TNT Signal
Internal viewThe assessment, as a buyer sees it
Internal view. It should feel like the first hour with a good consultant. The buyer gives a website; we read it and research the company in the background while they tell us where they are with AI and what they're trying to solve. Then we play back our understanding with up to five places to explore, and go deeper on the one they pick. Internal working model, not the public copy.
Every pathway
The whole decision tree. Click a question to fold it, an outcome to see its detail, or pick a vendor to light up every path that leads to it.
By business type
The website lookup picks the type; the buyer confirms it. Each type opens its own plays.
"We tried AI and it stalled"
"We already run AI"
No vendor fits: build, expert or scope
Every "something else" and "none of these fit" answer lands here.
Who we can push buyers toward
The 17 TNT companies in the catalog, grouped by buyer. Each card lists the quiz outcomes that route to it; click one to replay that path. None of these companies has agreed to partner terms yet.
Routing logic
The rules behind the tree: what each route means, when we build versus bring in an expert, and how readiness decides the next action.
The consulting flow
1. Website
Starts two background jobs: an AI read of the site (business type, one-line summary, ranked plays; ~10s) and public research with web search (ownership, size, news, leaders, AI mentions; ~60s).
2. Six quick questions
AI journey, what they're solving, which teams carry repetitive work, where the information lives, who would own it and when, then role, size and ownership (prefilled from research). Each one feeds a framework dimension, and together they cover the research time.
3. Drafted while they answer
The playback starts as soon as the owner/timing screen is done and never waits on research; if research lands later, the playback updates in place.
4. Our understanding
A playback of the business, where they are, what matters, and up to five opportunities picked from business types' pre-baked plays (or one idea specific to them). They can correct it and we regenerate.
5. Go deeper
Picking an opportunity opens the scoping questions and the route: vendor, build, expert, scoped discovery or not yet. Then a review call.
Engine
Prototype: local server.py calling Claude through the Claude Code CLI (research uses only web search and fetch; everything else runs with no tools). Production: the same calls via the API, with Exa or Apollo for faster research and firmographics.
Sent with every assessment, so the AI rates against our definitions, not its own.
Routes
No vendor fits: the checks, in order
| # | Check | If no | If yes |
|---|---|---|---|
| 1 | Is there enough work to matter? 40+ hours a month, or a revenue, risk or cash line attached. | Defer or self-serve | Next check |
| 2 | Is the problem mainly technical, not process, ownership or training? | Fix foundations first | Next check |
| 3 | Can we get access to the data and systems? | Validate feasibility | Next check |
| 4 | Is it a standard build pattern (documents, email, CRM, internal tool, voice intake, agent over a database)? | Deep domain or regulated science: expert network. Spans many teams: scoped discovery. | Next check |
| 5 | Will we see it again (3+ clients or the whole vertical)? | Custom build pilot | Build it as a TNT Labs product candidate |
What the quiz collects for every workflow
The work
- Desired outcome
- What starts the work, who does it, what the output is
- Volume per month and minutes per case (ranges are fine)
The environment
- Systems and inputs: email, documents, CRM, ERP, spreadsheets
- Where judgment and review happen
- Tools already licensed and pilots tried
The buyer
- Who owns the workflow
- Real deadline or not
- Readiness: exploring, building a case, ready to scope
- Optional budget range, including "not sure"
The estimate (only when inputs support it)
Monthly hours = cases × minutes ÷ 60. Released capacity = hours × an explicit assumed reduction, minus added review time. Capacity is not EBITDA until we say how it's realized.
Readiness decides the next action
| Buyer state | Next action | Stop rule |
|---|