DEPLOYMENT PROPOSAL · DP-2026-071
WORK SAMPLE — CONFIDENTIAL
Forward Deployed Engineering

Apex Home Services × Bland

PREPARED BY Benjamin Jacobs, FDE DATE July 2026 SCOPE Inbound voice — dispatch & booking READ TIME 3 min

You haven't hired me yet, so I deployed myself. This is the work product your customers would get.

01

Where Apex is bleeding

Apex runs 40 trucks across Phoenix — HVAC and plumbing. Phones are covered 6am–6pm by three dispatchers; overnight, an answering service takes messages. The leak is after hours, where the calls are most urgent and the coverage is weakest.

≈860
after-hours calls per month ring out unanswered
35% after-hours share × <18% pickup — ServiceTitan panel, ~7,000 operators
67%
of unanswered callers dial a competitor next — most never call back
Angi research; ~85% who hit voicemail leave no message (Z360)
$49k/mo
walk-away revenue, priced at the low end of emergency job values
$850 avg plumbing emergency — ServiceTitan dispatch dataset
Assumption, stated: 3,000 inbound calls/mo (~100/day) for a fleet this size. Every other figure is sourced (see §04). Every rounding goes against the proposal — the real number is likely worse.
02

The Pathway

The full flow, expressed in Bland's Conversational Pathways vocabulary — six node types, one Global Node, condition loops that hold until required variables are captured. This is the build plan, not a sketch.

Fig. 02 — Conversational Pathway · Apex inbound dispatch

intent = emergency intent = booking intent = question answered → back to triage safety override — from any node → transfer call DEFAULT Greeting & consent records notice · brand open
Default node. "Thanks for calling Apex Home Services! This is Riley on a recorded line — what can I do for you today?" Warm open, named persona, consent woven in — two-party-safe for multi-state fleets.
 GLOBAL NODE Safety override gas · smoke · CO keywords
Global Node — takes precedence over the agent's condition decisions, from any point in the call. Gas smell, smoke, CO alarm → safety script ("leave the building now") → immediate Transfer Call to on-call. No exceptions, no AI judgment.
 WAIT FOR RESPONSE Triage loops until {{intent}} + {{callback_number}} captured
Wait for Response with a condition: the agent stays on this node until the condition is fulfilled — intent classified and a callback number extracted. No caller proceeds unidentified; the #1 failure of naive phone bots.
 DEFAULT Emergency intake safety questions · address · severity read
Default node. No heat with an infant in the house is a different call than a warm fridge. Scripted severity questions, calm register, address confirmed before handoff.
 TRANSFER CALL On-call dispatcher final words → warm transfer
Transfer Call node — the agent says its final words ("connecting you to our on-call tech now, stay on the line") before the transfer fires. Caller never hears dead air.
 WAIT FOR RESPONSE Job details {{address}} · {{issue}} · {{time_window}}
Wait for Response loop. Three variables, extracted conversationally — not an IVR interrogation. Condition holds the node until all three exist.
 WEBHOOK Book in FSM/CRM returns {{job_id}} · {{confirmed_window}}
Webhook node — fires mid-conversation against the field-service system (ServiceTitan/Jobber/housecall). Returned variables are used directly in dialogue. If it fails: see §03 — the caller is never dead-ended.
 DEFAULT Confirm + SMS reads back {{confirmed_window}} texts confirmation
Default node. Read-back closes the loop; SMS confirmation lands before the call ends. Multi-channel is native — same platform, no bolt-on.
 END CALL Done clean close
End Call node. Terminates after dialogue completes — no awkward hang, no orphaned line time billed.
 KNOWLEDGE BASE Price book & FAQ service area · hours · membership plans
Knowledge Base node — grounded in Apex's actual price book and service-area docs, so answers are Apex's answers. Then back to triage: an answered question is a booking opportunity.
Global Node
Safety override
Gas / smoke / CO keywords override every node, from any point in the call → safety script → immediate transfer to on-call. No exceptions.
Default
Greeting & consent
Recording notice first, then the brand open. Two-party-safe script.
↓
Wait for Response
Triage
Condition loop holds until {{intent}} and {{callback_number}} are captured. Nobody proceeds unidentified.
A · Emergency
Default
Emergency intake
Severity questions, address confirm, calm register.
↓
Transfer Call
On-call dispatcher
Agent says its final words, then warm-transfers. No dead air.
B · Booking
Wait for Response
Job details
{{address}}, {{issue}}, {{time_window}} — extracted conversationally, loop holds until all three exist.
↓
Webhook
Book in FSM/CRM
Fires mid-conversation; returned {{job_id}} and {{confirmed_window}} are spoken back to the caller.
↓
Default
Confirm + SMS
Read-back plus SMS confirmation before hangup.
↓
End Call
Done
Clean close after dialogue completes.
C · FAQ
Knowledge Base
Price book & FAQ
Grounded in Apex's real docs, then returns to triage — an answered question is a booking opportunity.
Watch a call run
simulated transcript.
the live line in §04 is the real thing.
hover any node for build notes
The same pathway, built in Bland's Pathways editor
Fig. 02b — The same pathway, rebuilt in your Pathways editor.
03

Edge cases & failure modes

A pathway that only handles the happy path is a demo, not a deployment. These six are designed in from day one.

EC-01

Gas leak mid-booking

Caller books a tune-up, then mentions a smell. The Global Node overrides whatever node is active — safety script, immediate human. AI never adjudicates a safety call.

GLOBAL NODE → TRANSFER CALL
EC-02

Angry repeat caller

Second call about the same no-show: de-escalate once, acknowledge history by callback number, then hand to a human. One apology loop max — nothing enrages like a bot that won't yield.

CONDITION → TRANSFER CALL
EC-03

Dead air

Silence past 6 seconds: one re-prompt, then offer a callback to {{callback_number}}, then a graceful close. Never an infinite "are you still there?" loop.

RE-PROMPT → END CALL
EC-04

Recording consent

Arizona is one-party, but the script is written two-party-safe — consent in the greeting — so the same pathway survives a multi-state fleet without legal review per state.

GREETING SCRIPT · COMPLIANCE
EC-05

Booking webhook down

FSM API fails mid-call: the agent captures the job details anyway, promises a confirmation text within the hour, and flags dispatch. The caller never learns the backend hiccupped.

WEBHOOK FALLBACK · NO DEAD END
EC-06

2am non-emergency

After-hours rules: true emergencies dispatch the on-call tech now; everything else books the first morning slot with SMS confirmation. The on-call tech's sleep is a resource too.

AFTER-HOURS RULESET
04

Proof you can call right now

Proposals are cheap. Working systems aren't.

Apex dispatch line — demo LIVE
+1 (602) 637-1818
Try to break it: book a job, invent a gas leak, go silent, interrupt it, ask if it's an AI.

This is a reference implementation running on my own production stack — I built NinjaOtter, a voice + SMS receptionist, solo, from Twilio up. It maps node-for-node to the pathway in §02. Rebuilt in Pathways, it inherits what I can't give it alone: your enterprise controls, testing harness, and self-hosted models.

The math, shown

STEP 13,000 calls/mo × 35% after-hours share (low end of 35–45%)1,050 calls
STEP 2× 82% unanswered (pickup <18%, traditional shops)≈860 missed
STEP 3× 67% who dial a competitor next≈575 lost callers
STEP 4× 1-in-10 bookable emergency × $850 (low-end plumbing; HVAC avg is $1,400)≈$49,000/mo
Annualized walk-away revenue≈$588,000

Sources: 2026 Contractor Missed Call Report (higrovi.com — Invoca call logs; ServiceTitan panel of ~7,000 operators; ServiceTitan dispatch dataset; Z360 SMB benchmark) · Angi competitor-call research via agentzap.ai. Every rounding favors the skeptic.

05

30 days to production

Mirrors Bland's own deployment cadence. Pilot the bleeding artery first — after-hours — with the human fallback intact until the gates say otherwise.

DAYS 1–5

Discovery

Shadow the dispatchers. Pull 90 days of call logs. Map the booking flow into the FSM.

DAYS 6–12

Build

The §02 pathway in the Pathways editor. Webhook against the FSM sandbox. Knowledge Base from Apex's price book.

DAYS 13–19

Pilot

After-hours calls only. Human fallback intact. Every call reviewed.

DAYS 20–26

Measure

Against the go/no-go gates below. Miss a gate → fix the pathway, not the report.

DAYS 27–30

Scale

Daytime overflow. Dispatcher training. Handoff runbook — Apex owns it.

Go / no-go gates
  • ≥ 60% containment
  • ≥ 25% after-hours calls → booked jobs
  • < 5% false-transfer rate
  • 0 safety-protocol misses

"If the pilot misses the gates, we fix the pathway, not the report."

06

Deployment notes

The part a proposal usually hides: what I'd distrust about my own plan. Written for people who deploy these for a living.

DN-01

What this page can't know

Every dollar figure downstream of the 3,000 calls/mo assumption is provisional. Discovery days 1–5 exist to kill that assumption with Apex's real call logs and booking flow — and re-derive the math from whatever survives.

DN-02

Where this is weakest

The agent's tone was tuned by ear against live calls on my own stack — not against a regression harness. First move inside Bland: put the pathway under the platform's testing framework and simulate callers before the pilot takes a real one.

DN-03

What breaks first at 10×

The webhook path. Field-service APIs rate-limit and slow down mid-call exactly when volume spikes. EC-05 dead-end-proofs the caller experience, but capacity planning against the FSM needs real traffic, not this page.

DN-04

Built from public docs alone

Everything here — node semantics, Global Node behavior, condition loops, variables — comes from docs.bland.ai read cold, with no console access. That's the point: it's what self-onboarding looks like before day one.