TL;DR: Developer Manthan didn’t want to call 30 car dealerships — so he delegated the task to Claude AI. The AI independently found the contact details of dealership managers, drafted 20 personalized emails, and secured a $2,000 discount on a car that in California is usually sold above MSRP. The case shows that AI’s main value is not writing more elegantly, but being willing to do boring work at industrial scale.
1. The problem: the car of your dreams and a market with no discounts
The Volkswagen Golf R is one of the best hot hatches on the market. Practical, fast, discreet from the outside. The ideal car for anyone who doesn’t want to shout about their 300 horsepower.
But in California, getting a Golf R at the standard price — MSRP (manufacturer’s suggested retail price) — is almost impossible. The model is produced in limited batches, sells out quickly, and dealers have no reason whatsoever to offer discounts. The standard response when you call is: “We have a $1,500–3,000 markup over MSRP.”
Stories occasionally appeared on Reddit about someone managing to get the car with a small discount at the end of the year — when dealers are closing quarterly targets. But you had to call at the right moment, to the right dealership. Almost impossible to guess.
Developer Manthan found himself in exactly this situation. He wanted to buy a Golf R, and he had the money. What he didn’t have was the desire to methodically call 20–30 dealers across the state in hopes of finding one willing to give way. So he decided to try something unconventional.
2. The idea: what if AI handled the negotiations?
Modern AI assistants — especially Claude from Anthropic — can do more than just answer questions; they can also work in a browser on their own: search for information, open websites, and draft emails.
Manthan ran a simple experiment: what if I just tell the AI that I want to buy a Golf R, and let it sort out the rest itself?
The task was phrased roughly like this:
“I want to buy a 2026 Volkswagen Golf R. Ready to buy immediately. I need at least $3,000 off MSRP. I’m willing to finance through the dealer. I need confirmation of the deal by email or SMS before I come to the dealership.”
He created a separate email address specifically for this task — so dealership spam wouldn’t mix with his work email. Then he launched Claude and got back to work.
3. How Claude searched for dealers and wrote emails
While Manthan was busy with his own things, Claude worked in the background. Here’s what it did on its own, without additional instructions:
- Dealer search. The AI started by searching for all official Volkswagen dealers in California that had a Golf R in stock or on the way.
- Finding the right contacts. Claude didn’t send emails to generic addresses like info@dealer.com. It independently decided to find specific people — sales directors and general managers. The people with the authority to approve unusual discounts.
- Personalized emails. In the end, Claude drafted about 20 email drafts — one for each dealership.
Here is an example of the email that ultimately got results:
“Hi [name], I’m looking for a 2026 Golf R — base trim, no Euro package. Color doesn’t matter. I’m willing to finance through the dealer. I’m reaching out to several dealers in the region, and my target is $3,000 below MSRP. The first dealer to give me that number gets the deal. Please communicate only by email — no calls. As soon as we agree on the price and I receive a written order, I’ll come in to close the deal. Thanks, Manthan”
The email contains everything needed: specifics, urgency, respect for time, and directness. When the AI finished the work, Manthan simply opened Gmail and hit “Send” on all the drafts.
4. The result: $2,000 below MSRP on a scarce model
Most dealers declined. Some didn’t reply. One manager wrote that in all his years of work he had never once sold a Golf R below MSRP. But one dealer said “yes.”
Manthan handled the subsequent negotiations with Claude as well — the AI drafted reply emails directly in the thread. In the end, they managed to agree on a price more than $2,000 below MSRP.
To verify it, Manthan called several local dealers with the agreed price already in hand — could he find something better closer to home? The reactions were telling: one dealer said they could “probably” do a $2,000 markup, another joked: “Which dealer is that? Because if you don’t buy it right now — I will.”
After the purchase, the manager explained why he agreed: Volkswagen Corporate was pressuring dealers on sales volume without providing advertising budget. Several Golf Rs were sitting unsold, and the discount helped move inventory. As Manthan himself says: it was a combination of luck and good timing. But you can’t catch luck unless you make enough attempts.
5. Why it worked: 3 key principles
5.1. The volume of attempts changes everything
The main value of AI in this task is not the quality of the writing. Manthan could have written a similar email himself. But he wouldn’t have written 20 emails — it’s boring, time-consuming, and feels pointless. Claude removed that barrier entirely. When you make 20 attempts instead of zero, the chance of hitting the right moment rises dramatically.
5.2. AI found the right people
Claude independently understood that a request for an unusual discount should go to the person who can approve it. General managers, not inbox operators. It’s simple logic — but that logic turned 20 random emails into 20 targeted requests to decision-makers.
5.3. The email created the right dynamic
The phrase “the first dealer to give me that number gets the deal” creates gentle pressure without aggression and signals that the buyer is serious, the decision has already been made, and the only thing left is to choose who will handle it.
6. How to repeat this yourself: step-by-step instructions
You can apply the same approach — not only to buying a car, but to any negotiation where you need to reach out to many people at once.
- Step 1. Create a separate email address specifically for the task.
- Step 2. Give the AI a clear assignment: what you want to buy, the specific condition (price/discount), constraints (email only, written confirmation required), and readiness to act quickly.
- Step 3. Give the AI browser access (Claude with coworking mode, OpenClaw).
- Step 4. Review the drafts and send them — no need to edit every email.
- Step 5. When a positive reply comes in, take over. Use the AI to draft responses, but make the final decisions yourself.
7. FAQ
Can this approach be used in Russia?
Yes. The scheme works for any negotiation where you need to make many similar outreach attempts: buying a car, renting office space, finding contractors, requesting quotes. Adapt the email to the context.
Do you need dealer permission for this correspondence?
No. You’re simply sending an email with a request — this is standard business communication. Nothing illegal or inappropriate.
Which AI is best suited for this task?
Claude from Anthropic performed well in this case, especially with browser mode enabled. You can also use OpenClaw — a task automation platform based on Claude.
What if all the dealers refuse?
You spent an hour setting it up — and learned the real market situation. That is already valuable. Plus, you now have an email database of specific managers for the future.
Will this work for buying other products?
Yes — for anything where there is room for negotiation: equipment from major retailers, rentals, wholesale purchases, services. The more suppliers there are on the market, the higher the chances.
Conclusion
Mantkhan's story is a good illustration of how AI is changing everyday tasks. Not because it writes emails better than a person. But because it is ready to do tedious work at a scale that a person does not have the patience for.
Buying a car with the help of AI is neither science fiction nor a privilege of programmers. It is simply a correctly defined task plus a tool that can work autonomously. Try applying the same principle to something in your life — and you will be surprised how many "boring" tasks can be delegated.