Fourth Industrial · Prepared for BP Consulting / BenefitsPensions.com

Give every BDR an always-on team of AI agents.

One reasoning brain sits behind every channel: voice, email, and SMS. It works the top of your funnel across extended hours, into the evenings and weekends your desk doesn't cover. It books qualified meetings for your team to run. Your people spend their time on live conversations, not dialing.

PROPOSAL BRIEF · v1 · 2026-07-05
Your team, amplified
3 channels · extended hours
Voice · email · SMS. One brain per rep.
Speed to first touch
< 5 min · incl. after-hours
vs. hours, business-hours-only today
Cost per qualified meeting
~$8–$20
the same meeting your desk sets for ~$400+
The opportunity

Your 10 BDRs place 120–200 dials a day to set roughly 5 qualified meetings each per week. That is about 200 qualified meetings a month, on one channel, in business hours. Every rep hits the same ceiling: the number of dials one person can make in a day.

This engine lifts that ceiling. It gives each BDR a team of AI agents that works the tireless top of the funnel: the dialing, the instant follow-up across voice, email, and SMS, the after-hours speed-to-lead. Your reps stop grinding the phones. They start doing what only a human can: the live conversations, the judgment calls, the relationships. Same team, far more reach.

It works the consent-bearing surface of your list first: warm leads, existing-relationship contacts, inbound opt-ins. That is where the compliant, defensible motion already lives.

Multi-channel activation

One agent, working the channels your buyers actually use.

Same brain, same memory, across every channel. A prospect who takes a call, then gets an email, then replies by text is one continuous conversation, not three disconnected touches. We lead with the channels that fit a benefits and pensions buyer, and skip the consumer-DM noise.

Voice & callsflagship

Natural AI voice on warm and opted-in lists. It qualifies, handles objections, and books the meeting, or warm-transfers live to a rep. Sub-second, natural turn-taking, and it dials only inside permitted contact hours.

Email

Personalized sequences and inbox replies, CAN-SPAM and CASL-clean, tuned to each contact's history. The B2B workhorse alongside the phone.

SMS

Opt-in follow-up, meeting reminders, and two-way replies that keep booked meetings from slipping. STOP honored, every message logged.

The point isn't more tools. It's one relationship. Every touch on every channel reads and writes to the same HubSpot record, so your rep inherits a warm, fully-briefed conversation instead of starting from zero.

What we've built

Proof we can deliver this, not a slide deck of someone else's logos.

Selected delivery work from Fourth Industrial. Clients anonymized. Shown as capability: what we have actually designed, built, wired, and test-validated.

AI Voice · CRM-wired

A live AI phone agent on a natural, cloned voice

For a ~$90M home-services company we built and activated an Agentforce voice agent. An ElevenLabs voice answers a real inbound phone number, wired into the CRM to route and book service appointments through a governed booking pipeline.

Built and activated in a controlled environment; inbound; on Salesforce. Proves we stand up real AI phone agents wired to a CRM.
✓ Real number answered · ElevenLabs voice · CRM-routed booking
Outbound · the exact motion

A consent-gated qualify-and-book outbound agent

We designed and built a multichannel outbound agent whose flow is precisely what you're buying: a consent gate first (built to CASL / PIPEDA), then qualify, handle objections, and book the meeting or warm-transfer to a licensed human.

Text channels (SMS / email) on Salesforce; the same architecture we'd point at voice for you.
✓ Consent-first · qualify → book → human handoff
Compliance-grade automation

Agentic deal automation, 100% test-covered

For a national consumer-financing workflow we built an agent that reads a loan's fundability and safely advances deals inside the CRM. Delivered with 119 / 119 passing tests and 100% code coverage across the shipped logic, plus a fully-mocked integration layer.

Back-office rather than appointment-setting. Included as proof of our engineering and compliance rigor.
✓ 119/119 tests · 100% coverage · API-led integration
Systems depth

We architect multi-agent Claude systems end-to-end

Our own operation runs a 23-agent Claude orchestration with shared memory, guardrails, and cost-controlled inference (blended cloud + local models). It's the muscle behind building your engine: agent design, evals, and guardrails are our day job.

Internal capability, not a client outcome. Shown for systems-design depth.
✓ 23-agent orchestration · memory · guardrails · evals
Straight with you: we're a young firm, and you'd be an early, marquee client. So we're not going to wave invented production metrics at you. What we'll show is real capability and a build plan we stand behind, with fixed pricing and refund-backed milestones. The proof we care about is the baseline-to-lift we'll create together, measured in your HubSpot. (For third-party validation that Claude works in this vertical, HUB International and Newfront have both publicly deployed Claude across brokerage workflows.)
Proposed architecture

One brain, many channels, one HubSpot record.

HubSpot stays your system of record. The AI voice runs on its own runtime and writes every call, consent, and booked meeting back to HubSpot via API.

Prospects Warm · EBR · inbound opt-in CHANNELS Voice Email SMS Consent + AI-disclosure gate CASL · TCPA · CRTC the compliance moat REASONING BRAIN Claude Haiku · fast turns Opus · hard turns ElevenLabs voice + SIP runtime · sub-second Qualify → Book meeting or warm-transfer HUBSPOT · SYSTEM OF RECORD Contacts · consent log · call + transcript · Meetings booked / held ← the KPI Native meeting-outcome property. The performance fee attaches here, no custom instrumentation. read context write-back Licensed advisor takes the booked meeting LOAD-BEARING DESIGN DECISION HubSpot's native calling carries no live audio stream to a third-party AI. Its SDK is metadata-only. So the AI voice runs on its own runtime (ElevenLabs + SIP) and writes the call, recording, transcript and booked meeting back into HubSpot via API.
Consent-first, alwaysThe agent only ever engages consent-bearing contacts, opens every call by disclosing it's an AI on behalf of BP, and dials only inside permitted contact hours. That is the exact posture Utah, Colorado, the FCC, and the CRTC are moving toward. This is the moat a horizontal AI-SDR tool can't sell you.
Human-in-the-loop by designThe agent books meetings and hands off. It does not give regulated advice or make automated decisions. Any complexity trigger routes to a licensed human. That keeps you clean under Law 25 and your E&O posture.
You own everythingEvery account is yours, on your HubSpot: Claude, ElevenLabs, telephony. No rented seats, no lock-in, no markup. We build and operate. You own the asset and the data.
Why this stack

Three specialists, one system, chosen deliberately.

The brain

Why Claude

The reasoning that qualifies, handles objections, and decides to book, with the controls a licensed brokerage needs.

  • Two-speed routing: Haiku 4.5 runs fast conversational turns (~5× cheaper); Opus 4.8 handles the hard qualification turns.
  • Cost is negligible: ~$0.03–0.14 of model spend per call, a rounding error against a booked meeting.
  • Governable: per-workspace usage metering, HIPAA-BAA path, and data controls for when the back-office use cases follow.
The voice

Why ElevenLabs

The layer that actually talks and listens on the phone, natural enough that prospects stay on the line.

  • Sub-second turn-taking, with ~75ms voice synthesis (Flash v2.5): the difference between a natural call and an obvious robot.
  • $0.08/min flat, with speech-to-text, text-to-speech and turn-taking all included; Claude plugs in natively.
  • Scales on concurrency, not headcount: dozens of simultaneous calls across your Toronto and Miami hours, into the evenings and weekends your desk doesn't cover.
The record

How it bolts into HubSpot

HubSpot stays the source of truth. Nothing rips or replaces. The engine reads and writes around it.

  • Reads contact, consent and history to personalize every touch.
  • Writes back the call, transcript, recording and booked meeting via API, logged like any rep activity.
  • The KPI is native: "meeting held vs. no-show" is a standard HubSpot property, so the performance fee is measured directly, with zero custom tracking.
The economics

What changes when you add the engine.

Per rep, at your stated volumes: the same person, now with an AI team working every channel across extended hours. Illustrative; the real numbers get locked in Discovery.

 A BDR todayThat BDR + the engine
Channels workedPhone, business hoursVoice · email · SMS · extended hours
Speed to first touchhours< 5 minutes
Reach~120–200 dials/day, one at a timethousands of touches, in parallel
Cost per qualified meeting~$346 – $535~$8 – $20 marginal
After-hours & weekendsnonecovered
The rep's time goes todialing + settinglive conversations & relationships

The engine's own run-cost is ~$0.12–$0.20 a minute, all-in (voice, telephony, reasoning) and dominated by the voice line. The AI reasoning behind it is nearly free. That range is the independent, benchmarked figure, not a best case. So more reach and more booked meetings don't come with a bigger headcount bill. They come with a marginal per-minute one.

Commercials

A fixed, low-risk path. You never sign a blank cheque.

Discovery is fixed and credited. The build price is confirmed only after Discovery, so you sign a known number. Our upside comes from meetings you didn't have before.

STEP 01

Paid Discovery

$12.5K · 2 weeks

We measure your real baseline (meetings, speed-to-lead, current calling spend), audit telephony, consent and residency, and produce a fixed-price build plan. 100% credited to the build.

Full refund if we don't surface ≥ $100K/yr of defensible opportunity.
STEP 02

Confirmed build

Fixed, set at Discovery-end

Once Discovery measures the real inputs, we lock a fixed build price. You sign a known number, not an estimate. Milestone-gated payments across a ~45-day sprint to a live, evaluated engine.

50% fee refund if the agreed day-45 metric isn't hit.
STEP 03

KPI performance upside

Tied to meetings booked

Above your locked ~200/mo baseline, we earn on net-new held qualified meetings: a per-meeting bounty or a capped share of the gain. Self-funding: we earn more only when you book more than you do today.

Measured directly in HubSpot; "held vs. no-show" can't be gamed.