Key facts

Business: Pre-owned superbike dealer in Punjab, India.
Main problem: The owner has little time to search for good bikes online.
Sources in the design: OLX, Google search and Google Maps dealers, and Facebook Marketplace (forwarded by the owner, not scraped).
Output: One WhatsApp shortlist per day, plus a tracking sheet.
Human control: The owner approves every message before it goes to a seller.
Planned build time: About 6 weeks, with a first working version in week 2.
Expected owner review time: About 15–20 minutes a day (design target).

Problem

What problem did the dealer have?

The dealer does not have time to search the internet for good used superbikes. In his words (paraphrased): "I don't get much time to find good bikes on the internet. This is the main and major problem."

The business told us:

Good used superbikes are listed on OLX, Facebook Marketplace and by other dealers found through Google.
He runs the business himself, so time spent scrolling listings is time taken from buyers, the showroom and paperwork.
He wanted an AI agent that brings him sourcing leads, so he does not need to search.

Good bikes at a fair price sell fast. If the owner checks listings only late at night, the best deals are often gone.

Solution

What is an AI lead generation agent for a bike dealer?

An AI lead generation agent for a bike dealer is software that finds bikes for sale, checks them against the dealer's buy rules, and sends only the best ones to the dealer. It does the searching and sorting. The dealer still does the calling, checking and buying.

How would the daily work change?

The owner would stop searching and start reviewing. The table below compares a normal day before and after.

Step	Today (manual)	With the proposed agent
Find listings	Owner scrolls OLX, Marketplace and dealer pages when he has free time	Agent collects listings through the day
Check details	Owner opens each ad, reads model, year, km, price	Agent pulls these fields into one clean format
Remove repeats	Same bike seen on 2–3 sites, checked again each time	Agent marks duplicates and keeps one record
Judge the price	From memory	Agent compares with recent similar listings
Decide who to call	Owner's gut feeling, late in the day	Ranked shortlist on WhatsApp at 9 AM
First message to seller	Typed one by one	Agent drafts it; owner edits and sends
Keep track	Phone notes or memory	One shared sheet with status for every bike

How it works

How would the agent work, step by step?

Our proposed design has seven steps.

Collect. The agent gathers listings from three kinds of sources:
- OLX: saved-search alerts and public listing pages, read at a low rate and within OLX's terms (to be confirmed before build).
- Google: the Google Places API finds used bike dealers in chosen cities. The agent then checks their public stock pages.
- Facebook Marketplace: the owner keeps saved searches on his own account. When an alert looks good, he forwards the link or a screenshot to the agent's WhatsApp number. The agent never logs in to Facebook and never scrapes it.
Normalise. An AI model reads each listing and fills fixed fields: make, model, year, kilometres, asking price, city, seller type (owner or dealer), photos count and listing link.
Remove duplicates. The same bike often appears on more than one site. The agent matches on model, year, km range, city and photos, and keeps one record.
Score. Each bike gets a score from 0 to 100 (see the table below).
Send the shortlist. Every morning, the top bikes go to the owner on WhatsApp.
Draft the first message. For each shortlisted bike, the agent writes a short, polite message to the seller. The owner approves, edits or skips it.
Track. Every bike and every contact goes into one Google Sheet with a status: New, Contacted, Inspecting, Bought or Rejected.

How would the agent score each bike?

The score is a weighted sum. The owner sets the buy list and can change any weight.

Factor	What the agent checks	Example weight
Model fit	Is the model on the owner's buy list?	25
Price vs market	Asking price vs the median of similar recent listings (same model, year ±1, similar km)	25
Year and kilometres	Inside the owner's limits (for example, max 5 years old, max 25,000 km)	20
Distance	Distance from the showroom; nearby cities score higher	10
Seller type	Single owner often scores higher than a reseller	10
Listing quality	Clear photos, service history mentioned, RC details given	10

Bikes below the owner's cut-off (for example, 60) are saved in the sheet but not sent. Any red flag, such as "accidental" or "no papers", moves the bike to a separate "check carefully" list.

What would the WhatsApp shortlist look like?

It is one short message with a ranked list and links. Below is a sample with made-up data, not real listings.

Good morning. Today's superbike shortlist (5 of 38 new listings):

1. Kawasaki Ninja 650, 2022, 11,200 km
   Asking ₹5.45 lakh | Market about ₹5.9 lakh | Ludhiana, single owner
   Score 86 | Link: [listing]

2. Triumph Street Triple R, 2021, 14,800 km
   Asking ₹7.6 lakh | Market about ₹8.1 lakh | Chandigarh, owner
   Score 81 | Link: [listing]

3. Harley-Davidson Iron 883, 2020, 9,500 km
   Asking ₹7.9 lakh | Market about ₹8.0 lakh | Jalandhar, dealer
   Score 72 | Link: [listing]

Reply 1, 2 or 3 to see a draft message to the seller.
Reply "sheet" for the full list.

When the owner replies "1", the agent sends a draft such as: "Hello, I saw your Ninja 650 listing. Is it still available? Can you share the service record and RC details? I can come to see it this week." The owner sends it from his own number.

What tools would the agent use?

The design uses common, low-cost tools that the dealer owns.

AI model: Claude, GPT or Gemini to read listings and screenshots and to draft messages. We pick based on cost and Hindi/Punjabi text support.
Messaging: WhatsApp Business Platform (Cloud API) for the shortlist and the forward-to-agent number. See our guide to WhatsApp alerts.
Dealer discovery: Google Places API.
Data: a small database for listings and price history, plus Google Sheets as a simple CRM.
Scheduler: a daily job that runs collection, scoring and the 9 AM shortlist.

The dealer owns the code, prompts and data.

What is the rollout plan?

The plan is three phases over about six weeks. Each phase gives the owner something he can use.

Phase	Weeks	What the owner gets
1. Foundation	1–2	Buy list agreed. Google dealer list. WhatsApp forward-to-agent number. Tracking sheet.
2. Shortlist	3–4	OLX alerts added. Clean data, duplicate check, scoring. Daily 9 AM WhatsApp shortlist.
3. Outreach and tuning	5–6	Seller message drafts with approval. Scoring weights tuned from the owner's yes/no feedback.

What can this agent not do?

The agent does not inspect bikes, check documents or agree prices. Those stay with the dealer.

No Facebook scraping. Marketplace leads depend on the owner forwarding them. If he forwards none, the agent sees none.
No full market coverage. Many bikes sell through friends, WhatsApp groups and local mechanics. The agent will miss those.
Price estimates are rough at first. The "market price" comes from listings the agent has seen. It improves after a few weeks of data.
No automatic messages to sellers. Every message needs the owner's approval.
Platform rules can change. If OLX or Google change their terms or access, a source may need to be changed or dropped.
Data care. Seller phone numbers are personal data under India's DPDP Act. The sheet is private, and old seller data is deleted on a set schedule.

Expected outcomes

What is the expected impact?