The First 100 Customers Course #15: How Uplane Turned a One-Week Demo Into Its First Paying Customer

TL;DR

Uplane did not wait for a polished AI platform before testing demand. Founder Julius Körfgen says the three founders contacted strangers on LinkedIn, used the first call to learn rather than pitch, promised to return in one week, and built a real solution around the buyer's problem during that sprint.

An unnamed Berlin startup became the first paying customer. The amount is private, and no public source establishes a first-10 or first-100 count. Körfgen later reported that Uplane reached approximately $1 million in annual recurring revenue about six months after launch, but that revenue figure is not independently audited and should not be attributed to LinkedIn alone.

The reusable mechanism is lived pain → humble cold discovery → seven-day promise → real output with manual work behind it → payment gate → rapid founder support. The warning matters as much as the sequence: a free pilot can prove that someone likes your attention. It cannot prove that the problem has a budget.

What you will build

This course produces a paid-discovery loop for a founder who can build quickly but does not yet have customers. You will choose a narrow operational pain, recruit five discovery calls from a small cold list, convert one credible problem into a seven-day build commitment, deliver a working outcome without disguising manual work, and ask for money before expanding the product.

Your output is not a feature roadmap. It is an evidence ledger with one row per prospect: observed workflow, cost of the problem, promised output, build deadline, manual steps, payment decision, activation event, and next review. Every row must end in a pass, revise, or stop decision.

Who this is for, and who should not copy it

Use this motion when the founders understand the buyer's work, can produce a narrow result in seven days, and can remain personally involved during delivery. It fits workflow software, AI services becoming software, internal-tool replacements, and products where an ugly result can still save real time or money.

Do not use it for safety-critical software, regulated decisions, infrastructure migrations, or any product where a rushed prototype can damage data, money, health, or compliance. Do not promise seven days because AI makes coding fast. Promise it only when the useful output is small enough to verify and reverse.

Case snapshot and evidence limits

Stage	What the evidence supports	Limit
Founder insight	Körfgen says he had built roughly 10,000 ads manually before Uplane	Founder estimate, not an audited activity log
First-buyer channel	Completely cold LinkedIn outreach to a Berlin startup	Customer name and outreach volume are private
Validation loop	Discovery call, one-week return promise, real scrappy solution	No funnel-wide conversion data disclosed
Payment	The first customer paid; free pilots were rejected	Deal value is undisclosed
Later revenue	Approximately $1M ARR after about six months	Founder-reported; not independently audited
Independent context	$4.5M seed, YC F25, named later enterprise customers	Later traction does not prove the first-customer attribution

The independent record is useful but narrower than the podcast headline. Y Combinator lists Uplane as an active Fall 2025 B2B SaaS company with three founders. Business Insider reported a $4.5 million seed round, a 15-person team in April 2026, a target segment spending at least $100,000 per month on digital ads, and later customers including Deutsche Bahn. The article attributes operating-performance claims to Körfgen, so this course does too.

The model: turn a conversation into a falsifiable week

Most “sell before you build” advice ends at preorders or mockups. Uplane's useful move was different. Discovery created a specific deadline. The deadline forced the founders to choose one problem. A real output made the second conversation concrete. Payment separated urgency from politeness.

1. Access: a relevant stranger accepts a learning conversation.
2. Specificity: the founder can restate one expensive workflow in buyer language.
3. Deadline: both sides agree to review a working result in seven days.
4. Reality: the result works, even when humans perform hidden steps.
5. Payment: the buyer spends money to keep receiving the outcome.
6. Activation: the buyer uses the result in the real workflow.

A founder should not advance because a call felt good. Advance only when the next gate produces observable evidence.

Step 1: start from a pain you can inspect

Körfgen had operated the workflow himself. That did not remove the need for customer discovery, but it gave him a vocabulary and a bounded hypothesis. Write down the last ten times you performed the target job. Record the inputs, tools, handoffs, delays, errors, and the person who felt the cost.

When [role] tries to [job], work stalls at [handoff] because [missing information or manual step]. We can return [specific output] within [time] if we receive [inputs].

Pass condition: you can show five concrete instances and name the person who owns the result. Stop condition: the pain exists only in your imagined future, or the person experiencing it cannot authorize even a small purchase.

Step 2: build a 25-account discovery list

The transcript gives an illustrative 5% reply rate, not a measured first-customer funnel. Use it as a capacity assumption, not a promise. Start with 25 accounts that share the same role, workflow, and trigger. A small homogeneous list lets you learn whether your hypothesis is wrong; a large mixed list hides that lesson inside averages.

Required field	Admission rule	Reject when
Role	Owns the target workflow or budget	Only a generic senior title
Trigger	Visible change makes the job timely	No reason this month differs from last month
Evidence	Source dated within 90 days	AI-generated biography or guessed intent
Scope	One result can be produced in seven days	Requires an enterprise-wide rollout

Pass: 20 of 25 accounts meet every rule. Revise: fewer than 20 qualify. Tighten the role and trigger before enriching contacts or writing messages.

Step 3: ask to learn, not to disguise a pitch

Uplane's early opener was direct: the founder had left his job, was exploring an idea in the recipient's field, and wanted to ask questions. It worked because the call behaved like the request. The founder did not use “research” as a false wrapper for a finished sales deck.

Hi [name], I worked on [workflow] from the operator side and I am testing one narrow idea around [problem]. I am not asking you to evaluate a product. Could I ask how your team handles [specific handoff] today? Twenty minutes is enough.

Ask about the current process, recent failure, workarounds, time cost, budget owner, and what a useful output would look like. Do not show features during the first ten minutes. Do not ask whether they “would use” an imaginary product.

Pass: the buyer describes a recent incident, quantifies a cost, and agrees on a reviewable output. Stop: the conversation stays at trends, compliments, or hypothetical interest.

Step 4: close the call with a seven-day contract

The one-week promise is the mechanism's hinge. It converts discovery into a falsifiable commitment without pretending the whole product exists.

• Write the one output the buyer will inspect.
• List the inputs the buyer must provide by day one.
• State which steps will be manual.
• Set a 30-minute review exactly seven days later.
• Agree on the price or paid-pilot decision before building beyond the demo.

By [date], we will return [working output] using [buyer inputs]. [Manual steps] will still be done by us. At the review, you will test [activation event]. If it works, the next phase costs [price] for [scope]. If it does not, we stop.

Pass: the buyer supplies inputs and accepts the review. Stop: the buyer will not invest data, access, time, or a decision date.

Step 5: build the output, not the architecture

Uplane's first deliverables included marketing strategy, website work, persona-specific ads, and human execution behind incomplete automation. That is a service wrapper around a product hypothesis. It is legitimate when disclosed.

Keep a manual-work ledger with four columns: step, person, minutes, and failure mode. Automate only the repeated step that constrains delivery or quality. Do not spend the week building permissions, settings, analytics, or integrations that the first buyer does not need to judge the result.

Pass: the buyer can use the output without the founder narrating every click. Revise: the founder can deliver the result, but the workflow is too fragile. Stop: the demo is a mockup and the promised result cannot actually be produced.

Step 6: put money between interest and roadmap

Körfgen's rule was not to run free pilots. The reason is diagnostic: “no budget” often appears only after weeks of free work. A dollar sign forces the buyer to compare the outcome with other priorities.

Price the first phase around a bounded result, not lifetime access. The fee should cover the work well enough that continuing teaches a sustainable lesson. A symbolic payment can still be weak evidence if the founder absorbs unlimited custom work.

Pass: the buyer signs and pays for a defined period, result, and support level. Stop: the buyer wants indefinite access, bespoke features, or free labor without a budget owner and decision date.

Step 7: use founder support as an instrument

Uplane later adopted a founder-reported 120-second response benchmark. Do not copy that number blindly. Copy the principle: early support exposes where the product is missing context and where manual work remains.

For the first five customers, log every question and classify it as unclear onboarding, missing input, product defect, custom request, or value failure. Review the log twice weekly. Pass: repeated questions fall as the product improves. Stop: speed masks the fact that customers cannot succeed without permanent founder intervention.

What failed, and what not to copy

• Bulk outreach: Uplane stopped using outreach agencies for volume and returned to curated founder messages. Personalization tokens are not research.
• Free pilots: free usage can preserve a false positive because the buyer has not displaced another budget.
• Fake demos: a prototype may be rough, but the claimed result must work and manual steps must be disclosed.
• Premature enterprise attribution: Deutsche Bahn came through a warm contact and remained in a tender/proof phase at recording time. It is not evidence that the first cold motion closed every later account.
• Copying current ICP backward: Uplane now targets large ad spenders, while its first paying customer was a smaller startup.

Your seven-day implementation plan

1. Day 1: choose one workflow, write the pain ledger, and define one usable output.
2. Day 2: build a 25-account list with role, trigger, source, and exclusion reason.
3. Day 3: research the top 10 and send honest discovery requests.
4. Day 4: run calls and score incident, cost, urgency, access, and budget owner.
5. Day 5: select at most one demo promise and write the seven-day contract.
6. Day 6: build the narrow output and record every manual step.
7. Day 7: rehearse the activation test, confirm the review, and prepare the paid scope.

The week's pass condition is not a customer. It is one qualified buyer who supplies inputs and accepts a dated working-demo review. If nobody does, revise the problem or segment before building.

Lead Scorer implementation

This is a narrow cold-outreach motion. Lead Scorer can organize research, qualification, and draft review, but it cannot validate the pain for you or send without approval.

1. Run icp-offer-context with the workflow, buyer role, trigger, exclusion rules, evidence limits, and the one seven-day output.
2. Create separate lists for qualified accounts, watch accounts, and rejected accounts. Keep a source URL and date on every admitted signal.
3. Score before enrichment. Require at least 8/10 fit, an explicit role match, and one current trigger. Stop credit-spending when fewer than 20 of 25 accounts qualify.
4. Run signal-research-dossier for the top accounts. Require two verified, dated signals; skip honestly when none exist.
5. Use cold-email-first-touch or linkedin-connection-requests to create one discovery ask per lead. Keep the ask to one conversation, not a product demo.
6. Run outreach-qa-audit. A human reviews every claim, removes false familiarity, and approves each draft before any send.

Build a list of 25 [role] at [company type]. Admit only accounts with [trigger] dated in the last 90 days. Score fit before contact enrichment. For each account scoring 8 or above, collect two source-backed signals and draft one 20-minute discovery request about [workflow]. Do not promise the product, infer intent, spend enrichment credits on rejected accounts, or send without my approval.

After calls, tag every objection by problem, timing, trust, implementation, and budget. Feed repeated objections into Content Studio as research questions. Useful evidence can attract new signal audiences, but engaged people still pass through the same scoring and human-approval gates.

Saveable checklist

• One buyer role, workflow, trigger, and seven-day output.
• Twenty-five accounts, with evidence and explicit exclusions.
• Discovery request that behaves like discovery.
• Dated review before any sprint begins.
• Real output, with manual steps disclosed.
• Paid scope before roadmap expansion.
• Activation event and support log.
• Separate attribution for cold, warm, and former-employer paths.
• Stop when inputs, urgency, budget, or safe delivery are missing.

Sources and limits

The first-customer sequence comes from Körfgen's August 2026 interview on The SaaS Podcast and the publisher's derived sell-before-building playbook. Both belong to the same source family. The first customer, payment, relative deal size, outreach response example, 10,000-ad estimate, 120-second rule, and $1M ARR timing are founder-reported.

Y Combinator, Business Insider, and Uplane's current site verify company identity, founders, product category, funding, current positioning, and later customer context. They do not independently prove the unnamed first deal, customer retention, or the revenue attribution. No public first-10 or first-100 count, first contract value, CAC, payback, gross margin, or audited revenue record was found.