Home/Melverick Ng

Singapore · Agentic AI · Ex-SAP enterprise process lens

Digital Coworkers for ERP, CRM and SME operations.

Build AI systems that execute business processes safely and measurably.

I help growth-stage and enterprise-minded SMEs move from scattered AI experiments to human-controlled operating capability — with workflow design, approval gates, performance telemetry, and practical training for the people who own the work.

Melverick Ng
Operator-led / Agent-assisted / Audit-ready

Current work

Co-founder, Nexius Labs · Master Trainer, Nexius Academy

Process background

Former SAP consultant with ERP, CRM, operations and supply-chain workflow experience

Core principle

Orchestrate. Don’t operate. Humans own judgment; systems carry repeatable execution.

Ways to work together

Turn AI from a pilot into an operating layer.

Each engagement starts from the real workflow: where work begins, where it gets stuck, who approves it, and what evidence is needed before an AI system should act.

01

Digital Coworkers for ERP and CRM workflows.

Agentic assistants for lead intake, qualification, routing, follow-up orchestration, order flow, exception routing, and cross-functional handoffs.

Explore services →
02

AI transformation grounded in enterprise work.

SAP consulting experience gives me a process view of demand, planning, procurement, inventory, fulfilment, finance, reporting, and management control.

Read profile →
04

Governance, telemetry and human control.

Design approval gates, audit trails, role boundaries, task boards, and performance measures so automation stays useful, safe, and inspectable.

See projects →
✓

CRM lifecycle automation

Lead intake, qualification, routing, follow-up, account handoff, and revenue workflow coordination.

✓

ERP-linked process execution

Order flow, exception routing, inventory/logistics handoffs, finance visibility, and control points.

✓

Cross-functional coordination

Design workflows that connect sales, operations, finance, management reporting, and human decision gates.

✓

Capability transfer

Train business users and domain experts to operate with AI, not simply experiment with prompts.

The operating ecosystem

Labs, Academy, partners and practitioner work.

This site groups the work around the initiatives and collaborators that make execution possible.

Melverick Ng

Founder profile

Melverick Ng

Strategy, transformation, agentic workflows and practical AI implementation.

LinkedIn →

Nexius Labs

AI business automation

Digital Coworkers, workflow automation, Mission Control, and SME AI operating systems.

Visit Labs →

Nexius Academy

Practical AI training

Hands-on agentic AI programmes for non-technical business professionals and SME teams.

View courses →

Partner network

Business operators

Work with founders, finance teams, sales teams, operations leaders, and domain experts who own execution.

See Darryl →
Featured notes

Thinking behind the practice.

Three latest field notes. The full set is on the field notes page.

Field note · 7 November 2026

Your ERP Already Knows Who Is Slipping. Does Your Seller?

SugarAI is pushing CRM from record-keeping to next best sales action by reading ERP order history. Here is the SME habit I would copy: put purchase signals on the account, prompt the next human move, keep quote and order release with a person.

Read field note →

Field note · 6 November 2026

Put Your Finance Agent Next to the Invoice

Sage wants finance agents to sit inside AP and close work, not in a separate chat. Here is how I would design the same habit in an SME ERP stack: exceptions in, people approve, audit trail out.

Read field note →

Field note · 5 November 2026

Drafted, Proposed or Done? Know Which One Your Agent Just Did

As AI agents start working inside ERP and CRM systems, the most useful control is simple: every agent output should say plainly whether it is a draft, a proposed change, or something that already happened.

Read field note →

Field note · 3 November 2026

An Approval Button Is Not a Control

Putting a human sign-off in front of an AI agent only protects you if the reviewer can see enough, has the authority to say no, and only gets asked about the steps that matter. A practical guide for SME operators.

Read field note →

Field note · 1 November 2026

Your AI Workflow Does Not Need an Agent at Every Step

A practical guide for SME operators deciding which workflow steps need an AI agent, which need a smaller model, and which should remain deterministic.

Read field note →

Field note · 19 October 2026

Your AI Agents Are Not Saving Time If They Create a Human Attention Tax

A practical attention-budget framework for SME leaders managing AI agents: what they should complete, summarize, escalate, and hold for approval.

Read field note →

All field notes →

Open practitioner resources

Practical guides for building and operating AI agents.

Public, reusable field guides for efficient agent work and for designing an AI-native company with clear architecture, model routing, governance, and human control.

Public GitHub GistOperating guideEdition 1.0

Building an AI-Native Company with Buzz and OpenClaw

A practical guide to the architecture, providers, model routes, governance controls, and operating models needed to move from a solo AI setup to a resilient company-wide system.

  • Clear boundaries between workspace, runtime, providers, and business systems
  • Role-based model policies with credential rotation and resilient fallbacks
  • Least-privilege identities, secret handling, approval gates, and audit trails
  • Staged operating models for founders, small teams, and mid-size companies
Public GitHub Gist6 filesCC BY 4.0

Token-Efficient Agent Instructions

Use one lean core instruction file, load detailed procedures only when the task calls for them, make the narrowest safe modification, and verify in proportion to impact.

  • Progressive disclosure instead of loading every rule
  • Fingerprint-validated extraction and cache reuse
  • Narrow changes at the canonical source
  • Quality-gated verification with explicit evidence
Selected projects

Execution systems, training and operating infrastructure.

Representative work themes that connect strategy to daily operations.

AI workforce

Digital Coworker operating models

Role-based AI workers with goals, tools, task queues, approval paths, and audit-ready run history.

Meet the AI Workforce →

Mission Control

Human-controlled agent operations

Dashboards and workflows for supervising agents, decisions, tasks, telemetry, and evidence.

Academy

Agentic AI capability transfer

Training business professionals to design no-code workflows, govern agent activity, and improve real processes.

Tell me what you are trying to move

Map the workflow. Then build the Digital Coworker.

If your team is ready to operationalise AI around ERP, CRM, finance, operations, sales, training, or management workflows, start with the process and the control points.

WhatsApp Melverick