Measured outcomes from service operations.
Representative case studies show the workflow, the system built, and the operating result.
Harbor & Vale Accounting automated 58% of tax-season intake without hiring
Fourteen staff across two offices, a 14-week tax season, and an intake process — new clients, document collection, independence checks, engagement letter — that consumed 90 minutes of associate time per client. In peak March, associates were doing intake work instead of billable preparation.
The Problem
H&V's intake workflow involved 4 touchpoints across three tools (email, document portal, CRM), plus a manual independence check against a 12-year client record. Most of the work didn't require tax judgment — but all of it required a human, because that's how it had always been done. During March, senior associates were spending 30% of their time on admin instead of returns.
What We Built
A conversational intake agent that collects required information, requests and verifies documents, runs the independence check against their full history, drafts the engagement letter from the firm's approved templates, and schedules the kickoff meeting. A partner reviews any flagged independence issue. Everything else runs without them. Live in 3 weeks.
Operating Guardrail
Substantive tax questions route to a partner with context attached. The system owns intake preparation; the firm owns tax judgment.
Summit HVAC cut quoting from 20 minutes to 3
Fourteen trucks on the road, three dispatchers in the office, and every quote required a human to check parts availability, calculate labor, and write it up. In summer, that meant 40+ quotes per day — and the ones that took longest to send were the ones that went to a competitor.
The Problem
Summit's dispatchers spent 60% of their time on quotes — checking parts, calculating labor rates for different equipment, and writing proposals. During peak season, they were sending quotes at 9pm that should have gone at 2pm. Customers were getting three quotes the same day from competitors; Summit's was always last.
What We Built
An AI agent that receives the job details from a simple form the tech fills out on-site, checks parts availability against Summit's supplier APIs, calculates labor based on equipment type and truck location, and generates a formatted quote that goes to the customer within 3 minutes. The dispatcher approves it; they do not write it.
Operating Guardrail
Quote quality depends on clear job inputs. The form now captures equipment type, visible model details, and symptom checklists; incomplete submissions route back for clarification before a customer receives a quote.
Meridian Legal automated 60% of intake without losing the human touch
Eight attorneys, 12 support staff, and an intake process that required 3 phone calls, 2 emails, a conflict check, and 40 minutes of paralegal time per potential client. Most of that work didn't require legal judgment — but it all required a human because that's how they'd always done it.
The Problem
New client intake at Meridian took 5-7 business days from first call to signed engagement letter. Conflict checks required manual searches across 14 years of client records. The intake form was a Word document emailed back and forth. Attorneys were spending 30% of their time on administrative intake tasks — time that should have gone to billable work.
What We Built
An AI agent that handles the first 60% of intake automatically: collects client information through a conversational form (not a PDF), runs conflict checks against the full 14-year database, flags potential conflicts for attorney review, drafts the engagement letter template, and schedules the initial consultation. The attorney reviews the conflict flag — that's the checkpoint. Everything else runs without them.
Operating Guardrail
Fee conversations remain attorney-led. The system prepares intake, conflict context, and scheduling, then routes the relationship-sensitive conversation to the attorney.
Your workflow has an approval point like this.
Answer six questions and get a first-pass read on what to automate, what needs approval, and what should stay with your team.