Custom Dialer
One-line description. Twilio-based ISA dialer plus a Speech-to-Text and Vertex AI coaching pipeline to replace Mojo. Earliest realistic launch September 2026.
Why
Mojo creates three problems that get worse as the team grows.
No API for recordings. So coaching at scale is impossible.
Per-seat pricing. Every new ISA punishes the budget.
Vendor lock-in on the most critical workflow.
Owning the dialer means owning the data, the coaching loop, and the cost curve.
But the real prize isn't the dialer. It's the coaching pipeline. Twilio recordings flow to GCS, Speech-to-Text transcribes them, Vertex AI Gemini scores transcripts against the 2-gate qualification framework, ISAs get daily scorecards with timestamped feedback. No other team has this surface.
Current state
Captured but not active.
Spec is written. Cost math is done. Compliance posture is mapped.
May 1 launches on Mojo. Pulling this dialer forward would risk the launch. Discipline says don't.
The plan has two phases. First, build the coaching pipeline against Mojo recordings (May–July). That's the cheap version of the prize. If it moves the needle on ISA performance, the dialer build pays back. If it doesn't, save 6-8 weeks and stay on Mojo.
What pulls this forward: Phase 1 coaching pipeline shows real ISA improvement at the end of July. Or Mojo hits a hard limit (recording bulk-export breaks, pricing jump).
Earliest realistic dialer build: July 2026. Earliest production cutover: September 2026.
- Status: planning
- Last update: 2026-04-26
- Blocked on: May 1 sales motion launch must ship first. Coaching pipeline against Mojo recordings is the gate to the dialer build itself.
Next 3 actions
(none — surface when relevant)
Decisions log
- Portfolio organization — vault README is the front door, every project gets a hub from the template
Locked decisions (captured in idea-inbox doc):
- Platform is Twilio. Not Vonage, Bandwidth, or Plivo. Premium pricing buys ecosystem maturity.
- ISA UI is Streamlit. Lowest build cost. Twilio Voice SDK works in browser.
- Coaching pipeline is post-MVP, after 30+ days of call data exist.
- Voicemail strategy at MVP: live voicemails only. Auto-drops are TCPA-gray. Defer pending attorney review.
Open issues
- Verify Mojo recording bulk-export is viable (call Mojo support, test sample download into GCP STT). Pre-May-1 task.
- Decide number strategy: single Capri DFW number vs rotating pool (recommendation: start single, expand if flagged).
- Decide coaching pipeline scope at v1 vs v2.
- Phase 2 decision point at end of July: did the coaching pipeline move ISA performance? Yes = build dialer. No = stay on Mojo.
References
- Idea-Inbox doc:
~/CCPJ/projects/Idea-Inbox/idea/2026-04-26-custom-dialer-mojo-replacement.md - Related: External API Platforms (Twilio section), GCP Enhancement Candidates (Speech-to-Text + Vertex AI items)
- Code home (when built):
~/CCPJ/projects/pos-dialer/(new service)