Halo AI · YC S22 · Active Role

Mulya Sakti Muhammad

B2B Product Support Specialist
I sit between Account Managers and Project Engineers — converting unstructured client business requirements into structured AI logic, so AI agents are configured correctly before they ever reach the client's hands.
📍 Jakarta, Indonesia
📅 May 2026 – Present
🏢 SaaS · Agentic AI · CX Automation
🎯 Scope: Phase 1–4 of 6-phase pipeline
30
Active B2B Projects
Managed Simultaneously
24+
B2B Client Accounts
Onboarded
~95%
Onboarding
Completion Rate
5–14
Working Days
Time-to-Value (TTV)
My Core Workflow
4-Phase AI Implementation Pipeline
What I own end-to-end — from the moment AM closes a deal, to a fully configured AI agent ready for UAT
Entry Point
Client closed
& Kickoff
completed
Phase 01
Post-Kickoff
Alignment
  • Review AI kickoff transcript — map business flow, industry constraints & automation scope
  • Cross-check with AM/PE on remaining data gaps
  • Analyze client "character type" to tailor call vs chat approach
↓ Output Validated requirement checklist + communication strategy
Phase 02
Requirement Gathering
& SLA Management
  • Manage project tracker dashboard for data-gathering status
  • SLA-based structured comms templates to chase client docs
  • Proactive escalation (H-2 alignment call) to prevent SLA slippage
↓ Output Complete client data package, collected on-SLA
Phase 03
KB Structuring &
Guardrail Definition
  • Simulate end-to-end AI–customer conversations to find gaps
  • Define AI guardrails so data is machine-readable & hallucination-proof
↓ Output Finalized Knowledge Base with validated logic, e.g. dental case: schedule params, service duration, "1 Calendar = 1 Branch" double-booking prevention
Phase 04
Technical Coordination
(Business → Engineering)
  • Handover finalized KB + updated timeline to Project Engineer
  • Translate API/ERP integration (Accurate, Jubelio, Shopify) into client-friendly instructions
  • Monitor progress with PE to clear operational blockers
↓ Output AI Agent technically integrated — Go-Live ready
Final Output
AI Agent
Ready for
UAT ✓
Behind The Scenes
Case Study Articles
Deep-dive breakdowns of how I executed the 4-phase pipeline on real client projects
🦷
Healthcare · Dental Clinic
Behind the Scene AI Implementation SaaS: Bridging Business Needs With Techmical Logic
A full phase-by-phase breakdown of how an AI appointment system was built for a dental clinic — from post-kickoff alignment to Go-Live, including the Google Calendar guardrail architecture that prevents double-booking.
Read Full Article ↗
📝
Coming Soon
Next Case Study — In Progress
A new breakdown is in the works, covering another industry vertical and the unique implementation challenges encountered along the way.
Coming Soon
Client Portfolio
Industries & Clients Served
🏥
Healthcare &
Aesthetics
AFK Beauty One Icon Dental
Medical guardrails, appointment scheduling (1 Calendar = 1 Doctor), no-diagnosis rules, 5-min escalation for red flags
👗
Retail &
Fashion
Mel Store 707 Store Brand ON
Shopify catalog sync, defect complaint flows, size exchange & PO logic, tracking number handling
🍜
F&B &
Hospitality
Bakmi Tiga Mangkok Pulas Villa Catering Fatimah
Dine-in reservations, bulk catering T&Cs, loyalty programs, property survey scheduling
🏢
B2B
Services
Lotus Group RUMIA Solsica Music
Tiered pricing & wholesale discount logic, CRM funnel flows, MOQ (Minimum Order Qty) rules
Operational Insight
Key Challenges & Solutions Applied
Real problems encountered during implementation — and how I resolved them systematically
Challenge 01
Data Ambiguity After Kickoff
⚠ Challenge
Ambiguity during requirement gathering — risk of re-requesting data that clients already shared with AM or PE at kickoff (via private chat, links, or files), causing unnecessary friction and eroding client trust.
✓ Solution Applied
Always review AI meeting transcriptions from kickoff sessions to map exactly what was discussed and what data was already shared — then align with AM/PE before requesting anything from the client again.
Challenge 02
Unclear AI Reference Chat Scripts
⚠ Challenge
Clients providing vague roleplay scripts instead of absolute logic, which causes the possibility of AI agents hallucinating prices or policies during live conversations.
✓ Solution Applied
Encourage clients to fill out a structured "Trigger & Reaction Matrix" — If condition A is met → Give Answer B → Else, Answer C — converting narrative scripts into deterministic logic the AI can execute precisely.
Challenge 03
Client Slow Response or No Response After Follow-Up
⚠ Challenge
Clients stalling the timeline by taking too long to provide required data, causing the data gathering phase to exceed its scheduled window and delaying downstream configuration.
✓ Solution Applied
Apply Anchoring (provide prefilled examples so clients only fill in the blanks), Feature Selling (explain why the data directly benefits their business), and A/B Closed Questions to reduce cognitive load and accelerate decision-making.
Technical Stack
Tools & Skills Applied
WABA / WAHA Boolean Logic Matrix If-Else Decision Trees Knowledge Base Config AI Guardrails Design Shopify Catalog Sync API / Webhook ERP Integration Google Calendar Arch. 5-Level Escalation Tree Halo Pulse Dashboard Trigger & Reaction Matrix