Someone Says, ‘We’re Building an AI Document Platform’—Is It a Real Project or an Early Idea?
Turn a vague Telegram discussion about an AI document platform into evidence about project stage, evaluation window, decision roles, and the right human next step.
You notice a message in a Telegram group:
“Our team is preparing to start a new project: a business-facing AI document-processing platform. We are still evaluating the technology stack, and the solution has not been decided.”
Note: This is an illustrative message assembled from common business-discussion patterns. It does not represent a real customer or conversation.
If you work in enterprise sales or business development, your first questions may be:
- Could this person become a customer?
- Should I message them immediately?
- What stage is the project in, and is there still time to participate?
Do not send a product link yet. The message contains no budget, owner, or date. It is enough to make a salesperson pause, but not enough to support the claim that a formal purchasing project exists.
Why these opportunities are often discovered too late
A new project almost never enters a community as a formal request for proposal.
It appears indirectly first:
- Someone complains that the current approach is too slow or unreliable.
- A team debates whether it should build something internally.
- A technical lead asks whether anyone has used a similar architecture.
- Several departments disagree about which direction to choose.
Each sentence looks like ordinary conversation on its own. Read together, however, they show a team approaching a decision: a vendor, technology stack, or architecture will be selected. The window is still open, but it will not stay open indefinitely.
That makes these Signals difficult. They are distributed, ambiguous, and rarely come with a clear closing date. By the time someone writes, “We selected vendor A,” the opportunity window has already closed.
“Preparing to start” can still describe three different situations
When someone says, “We are building an AI document platform,” the sentence says almost nothing about project confidence. It is an unverified description of a project, not a confirmed plan.
It may mean, “Our document workflow is too slow.” It may come from one engineer doing technical research. Or the team may already have approval and be filling out a vendor shortlist. All three situations can produce the phrase “preparing to start,” yet they deserve very different priorities from a salesperson.
The beginner path looks like this: see a project name → map it to your product → begin pitching. That response treats an unverified need as an approved purchase plan. The other person may conclude that you did not read the context—or that you were only waiting for a chance to sell.
The first message earns the item a place in the inbox. Later messages determine which section of that inbox it belongs in.
Four apparently clear statements—and what each one omits
Do not begin with, “Would you like to buy this product?” Ask questions that expose the project structure instead:
| Words used in the group | Facts still missing |
|---|---|
| “We are building an AI document platform.” | “What job must the platform complete first: document parsing, an approval workflow, or knowledge management?” |
| “We are still evaluating the technology stack.” | “Which approaches are under consideration? What has already been ruled out? When does the evaluation end?” |
| “The solution has not been decided.” | “Who is involved in the decision? Who approves the final choice? Is the primary criterion cost, performance, or delivery time?” |
| “The team is preparing to start.” | “Has the project been approved? Is there a preliminary budget or timeline?” |
You do not need to fire all four questions at the author. First read the surrounding discussion that is available in the connected group. Only then decide whether human contact is appropriate; an ordinary technical conversation should not be turned into an interrogation.
Put it in “verify,” not directly in “opportunity”
Not every “we want to build X” statement is a commercial opportunity. A useful filter separates core evidence from supporting evidence.
Core Signals: high weight, and each missing item should increase caution
- ✅ The team explicitly says it is starting a project—not “I have an idea,” but “we are preparing to start.”
- ✅ The project names a concrete business use case—“document-processing platform” is more specific than “AI platform.”
- ✅ Infrastructure or technology remains under active evaluation, so the selection window has not closed.
- ✅ The team structure is visible, with both a technical owner and a business stakeholder involved.
Supporting Signals: useful, but not decisive by themselves
- ✅ The discussion identifies a business pain and the limits of the current approach.
- ✅ It mentions a timeline or milestone, such as “we need a proposal next month.”
- ✅ A budget is mentioned, even if the team is still estimating it.
When at least three core Signals are supported, place the item in a human verification queue. With fewer than three, the highest-value action is still information gathering, not sales progression. This threshold is an editorial triage method—not a probability of purchase or a universal TOP Prospect product score.
On Monday morning, preserve the wording before anything else
Suppose the evidence reaches the threshold: there appears to be a real project, and the evaluation window remains open. The next move is not sending a product link. It is completing three steps.
1. Preserve the complete evidence record
Record:
- The original message: preserve the exact wording rather than summarizing it as “the customer wants an AI platform.”
- Permitted context: retain the surrounding discussion so the message is not interpreted in isolation.
- Source information: identify the group, the author, and their observable participation history.
- Your reasoning: note which core Signals are supported and which point remains uncertain.
Why does this matter? If the judgment proves wrong, the record helps you improve the rule. If it proves useful, the evidence supports a more relevant human opening instead of a generic pitch.
2. Verify before intervening
Before making formal contact, verify:
- Whether the project appears to exist beyond one statement, using later permitted discussion as possible corroboration.
- The likely project stage and budget range.
- Which roles participate in the decision path.
The right sequence is to provide value before discussing your product. If the author says document processing is too slow, share a relevant implementation pattern or case—not an unsupported claim that your product solves everything.
3. Decide whether follow-up is immediate
- Clear Signal + at least three supported core conditions → mark it “pending follow-up,” complete human verification within 24 hours, and then decide whether to make contact.
- Ambiguous Signal + one or two supported core conditions → mark it “watch,” set a review date one or two weeks later, and observe related follow-up discussion.
- Clear statement but insufficient confirmation, such as a casual idea → save it as an informational record without placing it in the follow-up queue.
These statuses manage verification priority. They do not prove that a project is approved or that the author has become a customer.
A four-line note is better than “AI project customer”
The next time you see a new-project discussion, do not start with, “Is this a customer?” Write four lines first:
Observation: Someone in an Indonesia engineering group says, “We are preparing to build an AI document-processing platform.”
Business impact: If the project is real, it may involve a document-parsing engine or approval-system purchase.
Next deadline: The stack is under evaluation, and the selection window may close within two to four weeks.
Still to verify: Project approval, budget range, and decision owner.If you cannot complete all four lines, the most valuable next action remains gathering information—not pushing the opportunity forward.
Conclusion
New-project Signals have unusually high variance. A correct judgment can let your team contribute early in a complete buying cycle. A weak judgment can waste hours on something a person mentioned casually.
The advantage is not contacting someone first. It is being able to say who raised the project, what remains undecided, when a choice may be made, and which part of your reading is still a guess. An ambiguous discussion is not a confirmed requirement, and a score is not a substitute for the original words.
Write the four lines first. Verify them, then decide whether the Signal should move to the next action.
This article extends the TOP Prospect scenario library. Continue with the complete methodology:
- How Telegram Group Messages Become Actionable Business Signals
- Keyword Rules or Semantic Judgment? A Stable Approach to Telegram Signal Extraction
- How to Score Business Signal Confidence: Source, Recency, Certainty, and Corroboration
TOP Prospect only analyzes messages from Telegram groups the user intentionally connects. It identifies candidate events in the user’s business language while retaining the original message, source, reasoning, and human review boundary. It does not read private chats, send messages automatically, or present a Signal score as proof of a deal.
How a Signal worth attention is found
See how Top Prospect finds and organizes Signals worth checking, keeps the original Telegram context, removes duplicates, and helps you decide what to review first. You decide whether to follow up and what to do next.
