- The Next Input by Cylentis AI
- Posts
- š® The Next Input ā Issue #103
š® The Next Input ā Issue #103
Stop Trying to Fix Your ERP

ā” The Briefing ā 60 sec
DoorDash rolls out āZesty,ā an AI social app for discovering restaurants
Weāre all for AI here at The Next Input but⦠seriously? Feels like one of those ideas that either slaps or disappears quietly in six months. Hope itās a banger.Google tests an email-based productivity assistant
This, though? We can get behind. If AI can clean up email without inventing new work, thatās a real win.ANZ executives shift focus from ERP upgrades to agentic AI
ERP what? Sorry NetSuiteālooks like youāre riding shotgun while agents take the wheel.
š ļø The Playbook ā The Agentic Ops Layer (Goodbye ERP Drag)
MissionāReplace slow, brittle ERP workflows with task-specific AI agents that actually get work done.
DifficultyāAdvanced
Build timeā3 hours
ROIāCuts operational friction fast without multi-year ERP projects.
0) Why This Matters
Most ERP systems are great at one thing: slowing teams down.
The shift ANZ is making signals a broader moveāorganisations donāt want massive upgrades anymore. They want agents that act, not systems that just store data.
This playbook shows how to layer agents around your ERP instead of waiting years to replace it.
1) Architecture
Component | Tool | Purpose |
|---|---|---|
Task Intake | Email / Form / Slack | Capture real operational requests |
Orchestrator | LangGraph | Break requests into agent tasks |
Decision Engine | GPT-5-mini | Decide next actions and routing |
Action Agents | Claude 4.5 Sonnet | Execute domain-specific steps |
System Bridge | APIs / RPA | Read/write to ERP or legacy tools |
Audit Log | Postgres / BigQuery | Track every action and outcome |
2) Workflow
A request comes in: invoice issue, vendor update, approval request.
LangGraph decomposes the request into steps.
GPT-5-mini decides which agents are required (finance, ops, compliance).
Claude 4.5 Sonnet agents execute actions:
fetch records
validate data
draft responses
prepare updates
System writes changes back into ERP or connected tools.
Human approves final action when required.
3) Example Prompts
Task Routing (GPT-5-mini)
Given this request, decide:
- required sub-tasks
- agent types needed
- which steps require human approval
Return an ordered task list.
Execution Agent (Claude 4.5 Sonnet)
You are an operations agent.
Complete the assigned task using available system data.
Do not assume missing information.
Prepare outputs for human approval where required.
4) Guardrails
No agent writes directly to financial systems without approval.
Log every action with before/after state.
Keep agents narrowly scopedāno ādo everythingā bots.
Fallback to human review on ambiguity.
5) Pilot Rollout ā 3 hours
Identify one painful ERP workflow (approvals, reconciliations, updates).
Map steps manually once.
Build agents for each step.
Test with real but low-risk data.
Add approval checkpoints.
Measure speed and error reduction.
6) Metrics
Time to complete operational tasks
Number of manual handoffs removed
Error rate vs ERP-only workflow
Agent actions per day
Human approval turnaround time
Pro Tip: Donāt replace the ERP. Starve it. Let agents do the thinking while ERP stays as the record system.
šÆ The Arsenal ā Tools & Platforms
LangGraph Ā· Multi-agent orchestration and task flows Ā· https://langgraph.dev
Temporal Ā· Durable workflows for long-running ops tasks Ā· https://temporal.io
n8n Ā· Lightweight automation and system glue Ā· https://n8n.io
UI Bakery Ā· Fast internal approval interfaces Ā· https://uibakery.io
Copy-paste prompt block:
You are an operational agent.
Execute one task only.
Log inputs, outputs, and uncertainties.
Escalate for approval when required.
š” Free Office Hours
Want help implementing anything? Book a free 15-minute Office Hours slotāno sales pitch, just workflows solved.
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
š¹ļø Game Over
ERP keeps the books. Agents keep the business moving.
ā Aaron Automating the boring. Amplifying the brilliant.
Subscribe: https://cylentisai.beehiiv.com/subscribe

