10 min read
AI Chatbot for Lead Generation: Why We Built ConGreeto
Most website chatbots answer questions. ConGreeto qualifies the visitor, scores the lead, and drops it into your pipeline — grounded in your real business data. Here is why we built it, and what changed in 2026.
The enquiry nobody was awake to answer
It is 11:14pm. Someone lands on your site from a Google search, reads two pages, and has one question standing between them and a decision. There is a contact form. They look at it, decide it is not worth the wait, and go back to the results page — where three competitors are still open in other tabs.
Nothing broke. No alert fired. You will never know that visit happened.
That is the specific loss ConGreeto was built to stop, and today it is live.
What actually changed in 2026
Website chatbots are not new. What is new is that the generic ones stopped working, and everyone noticed at once.
The old pattern — a pop-up bubble with five canned buttons and a fallback to "leave your email and we'll get back to you" — trained an entire generation of buyers to close the widget on sight. The 2026 shift is away from generic pop-up chat and toward purpose-built assistants that understand what the visitor is actually doing on the page they are standing on.
The numbers behind that shift are hard to argue with:
- Shoppers who engage with AI chat convert at 12.3%, against 3.1% for those who do not — roughly a 4x lift.
- 64% of businesses running AI chatbots report an increase in qualified leads, not just raw volume (Silvertouch).
- Responding to a prospect within one minute is associated with a 391% improvement in conversion. Within five minutes, you are 100x more likely to make contact and 21x more likely to qualify the lead than at the thirty-minute mark.
- Roughly 28.5% of inbound enquiries arrive outside business hours — and in home services, healthcare and legal, those carry the highest intent and the highest revenue per contact.
Read those together and the conclusion is uncomfortable: for most businesses, the single biggest conversion lever is not the copy, the offer, or the ad spend. It is being awake.
But there is a catch, and it is the reason we did not just ship another bubble.
Why most "AI chatbots" still fail
Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027 — not because the models are bad, but because of escalating costs, unclear business value and inadequate controls. Gartner has a name for part of the problem: "agent washing", vendors rebranding an existing scripted chatbot as agentic AI. Of the thousands of vendors claiming agentic capability, Gartner estimates around 130 actually have it.
From the buyer's side, the failures are far more mundane than the analyst language suggests. A chatbot fails because:
- It does not know your business. It was trained on nothing, so it answers in generalities, invents a price, or says "I'm not sure, please contact us" — which is the contact form again, with extra steps.
- It answers instead of qualifying. It resolves the visitor's question perfectly and lets them leave. The visitor got value. You got a transcript.
- It captures a name and email, and stops. Now you have a row in a spreadsheet with no context, no urgency signal, and no idea whether this is a buyer or a student doing research.
- It cannot prove it did anything. No attribution, no revenue line — so at renewal it looks like a cost.
Every one of those is an architecture decision, not a prompt. That is where our build went.
What ConGreeto does differently
ConGreeto is an AI sales assistant and sales pipeline in one. Five things define it.
1. It is grounded in your actual business
Before your assistant answers anything, it learns your business: we crawl your website, ingest your PDFs and documents, and index your listings or product data. That knowledge is retrieved at answer time rather than baked into a prompt, which is what keeps answers correct when your pricing page changes.
Under the hood this is hybrid retrieval — dense vector search and keyword full-text search, merged with Reciprocal Rank Fusion. That is not engineering for its own sake: hybrid methods combining keyword and semantic search improve precision by around 30% and cut irrelevant results by 40% compared with vector search alone. It is also the most effective defence against the thing buyers fear most — nearly 70% of users report fewer hallucinations when responses are retrieval-grounded.
Concretely: your assistant quotes your delivery window, your warranty terms, your unit that is still available — or tells the visitor it will find out, and captures the lead while doing it.
2. It qualifies like someone who knows the industry
A conversation on a real-estate site should not go the same way as one on a dental practice site. ConGreeto ships with industry-specific conversation playbooks: the qualifying questions that matter, the order to ask them in, the fields worth capturing, and pipeline stages that map to how that industry actually sells.
It also reads the moment. Time of day, weekday versus weekend, and whether this visitor has been here before all change how the assistant opens and how hard it pushes. A returning visitor at 9am on a Tuesday is a different conversation from a first-timer at 11pm on a Saturday, and treating them identically is how you burn the good ones.
The assistant works across 30+ languages including right-to-left rendering, and detects the visitor's language rather than making them hunt for a switcher.
3. It scores and tiers every lead
When a conversation ends, ConGreeto analyses it and returns a structured lead: a summary of what the person actually wants, the fields it captured, and a tier — Hot, Warm, Nurture, or Cold — judged against that industry's own conversion norms rather than a generic template.
This matters more than it sounds. High-intent messages in a bot conversation make a visitor five times more likely to become a real opportunity. The value is not in capturing more leads. It is in your team knowing, at a glance on Monday morning, which four of the weekend's nineteen conversations to call first.
Contact details are captured during the conversation, as they surface, rather than gated behind a form at the end — so a visitor who leaves halfway through still leaves you something to work with.
4. It hands off into a pipeline, not an inbox
Leads land on a drag-and-drop board with owners, statuses, follow-up scheduling and notes. Roles (admin, agent, viewer), a team leaderboard and a full audit trail are built in, so a growing team can genuinely run on it. Idle conversations are swept up automatically, and end-of-chat summaries reach the right people by email.
The point is that a lead never has to be re-typed into another system to become work.
5. It shows you which campaigns produced revenue
ConGreeto matches the page a visitor arrived on against your active campaigns, links every matched campaign to that conversation, and tracks cost against closed revenue. It also adapts the assistant's goal and tone per campaign — the conversation on your Google Ads landing page can push a different call-to-action than the one on your homepage.
The reporting question stops being "how many chats did we get" and becomes "which campaign paid for itself".
Under the hood, briefly
We are a product company that also builds AI systems for other people, so we are opinionated about this. ConGreeto runs on FastAPI and PostgreSQL with pgvector, with background workers split across four queues by latency budget so a slow document ingest can never delay a live reply. Around the AI calls sit a Redis-backed circuit breaker, prompt-injection sandboxing of visitor input, output moderation, per-tenant token and conversation quotas, and correlation IDs that trace a single request from the browser through the worker to the database.
None of that is visible to your visitors. It is the reason the assistant is still answering at 3am on the day one of your suppliers has an outage.
If you want the long version of that reasoning, our parent company wrote it up: Why 40% of AI Agent Projects Get Cancelled — And What We Did Differently.
Who this is for
ConGreeto is multi-industry by design — real estate, automotive, healthcare, professional services and e-commerce are where it fits most naturally — and it scales from a single site to a multi-brand network.
It is a particularly strong fit if you recognise any of these:
- Your enquiry volume is fine but your response time is not.
- A meaningful share of your traffic arrives outside working hours.
- Your team spends the first hour of every day triaging rather than selling.
- You are spending on ads and cannot say which campaign produced revenue.
- You have tried a chatbot before and quietly turned it off.
Getting started
Installation is one line of code — a single script tag, identical on every page, no developer required. If you are on Shopify, WordPress, Webflow, Squarespace, Wix or one of twenty-odd other platforms, our embed assistant walks you through exactly where that tag goes on your platform and verifies the widget actually loaded.
Training the assistant is a URL and, if you have them, a few PDFs. Most setups are answering questions the same afternoon.
Plans start at $37.50/month (billed annually) for Starter, with Growth and Scale tiers for larger teams and Enterprise for multi-brand networks. Every plan includes the full feature set — the tiers scale seats, chatbots, conversations and AI tokens, not capability.
Frequently asked questions
How is ConGreeto different from a normal website chatbot? A normal chatbot answers questions. ConGreeto is built to qualify: it runs an industry-specific conversation, captures contact details as they surface, scores the lead into Hot/Warm/Nurture/Cold with a summary, and pushes it onto a sales pipeline with an owner and a follow-up. It also attributes the conversation to the campaign that produced it.
Will it make things up about my business? It answers from your indexed content — your site, documents and listings — using hybrid semantic plus keyword retrieval, and it is instructed to capture the lead rather than guess when it does not have an answer. Retrieval grounding is the mechanism that keeps answers tied to your real data, and industry research consistently shows a large drop in hallucination when responses are grounded this way.
How long does setup take? One script tag on your site, then point it at your website URL and any documents. Most customers are live the same day, and no developer is required for the install.
What languages does it support? Over 30, with right-to-left rendering. The assistant detects the visitor's language rather than requiring them to choose one.
Does it work outside business hours? That is the main reason it exists. Around 28.5% of enquiries arrive after hours, and those leads are frequently the highest-intent ones. ConGreeto answers immediately, at any hour, and the qualified lead is waiting when your team logs in.
Can I control what it says? Yes. You set the assistant's personality, response length, initiative level and emoji policy, and campaign overlays let you change its goal and call-to-action per landing page. Restricting it to specific domains is enforced server-side.
How do I know it is working? Campaign attribution ties conversations to campaign spend and closed revenue, so you can see cost against outcome per campaign rather than a chat-volume vanity metric.
ConGreeto is built by Brillnex Systems, a software and AI engineering firm building production AI systems for startups and growing businesses.