. We also dig into the actual playbook for getting a brand cited in AI answers, from finding lookalike publishers to building a surround-sound presence around the models. I hope you enjoy the conversation.