AI phone coverage, governed by your operating rules
This is an AI voice assistant, not a human. Starting a call records and transcribes your audio and shares it with ElevenLabs and its model and service providers so the assistant can respond, as described in our Privacy Policy at /privacy. Your microphone is used only while the call is open and is released when it ends.
Pick a starting line
Every call gets an answer. Every opportunity gets an owner.
EchoCody answers the questions you approve, qualifies the caller, books or routes them, writes the outcome back to your CRM, and leaves the record with exactly one owner and one next action — with a due time attached.
The demo on this page is a synthetic voice. It never starts on its own — it opens only after you tap, and only after you acknowledge the recording notice. Prefer a human? (801) 980-0308
02Industry intelligence
It opens the call already knowing your rules.
Each vertical has its own opening, its own discovery set, one line the agent will not cross, and a finite set of end results a conversation is allowed to reach.
Home · how the call opens
Opens by separating no-heat / no-cool emergencies from maintenance and quotes before anything else.
What it must learn
- Emergency or scheduled
- System type and age
- Service address in area
- Under warranty or contract
- Access window
The line it will not cross
No diagnosis over the phone and no firm repair price. Gas smell or CO concern gets emergency direction immediately.
Definite end results
- Job booked
- Emergency dispatch direction
- Callback scheduled
- Estimate scheduled
- Out of service area — closed
Every conversation must land on exactly one of these, with one owner and one next action written back to the CRM. Nothing is left open without a date.
03The cost of missed demand
Model it with your own numbers.
Nothing below is a customer result. Change the inputs and the arithmetic moves — the formula is printed underneath so you can argue with it.
- Missed opportunities
- 10/wk · 43/mo
- Modeled recoverable range
- $772 – $1,479 / mo
- Break-even captures
- 27.8 closed jobs / mo
- Annualized (midpoint)
- $15,432
- Net of coverage cost
- $-213
Assumptions and formula
missed/mo = calls/wk × missed% × 4.33
recoverable/mo = missed/mo × reachable% × qualified% × book% × show% × close% × value
range = midpoint × 0.6 … midpoint × 1.15
break-even = coverage cost ÷ (value × qualified% × book% × show% × close%)
The ±range exists because reachability and close rate are the two inputs a model cannot verify. This is a scenario, not a forecast, and not a guarantee of revenue.
A fair comparison
Three ways to cover the phone. Each is genuinely better than the others under some conditions.
| Internal team | Answering service | EchoCody | |
|---|---|---|---|
| After-hours coverage | Only if you staff it | Yes, message only | Yes, within your approved rules |
| Business context | High once trained | None | Loaded from your CRM before each action |
| Books on your calendar | Yes | Rarely | Yes, when booking rules allow |
| CRM writeback | Manual, inconsistent | Rarely | Every call, structured disposition |
| Human oversight | Is the human | Limited | Review, approve, version, deploy |
| Judgment on unusual calls | Strongest | Weak | Escalates to a human by design |
| Best fit when | Volume is steady and staffed | You only need a message taken | Demand is spiky and every call needs an owner |
04Same call, two systems
Illustrative call reconstruction.
One inbound call, run twice. No dollar figures — only the observable steps and what exists in the record afterwards.
- 00:00Call arrives after hours. Rings out.Call arrives after hours. Answered, with an AI disclosure.
- 00:12Voicemail greeting plays.Caller states the problem. Intake fields captured.
- 00:40Caller hangs up without leaving a message.Qualification runs against approved rules for the trade.
- 01:05No record exists that the call happened.Booking rules allow a slot. Window offered and confirmed.
- 01:30—Disposition written to the CRM: stage, owner, next action, due time.
- 01:45—Confirmation sent to the caller. Owner notified.
- Next AMNobody knows to call back. No owner, no next action.Appointment on the calendar with evidence attached to the record.
Result — conventionalVoicemail, no record, no owner, no next action.
Result — governedDisposition written, next action owned with a due time, confirmation sent.
Read the transcript
- Caller
- Hi — my system quit and it's late. Is anyone there?
- Cody (AI)
- You've reached the after-hours line and I'm an AI assistant. I can get this on the board. Is it out completely, or running weak?
- Caller
- Completely out.
- Cody (AI)
- Understood. Is the address the one on file, and is tomorrow morning workable if I can hold a window?
- Caller
- Morning works.
- Cody (AI)
- Holding it. You'll get a confirmation, and a named owner is attached to this before anyone starts the day.
05Stage controller
The operating spine, not a chatbot.
Every stage declares its trigger, guards, owner, SLA, measurable end result, authorized exits and failure route. Select a stage to read its contract.
AI Call — Active
- Trigger
- Call connected.
- Action owner
- Agent
- SLA
- Disposition written within 60 seconds of hangup
- Measurable end result
- One disposition, one owner, one next action, one due time.
Guards
- AI disclosure delivered
- Recording notice given
- Approved script and boundaries loaded
Authorized exits
- Engaged
- No Answer
- Callback Scheduled
- Booked
- DNC
Failure route
Tool failure mid-call → graceful close, owner task, retry queued.
Priority overrides — highest wins
0 · Compliance / terminal
STOP, DND, wrong person, Closed Won/Lost, Do Not Contact. Suppresses everything below it.
1 · Customer events
Reply, booking, reschedule, cancellation, no-show. Interrupts the cadence before responding.
2 · Human control
Manual-control tag or an active owner conversation pauses automation.
3 · Disposition router
Signed, native call outcomes route the record.
4 · Stage-entry worker
Runs the entry action for the stage the record just landed in.
5 · Timed cadence and audit
Lowest priority. Only runs when nothing above it applies.
Any substantive inbound reply interrupts the cadence before anything else runs. Target acknowledgment is under 15 seconds, with a hard exception raised at 60 seconds. Automated outbound runs only between 08:00 and 20:59 in the prospect's local time; at 21:00 it stops and holds until 08:00.
06Operating intelligence
Turn every conversation into governed operating intelligence.
A supervised loop, not a self-rewriting agent. Changes are reviewed, approved, versioned, deployed and measured — or rolled back.
01
Capture
Every conversation is transcribed and reduced to structured facts and a disposition.
02
Classify
Facts are tagged: objection type, exit point, missing field, boundary hit, tool failure.
03
Review
A human reviews proposed changes against the classification and the evidence behind it.
04
Approve
Nothing reaches a live call without an explicit approval. The agent never rewrites itself.
05
Version
The approved change is versioned so any behaviour can be traced to a decision and a date.
06
Deploy
The version is promoted after certification scenarios pass.
07
Measure
The change is measured against the metric it was supposed to move, or it is rolled back.
Your CRM stays the source of truth. The agent reloads approved context — stage, consent, owner, notes, appointments, DND — before every action, and never treats its own memory as authoritative.
Why prospects exit — diagnostic categories
Illustrative example — not a customer result- Evidence linked
- Transcripts where a price question appears in the first 30 seconds and the call ends without a booking.
- Owner
- Sales lead
- Recommended correction
- Approve a bracketed range response plus one qualifying question before any number is given.
- Measurable end result
- Share of price-first calls reaching a scheduled next step.
07Pre-call journey
The Intelligence Brief lands before the call does.
Booking stops acquisition and starts preparation. Nothing is sent until a human has checked every claim against its source.
- 01
Booking confirmed
Acquisition cadence stops. The record moves to Report Pending with the assigned advisor.
- 02
Intake gaps only
The prospect is asked only for what is genuinely missing — never a re-run of what they already answered.
- 03
Intelligence Brief generated
Built from booking facts, CRM notes, transcript facts and approved sources. No invented figures.
- 04
QA
A human checks every claim against its source. A failed QA holds delivery — a placeholder is never sent.
- 05
Delivered before the call
More than 24h out: delivered at T-24h. Inside 24h: delivered immediately after QA passes.
- 06
Priority chosen
The prospect is asked which section matters most, and a reminder goes out at T-2h.
- 07
Advisor briefing
Closer receives a one-screen briefing: context, priority, open questions, and the decision on the table.
- 08
Strategy call
The conversation starts from shared evidence instead of discovery.
- 09
Structured recap
Notes within 15 minutes, personalized recap within 30, and one next step with an owner and a date.
Intelligence Brief · structure preview
Pre-call operating brief
Version and QA stamp recorded on delivery
- 01
Situation as stated
The problem in the prospect's own words, sourced to the call or form.
- 02
Systems in play
What they already run, and where the handoffs break.
- 03
Where demand leaks
Modeled from their inputs, with the assumptions printed.
- 04
Options on the table
Two or three routes, each with what it requires and what it does not solve.
- 05
Decision to make
The one decision the strategy call exists to reach.
- 06
Open questions
What the report could not verify, stated plainly.
Sections shown are the report's structure. No customer data appears on this page.
08Weekly Operations Brief
Not a dashboard. A decision.
What changed, where prospects exited, what it costs on your modeled inputs, and the one fix to make first — with an owner and a date.
- 01
What changed
Inbound volume shifted toward after-hours, and a larger share of conversations reached a dated commitment than the prior week.
- 02
Where prospects exited
The largest single exit category was price asked before value was established, followed by scheduling friction on same-week windows.
- 03
What it cost — modeled
Cost is derived from the operator's own inputs in the scenario model. No revenue figure here is observed; it is arithmetic on the inputs provided.
- 04
Fix-first priority
Approve a bracketed price response with one qualifying question before any number is given.
- Owner
- Named human, not a queue
- Due date
- Set on the brief, tracked to close
- Confidence / evidence
- Stated with the transcript set it came from
- Expected end result
- The one metric the fix should move
This is decision intelligence, not a chart pack. Every line has an owner, a date, the evidence behind it, and the measurable result it is supposed to produce. Where verified data does not exist yet, the brief says so.
09Platform standard
Twelve configured agents, one operating standard.
Different brands, different boundaries, identical governance: reply interruption, local-hours limits, idempotent writes, terminal suppression, one owner and one next action.
MilesReal estate
Buyer / seller qualification and handoff
LucaRecovery
Lead recovery, re-engagement and appointment rescue
BenIntake
New opportunity intake and owner routing
ColeHVAC
Emergency triage and job booking
NoahRoofing
Leak, inspection and claim routing
JackPlumbing
Containment triage and dispatch
LucasDental
Pain vs. routine, appointment booking
MasonSolar
Homeowner qualification, long-cycle nurture
GrantMed spa
Treatment interest and consultation booking
DrewPest control
Pest type, severity, treatment booking
EliLegal
Matter intake and conflict routing
MattFranchise
Franchise candidate qualification
One approved EchoCody male voice identity across the roster. Brand prompts, discovery sets and safety boundaries are distinct per agent and versioned separately.
20 certification scenarios per bot · 260 deterministic routing examples across the roster.
These are configuration tests an agent must pass before deployment — no-answer progression, duplicate webhook, reply interrupt, booking, report QA failure, no-show, reschedule, STOP, wrong person, manual control, vertical restriction and tool failure. They are not customer performance metrics.
10Security and control
What we do, and what we do not claim.
Controls that exist today, stated plainly. Where we have no certification, we say so.
Secrets stay server side
API keys, webhook URLs and proprietary prompts never reach browser code. The page holds no credentials.
Signed webhooks
Inbound call and disposition events are verified before anything is written.
Least privilege
Each integration gets only the scopes its job requires.
Consent and DND
Consent and do-not-contact state is checked before every automated action, and terminal states suppress outbound.
Idempotency
Every write carries an idempotency token and a compare-and-set on stage, so a duplicate event cannot double-move a record.
Redacted logs
Phone numbers, transcripts and personal details are redacted in operational logs.
Human takeover
A manual-control tag or an active owner conversation pauses automation immediately.
Auditable decisions
Prompt and rule changes are reviewed, approved, versioned and traceable to a date and a person.
We do not claim SOC 2, ISO, HIPAA or PCI certification. If a compliance review is part of your process, ask and we will tell you exactly what exists today. Privacy · Terms
11Capture & Workflow Review
See your actual call flow before you buy.
You get a modeled gap based on your numbers, a stage map of your current flow, and one fix-first recommendation with an owner and a measurable end result.
- 01You submit the form. It goes straight to our CRM — nothing is stored in your browser.
- 02We reply by email to confirm the review and ask for anything missing.
- 03We map your current call flow and build the modeled gap from your inputs.
- 04You get the review and one fix-first recommendation. If it is not a fit, we say so.
Or skip the form
(801) 980-030812Questions
What people ask before they buy.
Does the agent tell callers it is AI?
Yes. The agent identifies itself as an AI assistant at the start of the conversation, and again any time a caller asks.
Are calls recorded, and how is consent handled?
Calls with the agent are recorded and transcribed so outcomes can be verified. Callers are told before the conversation continues, and recording rules are configured per state and per channel.
Can a human take over a conversation?
Yes. A manual-control tag or an active owner conversation pauses automation immediately, and the record keeps its owner and next action.
Who owns the data in the CRM?
You do. Your CRM stays the system of record. The agent reads approved context from it and writes structured dispositions back to it.
What happens after hours?
Inbound calls can be answered at any hour within the rules you approve. Automated outbound is different — it runs only between 08:00 and 20:59 in the prospect's local time.
Why the 8am–9pm outbound boundary?
It is the conservative window for automated outbound contact. At 21:00 local the cadence stops and holds until 08:00. If a timezone is unknown, the system holds and creates an owner task instead of guessing.
Can it actually book appointments?
Yes, when calendar access and booking rules are configured. Where booking is not allowed, the conversation ends in a callback with a named time or a qualified handoff — never an open loop.
How does report delivery work?
When a report is part of the flow, it is generated from real booking facts and notes, QA-checked by a human, and delivered before the call — at T-24h when the meeting is more than a day out, immediately after QA when it is sooner. A placeholder report is never sent.
What happens if a tool or integration fails mid-call?
The agent closes the conversation gracefully, records what it did capture, creates an owner task, and queues the failed write for retry with an idempotency token so nothing is duplicated.
Which systems does it integrate with?
It is built around GoHighLevel as the system of record. Connections to other calendars, CRMs and phone systems are available and confirmed during setup — we do not list an integration as connected until it actually is on your account.
How is pricing structured?
A one-time build fee that scales with the size of the configuration, plus a monthly coverage plan based on headcount and volume. Pricing detail is in the pricing menu in the header.
What will the system not do?
It will not give medical, legal or financial advice, quote outside approved ranges, make earnings or savings guarantees, apply pressure or false urgency, contact anyone on DNC, or continue after a caller says stop. Those paths escalate to a human or end the conversation.