A-101 · First contact
The AI front desk — how Autotelic answers the phone and the chat
Eve answers the phone and Chloe runs the chat; both hand every first contact to the next real step and neither can book anything.
- Project
- HOW IT WORKS
- Sheet
- A-101
- Sheet title
- FRONT DESK
- Rev
- 2026-09-26
- Set
- 17 SHEETS
- Status
- RELEASED
- 1ADU
- 2Home remodel
- 3New build
- 4Commercial
- 5Venue
- 6Front desk
- Released
- In progress
- Not in service
- Internal — shown, never linked
- Seat on the base (alignment)
- Item no. → parts list
How Eve, an AI voice line, and Chloe, a scoping chat, answer, summarize, and route every first contact at Autotelic — and exactly what each can and cannot do.
Released
Every construction project starts with a first contact — a phone call or a message from someone who doesn't yet know what they need, what it will cost, or who should handle it. A good front desk does three things well: it answers quickly, it captures the right information the first time, and it routes the person to the next real step without making them repeat themselves. In a construction business that first contact usually decides whether a lead becomes a scoped project or drops off entirely, since most callers are comparing more than one contractor and most website visitors leave the moment a form feels like a dead end.
An AI receptionist or a scoping chat can listen, ask clarifying questions, summarize what it heard, and hand a clean record to a person — the same job a well-trained office manager does. What it should not do is pretend to have authority it doesn't have: quoting a firm price before anyone has seen the site, promising a schedule, or committing to hold a date. The honest version of this tool says plainly what it can and cannot do, and it hands off to a person for anything that needs judgment.
At Autotelic, an AI voice line named Eve answers inbound calls first. Every call gets an AI-written summary afterward, and a filter step decides whether the call is ready for a warm transfer to a person or should be routed differently — so a team member isn't pulled off other work for every call, only the ones that are ready.
On the website, a scoping chat named Chloe handles the same first-contact job in text. Chloe's own instructions say it plainly: she has no tools, and she cannot book, hold, or reserve anything — her job is to ask about the project, note the trade and the city, and, on booking pages, hand over the specific booking details a person or the booking system needs next. When a scoping conversation finishes, it becomes a lead in the CRM, tagged so the team knows it came from the chat.
The front desk is the same door for every project type — whether the call or chat is about an ADU, a home remodel, a new build, a commercial tenant finish or a venue, Eve or Chloe take the first contact and route it to the right next step: a live tool, a scoped consult, or a person.
What it does
- Eve, an AI voice line, answers inbound calls and produces an AI-written summary of every call afterward.
- A mill filter step decides whether a call is ready for a warm transfer to a person.
- Chloe runs the scoping chat on the site; her own instructions say she has no tools and cannot book, hold, or reserve anything.
- Chloe's page context carries the trade, the city, whether design-assist applies, and the booking details a `/book` page needs next.
- A finished scoping chat becomes a CRM lead, tagged so the team knows it came from Chloe.
- Every summarized call and finished chat feeds the same lead pipeline the rest of the site uses.
What it sees
Three turns of a real, scrubbed scoping chat: Chloe asks, does not invent.
- ChloeHey! I'm Chloe 👋 Welcome to **Autotelic**! What can I help you with today?
- VisitorWhat do you think of my floor tile picks so far?
- ChloeI’d love to help you fine‑tune those tile choices! 😊 What style or color palette are you leaning toward, and where in the space will the tiles be installed?
- VisitorWhat do you think of my floor tile picks so far?
- ChloeSounds great! Could you tell me a bit about the tiles you’ve chosen—like the color, pattern, or material? That’ll help me give you the best feedback.
- VisitorWhich tile did I lock in?
Show the raw response
{
"page_context": "general",
"status": "abandoned",
"message_count": 5,
"history": [
{
"role": "assistant",
"content": "Hey! I'm Chloe 👋 Welcome to **Autotelic**! What can I help you with today?"
},
{
"role": "user",
"content": "What do you think of my floor tile picks so far?"
},
{
"role": "assistant",
"content": "I’d love to help you fine‑tune those tile choices! 😊 What style or color palette are you leaning toward, and where in the space will the tiles be installed?"
},
{
"role": "user",
"content": "What do you think of my floor tile picks so far?"
},
{
"role": "assistant",
"content": "Sounds great! Could you tell me a bit about the tiles you’ve chosen—like the color, pattern, or material? That’ll help me give you the best feedback."
},
{
"role": "user",
"content": "Which tile did I lock in?"
},
{
"role": "assistant",
"content": "Got it! I don’t have the exact tile info on hand—could you let me know the style or name of the tile you selected? That’ll help me confirm which one you locked in."
},
{
"role": "user",
"content": "Which tile did I lock in for my flooring?"
},
{
"role": "assistant",
"content": "I’m not seeing the specific tile you locked in on my end. Let me connect you with our team—they’ll have the exact details and can confirm the tile selection for you!"
},
{
"role": "user",
"content": "What project am I working on right now?"
},
{
"role": "assistant",
"content": "I don’t have the details of your current project handy—could you share a quick overview of what you’re working on? (e.g., kitchen remodel, smart‑home upgrade, etc.)"
}
]
}The shape of what Eve writes down after a call — keys and types only, no values.
- Calls sampled
- 50
- action_items
- array
- budget_signals
- null
- caller_intent
- mixed
- drafted_followups
- array
- follow_up_date
- null
- next_action
- mixed
- project_type
- mixed
Show the raw response
{
"rows_sampled": 50,
"keys": {
"action_items": "array",
"budget_signals": "null",
"caller_intent": "mixed",
"drafted_followups": "array",
"follow_up_date": "null",
"next_action": "mixed",
"project_type": "mixed"
}
}What it replaced
Nothing here replaced an old phase.
Proof
- 7,594 Calls with an AI-written summary
Where a first contact goes next
Eve and Chloe note the trade and the city, then hand the contact to the step that fits the project: a live tool, a fixed-price study, or a person.
The system layer
The AI and automation behind the front desk: how a call or a chat gets answered, summarized, and routed.
Topology
- Eve voice runtime
- Call intelligence summaries
- Mill filter
- Chloe sidecar
- Scoping session store
- Site-visit AI packet — Not in service
- Thread-intel drafts — Not in service
- Eve voice runtime → Call intelligence summaries
- Call intelligence summaries → Mill filter
- Chloe sidecar → Scoping session store
- Mill filter → Thread-intel drafts (planned draft generation (not running))
- Scoping session store → Site-visit AI packet (planned handoff (not running))
The one-day clock
- 17:00 — COO digest
- every 2 hours — Build-desk PR pulse
- every 5 minutes — Fleet watch
Questions
Can I book a project directly through Chloe?
No. Chloe's own instructions tell her she has no tools and cannot book, hold, or reserve anything — she scopes the conversation and hands it to a person or the right tool.
Does a human ever review what the AI wrote down?
Yes. Call summaries and finished scoping chats become leads a person reviews; nothing skips straight to a schedule or a crew.
What happens if Eve or Chloe can't answer my question?
Both are built to hand off rather than guess — a call routes to a person when it is ready, and a chat can always end in a phone call or a form to the team.
Is my call recorded?
Calls are transcribed and summarized by AI as part of how the front desk works; the summary is what a team member reviews afterward.