Geode: catching live demand

When ads stop delivering enough enquiries at an acceptable cost, businesses start looking for other ways to find clients.

Discuss a task

The Geode dashboard: found enquiries with the original quote, the source and a «Take it» action

The problem

Part of that demand already shows up in Telegram chats and Instagram comments: people describe their problem, ask for advice and look for someone to do the work. Those messages disappear down the feed quickly, and the sales team usually never sees them.

At Traffic Rock we built Geode to find these requests and bring them to a manager. The system works across two platforms, checks messages with two LLMs and collects the enquiries in its own CRM.

The brief

Telegram parsers have been around for years. Most of them match on keywords, so the manager ends up with messages where the topic is merely mentioned: a contractor’s ad, a job posting, a question to colleagues or an old thread. Everything has to be read manually again to decide who is actually worth contacting.

We wanted a system that takes the meaning of a message into account along with the rules of a specific business. It has to tell a potential client’s request apart from topical noise, work with Instagram comments as well as Telegram, and then store the enquiry together with its context and handling history.

What we built

Geode monitors selected Telegram communities and Instagram comments. For each company we define the instructions: what services it provides, which regions it covers, who its client is, which wordings signal demand and which messages should never reach sales.

The criteria can be adjusted to the niche and the business model. One company cares about flat renovation requests, another about complex integration enquiries, a third about finding a carrier for a particular route. Geode evaluates messages against the conditions set for that specific task.

A candidate goes through two sequential checks. The first LLM breaks the message down: who wrote it, what the task is and whether there is any intent to act. The second LLM verifies that conclusion against the client’s rules. The card keeps the original quote, the author, the source and the link, so the manager can judge the context for themselves.

This is where Geode differs from a plain keyword parser: it combines Telegram and Instagram, allows detailed instructions, and uses a second AI stage to verify the demand it found.

The enquiry lands in the CRM

Finding the message is not the end of the job. Geode includes a CRM where the team sees the feed of enquiries and takes each one further. There are lead cards, statuses, owners, notes, an action history, a list of sources, prompt settings and dashboards.

The manager opens the original request, decides how to reach the author and records the outcome. Management sees which sources produce suitable enquiries and what happens to them once they are handed over. If messages come in but never turn into conversations, that becomes visible too.

The outcome

We brought together monitoring of Telegram and Instagram comments, flexible search rules, verification by two LLMs and a CRM for the sales team in a single product. Instead of manually scanning dozens of sources, the team receives messages with clear context and can follow the path of every enquiry.

Geode helps you work with demand that is already expressed in open discussions. It does not create demand and does not guarantee sales: the number of suitable messages depends on the niche, how active the sources are and how quickly the team replies to people.

Try Geode on one of your services

Name a single service your business offers. We will show which sources can be monitored, how the two LLMs read messages against your criteria and what the manager sees in the CRM.

Request a Geode demo and see which enquiries your business might be missing right now.