Sheppard

Build it. The systems mode of Sheppard's practice.

Every paid click becomes a booked job.

We build the customer-journey systems underneath your marketing. AI call answering, booking automation, CRM and dispatch integration (ServiceTitan, FieldEdge, Housecall Pro), attribution warehouses that tie ROAS to booked revenue, applied AI workflows. Owned by your operation. Transferable at exit.

Operating Thesis

The marketing engine is software now.

Five years ago the marketing P&L was paid media and a few agencies. Today the function lives in AI voice agents, attribution warehouses, dispatch integrations, lead-scoring models, and the operational tooling that ties it all back to booked revenue. The operator that owns its software compounds faster than the operator that rents it.

We build that software inside your operation. We don't sell it as a SaaS. You own it. The next buyer pays for it on exit.

01

AI Call Answering & Booking

Every paid click becomes a booked job.

The largest leak in residential home services marketing is the call that doesn't get answered or doesn't get booked. We wire AI voice answering, live-agent overflow, and CRM-integrated booking into the marketing operating system so spend converts to revenue, not abandoned calls.

What We Ship

  • 24/7 AI voice answering with natural-language intake and qualification
  • Live-agent overflow for emergency, high-ticket, or complex calls
  • Direct booking into the dispatch system (ServiceTitan, FieldEdge, Housecall Pro, Service Fusion)
  • Call recording, transcription, and sentiment analysis for QA and coaching
  • Speed-to-answer monitoring as a marketing KPI (booked-call rate by hour)
  • Missed-call recovery flows via SMS and outbound callback

Stack

AI voice agents (Synthflow / Bland / Vapi class) · twilio · ServiceTitan / FieldEdge / Housecall Pro APIs · CallRail or equivalent · SMS recovery · Postgres event log

Recent Build · End-to-end call infrastructure

We rebuild the call layer so booked-call rate becomes the operating metric the CFO reads, not impressions or clicks. The infrastructure sits between ad spend and dispatch and quietly compounds revenue per paid dollar.

02

Lead Operations Infrastructure

The unified lead pipeline most operations don't have.

The lead-ops layer connecting marketing channels, CRM, dialer, and dispatch into one auditable pipeline. CFO-grade. Sponsor-grade. Built so the next owner inherits a working system, not a vendor relationship.

What We Ship

  • Unified lead taxonomy across channels and locations (source, type, disposition, channel, DMA)
  • Salesforce / HubSpot / ServiceTitan / LeadSquared sync layer
  • Sync-health monitoring with failure alerts and replay tooling
  • Audit trail and reconciliation against the financial system
  • CFO-readable dashboard with monthly EBITDA-contribution view
  • Multi-location platform view for operators running roll-ups

Stack

Next.js · Supabase · Postgres · Salesforce / HubSpot APIs · cron-based sync workers · Vercel

Recent Build · Multi-DMA home services platform

Salesforce-synced lead operations dashboard with parallel cron-based sync, audit-grade failure tracking, auth revocation, and an admin sync-health UI. SQL aggregation tuned for sponsor-facing reporting cadence.

03

Attribution & Reporting Warehouses

Spend reconciled to booked revenue.

The data layer that ties every marketing dollar back to the financial system. Multi-touch attribution with audit trail. ROAS measured against actual booked-revenue, not platform-reported conversions. Reporting your CFO can sign off on.

What We Ship

  • Multi-touch attribution methodology reconciled to the financial system
  • ROAS calculated against booked revenue (not platform-reported conversions)
  • Server-side tagging, consent management, and Conversions API integration
  • Channel-mix dashboards normalized for trade economics and seasonality
  • CAC and LTV models by channel, DMA, and trade
  • Sponsor-facing monthly EBITDA-contribution reporting on the close cadence

Stack

Postgres · dbt-style transforms · GA4 / server-side tagging · CallRail / dynamic number insertion · Meta Conversions API · custom warehouse layer

Recent Build · Attribution warehouse for multi-DMA platform

Replaced a default GA4-only attribution stack with a custom warehouse reconciling paid spend, call tracking, CRM dispositions, and dispatched revenue. Sourced-revenue numbers reconciled to the GL for the first time in the platform's history.

04

Applied AI Workflows

AI where it moves a KPI, not where it demos well.

The AI tooling woven into the marketing operating system. Lead scoring, predictive LTV, content generation pipelines, creative iteration, dispatch-side coaching, and the next layer of AI that's actually moving operators' numbers. Not generic AI for AI's sake.

What We Ship

  • Lead scoring and predictive booked-call probability
  • Predictive LTV by acquisition channel for budget allocation
  • AI-assisted content production for local pages, GBP, and ad creative at DMA scale
  • Creative iteration loops for paid social and paid search ad copy
  • Dispatch-side call review and tech coaching from recorded interactions
  • Evaluation harnesses so every AI workflow gets measured, not assumed to work

Stack

Claude / OpenAI APIs · LangChain / custom orchestration · vector store for content reuse · evaluation harnesses · human-in-the-loop review

Recent Build · AI scoring + content automation

Built into operating partners' marketing stacks where the AI workflow is measured against a KPI (cost per booked call, content production throughput) rather than installed as a category and assumed to add value.

Who We Build For

Built with operators. Owned by operators.

Operator Engagements

  • Independent home services operators

    HVAC, plumbing, electrical, roofing operators doing real volume who need the call layer, the booking layer, and the reporting layer wired together.

  • Multi-location operators and roll-ups

    Multi-DMA operators consolidating lead operations and attribution into one platform-grade system.

Builds scoped to the operation's existing stack rather than imposed as a SaaS vendor's seat license.

PE Sponsor Engagements

  • Lower-middle-market PE sponsors

    Diligence, value-creation, and infrastructure builds across portfolio companies.

  • Growth-equity operating partners

    Cross-portfolio marketing oversight, attribution standardization, and AI workflow rollout.

Engaged at the platform level. Production infrastructure built inside each portco's perimeter, owned by the operator, transferable at exit.

Frequently Asked

On the systems we build.

What does Sheppard build for PE-backed home services portcos?

+
Two core build categories: lead-operations infrastructure (Salesforce-synced lead pipelines, sync-health monitoring, audit-grade attribution, sponsor-facing dashboards) and performance operations tooling (attribution methodology, lead routing, lifecycle automation, content pipelines, and AI-assisted scoring where it moves a KPI). Builds are scoped to the portco's existing tech stack and deployed inside the portco's data perimeter, not behind a SaaS vendor's seat license.

Do you sell software as a SaaS?

+
No. Sheppard does not sell SaaS, does not bill by the seat, and does not lock portcos into our tooling. Everything we build is owned by the portco, documented, and transferable to the next sponsor at exit. The buyer at exit inherits a working system, not a vendor relationship to renegotiate.

What's the typical timeline for a Sheppard build engagement?

+
Most lead-operations and AI builds ship in 60-90 days. Larger multi-portco platform installs run 90-120 days. The Day-1-to-Day-100 cadence is intentional, sponsors need the system in production for the 100-day operating-partner reporting cycle, not delivered in month six.

What tech stack does Sheppard build on?

+
Modern, portable, production-grade. Typical stack: Next.js or React for dashboards, Supabase or Postgres for data, Salesforce / HubSpot / ServiceTitan / LeadSquared APIs for CRM integration, Anthropic Claude or OpenAI APIs for applied AI workflows. We build in the portco's existing cloud environment when possible, Vercel, AWS, or GCP, to keep the data perimeter clean and the exit transfer simple.

Does the portco own what Sheppard builds?

+
Yes. Source code, infrastructure, data, and documentation are all owned by the portco from Day 1. Sheppard retains no proprietary lock-in. This is structural, sponsors are buying assets that compound through hold and transfer cleanly at exit, not vendor dependencies that destroy negotiation leverage at sale.

Build it. One of three modes.

Have a system that needs to be built?

Call answering and booking infrastructure, lead-ops platform, attribution warehouse, or AI workflow. Most builds ship inside 90 days.