Transcribe client onboarding calls and generate engagement letter drafts
This solution transforms how accounting firms onboard clients by turning spoken conversations directly into ready-to-sign engagement letters. Pitch this to CPA firms to help them eliminate administrative bottlenecks, secure client signatures faster, and protect themselves from scope creep.
The problem today
5 days
delay in sending out new engagement letters
40
unbillable hours lost to manual note-taking monthly
Marcus Ellison is the managing partner of a 9-person CPA firm in Columbus, Ohio, handling tax prep, bookkeeping, and small business advisory for about 300 clients. He has a draft engagement letter sitting in his outbox right now that he started on Tuesday and still hasn't sent because he can't remember whether he quoted the client $1,800 or $2,200 for the advisory retainer.
01The Problem
Non-billable note-taking and recall effort consumes partner time before any deliverable exists.
Prospects sign with competitors or arrive at kickoff without an executed agreement in place.
Verbal fee exceptions reconstructed from memory become the foundation for expensive disputes months later.
Partners pull from revenue-generating work to fix letters junior staff cannot be trusted to draft correctly.
Disputed scope or fees resolve to competing recollections — the firm holds no documentary evidence.
Partners drafting inconsistently across service lines creates uneven client exposure and audit risk.
02The Solution
Solution Brief
Fictional portrayal · illustrative
- Marcus runs ~25 onboarding calls/month across 300 clients
- Every call ends with handwritten notes and an unsent draft
- One letter stalled since Tuesday — $1,800 or $2,200 quoted, Marcus can't confirm
- 3–5 day drafting lag costs signed clients — two went quiet this week
- Missed fee terms and verbal commitments surface later as scope disputes
- No call record means no evidence when a client contests what was promised
- Inconsistent drafting across partners creates both relationship and audit exposure
- Call records and transcribes automatically; draft appears before Zoom closes
- Structured letter — scope, fees, deadlines, responsibilities — pulled from actual conversation
- Staff edits in 10 minutes; DocuSign link out before lunch
- IRS §7216 consent and state recording law setup handled at implementation
- $149–$249/mo recurring after $5,000–$10,000 implementation; removal breaks the entire onboarding process
“I quoted a client a number on a call in February and by April neither of us could agree on what it was. We both thought we were right. I ended up eating $600 just to keep the relationship. That doesn't happen anymore — if there's ever a question, I pull up the transcript in thirty seconds and we're done arguing.”
— Marcus Ellison is the managing partner of a 9-person CPA firm in Columbus, Ohio, handling tax prep, bookkeeping, and small business advisory for about 300 clients
03What the AI Actually Does
Call Transcription Layer
Captures every onboarding conversation — in-person or virtual — and produces a clean, timestamped transcript automatically. No manual recording, no note-taking. The full conversation becomes a searchable, retrievable record the moment the call ends.
Engagement Detail Extractor
Reads the transcript and identifies every term that belongs in an engagement letter: service scope, fee amounts, payment schedules, filing deadlines, client responsibilities, and any verbal exceptions or special agreements made on the call.
Engagement Letter Drafter
Assembles the extracted details into a properly structured draft engagement letter, formatted to the firm's template and routed directly into their practice management system for partner review and e-signature — typically ready within minutes of the call ending.
Compliance Configuration Engine
Manages IRS Section 7216 consent language and state-specific recording consent workflows before each call, so every transcript the firm captures is legally defensible and audit-ready without requiring staff to remember the rules themselves.
04Technology Stack
Fireflies.ai Business Plan
$19/user/month (annual billing) or $29/user/month (monthly billing) — typical 5-user firm = $95–$145/month
Core AI transcription platform. Automatically joins Zoom, Teams, and Google Meet calls as a bot participant. Provides speaker diarization, searchable …
OpenAI API (GPT-5.4)
$5–$15/month for typical accounting firm volume (20–40 calls/month). Input: $2.50/1M tokens, Output: $10.00/1M tokens. A single engagement letter generation costs ~$0.02–$0.05.
Large language model API used to process transcription summaries and generate structured engagement letter drafts. GPT-5.4 provides the nuanced profes…
Make.com (formerly Integromat) Pro Plan
$16/month (10,000 operations) — sufficient for 40+ calls/month with multi-step automations
Integration middleware that orchestrates the end-to-end workflow: receives webhook from Fireflies.ai when transcription completes → sends transcript t…
Ignition (formerly Practice Ignition) — Starter or Pro Plan
Starter: $49/month (1 user); Pro: $229/month (billed annually) or $279/month (monthly). Includes e-signature, automated billing, and engagement letter templates.
Accounting-specific engagement letter and proposal platform. Receives AI-generated draft data via API/Zapier and populates professional engagement let…
SmartVault Business Plan
$20–$40/user/month depending on tier; typical firm = $100–$200/month
SOC 2 Type II certified document management platform purpose-built for accounting firms. Stores signed engagement letters, call recordings (if retaine…
Fireflies.ai API Add-on
Included in Business subscription — no additional cost
Programmatic access to completed transcripts, summaries, and speaker-labeled segments. The Make.com webhook integration uses this API to automatically…
05Alternative Approaches
Microsoft Teams Premium with Copilot
~$10/user/month for Teams Premium vs. $19/user/month for Fireflies Business
Instead of Fireflies.ai + OpenAI, leverage Microsoft Teams Premium which includes built-in intelligent meeting recap with AI-generated notes, action items, and task assignments via Microsoft Copilot. Engagement letter generation would still require a separate LLM step but could use Microsoft Copilot Studio or Azure OpenAI Service for a fully Microsoft-native stack.
Strengths
- Single vendor ecosystem (Microsoft)
- No third-party bot joining meetings (better client optics)
- Integrated with SharePoint/OneDrive for document storage
- May satisfy compliance teams more easily
Tradeoffs
- Teams Premium transcription quality is generally lower than Fireflies for specialized vocabulary
- Copilot meeting summaries are less customizable
- No speaker-labeled transcript export API as robust as Fireflies
- Requires Microsoft 365 E3/E5 or Business Premium as a base
- Engagement letter automation pipeline still needs custom development
Best for: Firms already all-in on Microsoft 365 and resistant to adding third-party tools, or when client-facing optics of a bot joining the call is a concern.
Otter.ai Business + Direct Integration
Approximately $20–$30 more per month for a 5-user firm
Replace Fireflies.ai with Otter.ai Business which offers real-time transcription with live captions during the call, stronger custom vocabulary support, and a dedicated OtterPilot feature that generates summaries and action items. Use Otter.ai's Zapier integration to feed into the same downstream GPT-5.4 pipeline.
Strengths
- Real-time transcription visible during the call (partners can see live captions)
- Stronger brand recognition with clients
- Excellent custom vocabulary for accounting jargon
- Slightly better real-time accuracy in many benchmarks
Tradeoffs
- $1/user/month more expensive than Fireflies at Business tier
- Webhook/API capabilities are less mature than Fireflies for automated pipelines
- Zapier integration (not webhook) adds latency and requires Zapier subscription ($19.99/month vs. Make.com $16/month)
Best for: Firms that value real-time live captions during calls for note-taking purposes, or when transcription accuracy is the primary concern and the firm is willing to pay a small premium.
Fully Custom Pipeline with Deepgram + AssemblyAI
API costs may be 50–80% lower than SaaS subscription at scale, but development time adds $5,000–$15,000 to initial project
Replace the SaaS transcription platform entirely with direct API calls to Deepgram (for real-time streaming transcription) or AssemblyAI (for batch transcription with built-in PII redaction). Build a custom web application or use a low-code platform like Retool to create a firm-branded transcription and engagement letter portal.
Strengths
- Lowest per-minute transcription cost ($0.0025–$0.0092/min vs. flat SaaS subscription)
- White-label opportunity for MSP (build once, deploy to many clients)
- PII redaction built into AssemblyAI (stronger than regex)
- No third-party bot joining meetings (direct audio file upload)
- Complete control over data residency and processing
Tradeoffs
- Significantly higher implementation complexity (80–120 hours vs. 40–65 hours)
- Requires developer skills (Python/Node.js, API integration, possibly a simple web frontend)
- Ongoing maintenance burden is higher
- No out-of-the-box meeting bot (must record locally and upload)
- Audio quality depends entirely on recording setup
Best for: MSPs serving 10+ accounting firms wanting to build a reusable branded product; firms with strict data sovereignty requirements; or when call volume exceeds 100+ hours/month making per-minute pricing advantageous.
Engagement Letter Platform Native AI (Ignition AI / GoProposal)
May be included in existing subscription or require a premium tier (+$10–$30/month)
Wait for Ignition and GoProposal to release their own AI-powered engagement letter generation features (both have announced AI roadmap items). Use a simpler transcription-only setup and rely on the engagement letter platform's built-in AI to generate drafts from imported meeting notes rather than building a custom GPT-5.4 pipeline.
Strengths
- Eliminates the custom GPT-5.4 pipeline entirely (no prompt engineering, no Make.com scenario for letter generation)
- Single vendor support for the engagement letter output
- Likely better integration with the platform's template system and e-signature workflow
- Lower ongoing maintenance
Tradeoffs
- These AI features are not fully available as of early 2025 (roadmap items)
- The quality and customizability of vendor AI may be inferior to a tuned GPT-5.4 prompt
- Vendor lock-in increases
- No control over the AI model or prompt
Best for: Firms already using Ignition or GoProposal and willing to wait 3–6 months for native AI features to mature, or when the MSP wants to minimize custom development and ongoing prompt maintenance.
Budget Approach: Fathom Free + Manual Letter Creation
$0/month for transcription + existing template tools. Implementation: 4–8 hours
Use Fathom's unlimited free transcription tier for all onboarding calls. Partners manually review the AI-generated summary and action items from Fathom, then create engagement letters using existing templates in Word/Google Docs. No automation pipeline, no LLM API costs.
Strengths
- Zero monthly software cost for transcription (Fathom free tier is genuinely unlimited)
- No complex automation to build or maintain
- No API keys to manage
- Fastest time to value (can be set up in 2–3 hours)
- Lowest risk
Tradeoffs
- No automated engagement letter generation (partners still manually create letters)
- No integration with practice management systems
- The entire time-saving benefit of the LLM engagement letter pipeline is lost
- Scales poorly as call volume grows
- Limited admin controls and no API access on free tier
Best for: Firms skeptical about AI, with very low call volume (under 5 onboarding calls/month), budget constraints, or wanting a low-risk proof of concept before investing in the full pipeline. Can serve as a Phase 0 before upgrading to the full solution.
Ready to build this?