AK·Let’s talk
Building Agentic AI Systems

Abubakar KhalidAgentic AI Solutions Architect

I Build AI SystemsThat AutomateReal Business Work.

AI agents, ecommerce automation, WhatsApp systems and intelligent workflows built around your existing tools.

View My Systems Discuss Your Project
Based in Pakistan LinkedIn
AGENT RUNTIMEARCHITECTURE / 01
AI AGENTReason. Decide. Act.
01 INPUT02 REASONING03 TOOL04 ACTION
WhatsAppCustomer input
AI AGENTReasoning & tool selection
ShopifyDatabaseAPIs

INPUT → REASONING → TOOL → ACTION

INTELLIGENCE, CONNECTED TO ACTION.EXPLORE THE SYSTEMS
THE BUILDING BLOCKSAgentic AIn8nShopifyPostgreSQLLLMsRobotics
01 / BUILT AROUND YOUR BUSINESS

What I Can Build
for Your Business.

Bring the workflow. I’ll help define the agents, integrations, and controls it needs.

AI Customer Support Agent

An AI support system that understands customer questions, retrieves business information, checks orders and escalates complex cases.

CUSTOMER EXPERIENCE

WhatsApp AI Automation

Connect WhatsApp with AI, databases, APIs and workflows for customer service, lead handling and operations.

CONVERSATIONS → WORKFLOWS

Ecommerce AI Automation

Automate customer support, order workflows, approvals, notifications and ecommerce operations.

CONNECTED COMMERCE

Custom AI Agents

AI agents that reason about requests, use tools and perform controlled business actions.

REASONING + ACTION

n8n Workflow Automation

Automate repetitive processes across apps, APIs, databases and communication platforms.

LESS MANUAL HANDOFF

AI + API Integration

Connect AI models with Shopify, databases, CRM systems, APIs, Google tools and other business software.

YOUR EXISTING TOOLS

Human-in-the-Loop AI

AI automation with approval checkpoints for actions that should remain under human control.

CONTROL BY DESIGN
02 / FROM MANUAL TO CONNECTED

Problems I Help Automate.

Practical starting points for a system designed around your operations.

THE MANUAL WORKTHE AUTOMATION APPROACH

Your staff answers the same customer questions repeatedly.

A support agent retrieves your business knowledge and passes complex cases to your team.

Business knowledge + escalation

Customers keep asking for order status manually.

Connect AI to your ecommerce order system so it can retrieve the relevant information.

Shopify + AI tools

Important actions need a person’s approval.

Add validation rules and approval checkpoints before the system performs sensitive actions.

Guardrails + human approval

Your business uses several disconnected tools.

Connect the handoffs between apps, data and messages in one coordinated workflow.

n8n + APIs + orchestration

Customer conversations lose context.

Design database-backed memory so the system can retrieve relevant previous interactions.

Persistent customer context
Abubakar Khalid — Agentic AI Solutions ArchitectABUBAKAR KHALID / PAKISTAN
03 / THE PERSON BEHIND THE SYSTEMS

Technical curiosity.
Practical execution.

I’m Abubakar Khalid, an Agentic AI Solutions Architect and Robotics & Intelligent Systems student focused on building AI systems that connect reasoning with real business tools.

My work focuses on AI agents, n8n automation, ecommerce systems, APIs, databases, persistent memory and human-in-the-loop workflows.

BUILD

Turn ideas into working systems.

CONNECT

Connect AI with APIs, databases and business platforms.

AUTOMATE

Reduce repetitive manual operations through intelligent workflows.

BS Robotics and Intelligence SystemsBahria University Islamabad
04 / SYSTEMS IN PRACTICE

Built to connect.
Designed to do real work.

Explore the architecture, tools, and decision logic behind my work. Every system has a clear development status.

Showing completed work, work in development, concept demos, and explorations.

SYSTEM / V1Completed

AI Ecommerce Customer Support System

Multi-tool AI support architecture for ecommerce operations.

Designed a multi-tool ecommerce AI architecture combining customer messaging, LLM reasoning, Shopify data retrieval and automated escalation.

  • Understand requests and select the appropriate business tool.
  • Retrieve store policies and Shopify order information.
  • Return structured results and route replies or staff escalations.
  • Communicate through WhatsApp.
n8nLLM / AI AgentShopify APIEvolution APIWhatsAppGoogle SheetsStructured Output ParserHTTP APIsSub-workflows
Explore the case study
V1 / TOOL ORCHESTRATIONInteractive illustration
“What is your returns policy?”
CustomerWhatsAppEvolution API
AI OrchestratorLLM reasoning · n8n workflow
TOOL RESULTGoogle Sheets / store policies

Retrieved policy information is used to prepare a customer reply.

STRUCTURED OUTPUT → CONTROLLED ROUTING
status
reply
reason
Store-policy information requested
priority
standard
customer_reply
Answer or request for missing information
AI response → Customer on WhatsApp

Illustrative requests and routing labels. No live messages or store data are used.

SYSTEM / V2In development

Agentic Ecommerce Operations System

AI orchestration, persistent memory, order actions, guardrails and human approval.

Developing an operations architecture with specialized agents, persistent customer context, and approval boundaries for order actions.

  • Coordinate requests through a main agent and order agent.
  • Design persistent customer memory in PostgreSQL.
  • Evaluate order actions with guardrails and staff approval.
n8nPostgreSQLShopifyEvolution APIWhatsAppRedisDockerLLMsAPIsHuman-in-the-loop workflows

In development. Order actions, memory, guardrails, and approval flows shown here are design targets, not a finished production system.

Explore the case study
V2 / CONTROLLED OPERATIONSDesign simulation
Customer request: “Cancel my order.”
Main AI AgentIntent detection
Order AgentShopify
Guardrail checkHigher-risk action → approval required
STAFF WHATSAPP APPROVALAwaiting human decision
Action paused until approval

In development. This local simulation does not perform order actions, store customer data, or send notifications.

FROM ARCHITECTURE TO YOUR WORKFLOW

Have a similar workflow in your business?

05 / CONNECTED INTELLIGENCE

More Than a Chatbot.

I focus on building systems that connect AI reasoning with real operational tools.

SIMPLE QUESTION–ANSWER FLOWUser asks a questionAI replies

Useful conversation.
A limited operational role.

MY AGENTIC SYSTEM APPROACH
01User request
02Intent analysis
03Tool selection
04Business data
05Guardrails
06Action
07Memory
08Response
ShopifyPostgreSQLWhatsAppAPIsGoogle Sheetsn8n

Architecture adapted to the workflow. Memory, action controls, and approval are part of the V2 direction and concept designs.

06 / WHAT YOUR NEXT SYSTEM COULD LOOK LIKE

Business scenarios.
Working through the possibilities.

Interactive concept demos showing how the architecture could work. These are not client deployments.

01 / LEAD OPERATIONSCONCEPT DEMO

AI Lead Qualification & Appointment Agent

A concept for turning WhatsApp and website enquiries into qualified, recorded leads and proposed appointment slots.

POTENTIAL USE CASES

Dental clinics · Aesthetic clinics · Real estate · Consulting · Service businesses

WhatsAppGoogle CalendarGoogle Sheets / CRMEmailn8nLLMAPIs
WORKFLOW WALKTHROUGHIllustrative · not connected
New enquiryWhatsApp / website form
Proposed appointmentCustomer confirmation required
Lead received

Receive a WhatsApp message or website form enquiry.

Step 1 of 6

Concept demonstration, not a paid client project or production deployment. No live data or external actions.

02 / SALES CONTINUITYCONCEPT DEMO

AI Sales & Customer Follow-Up System

A concept for keeping lead context between conversations and coordinating timely WhatsApp or email follow-ups.

POTENTIAL USE CASES

Service enquiries · Sales pipelines · Customer follow-ups

n8nPostgreSQLWhatsAppEmailLLMCalendarCRM / API
WORKFLOW WALKTHROUGHIllustrative · not connected
One conversation. Continuing context.

Requirement → saved context → follow-up → customer reply

PostgreSQL memory · WhatsApp / email
New lead

A lead contacts the business without making an immediate decision.

Step 1 of 8

Concept demonstration, not a paid client project or production deployment. No live data or external actions.

03 / INTERNAL OPERATIONSCONCEPT DEMO

AI Business Operations Assistant

A multi-tool assistant concept that gathers operational information and prepares a useful briefing for a business owner.

POTENTIAL USE CASES

Ecommerce owners · Operations teams · Internal business support

ShopifyPostgreSQLGoogle SheetsEmailSlack / WhatsAppn8nLLMsAPIs
OPERATIONS BRIEFINGIllustrative · not connected

“Give me today’s orders, identify pending issues and prepare a summary.”

Order dataPotential data source
Business contextPotential data source
Support ticketsPotential data source
BRIEFING STRUCTURE
Order overview
Issues requiring review
Recommended next actions
Owner request

“Give me today’s orders, identify pending issues and prepare a summary.”

Step 1 of 8

Concept demonstration, not a paid client project or production deployment. No live data or external actions.

Prototypes & agentic experimentsExplore the learning behind the systems +
Prototype / Exploration

WhatsApp Multimodal
AI Assistant

Exploring how an assistant can interpret text, voice, images, and documents in a shared conversation. This prototype helped develop my understanding of multimodal processing, memory, RAG, and multi-user handling.

Explore the concept

WhatsApp → Evolution API → input routing → speech-to-text or document / image processing → LLM + knowledge retrieval → response. These are explored architecture concepts; the complete feature set is not claimed as finished.

Text
Voice
Image
Document
Multimodal intelligenceEvolution API · LLMs · Memory · RAG concepts

Small experiments. Deeper understanding.

AGENTIC EXPERIMENTS
01

Two-tool AI agent

+

Selects between Google Sheets for stored information and a calculator for arithmetic.

User request→ AI agent→ Sheets / Calculator→ Response
02

Structured output agent

+

Explores predictable JSON responses for workflows such as appointment handling.

Request→ AI agent→ JSON output→ Workflow routing
03

Tool calling agent

+

Chooses a tool based on the user’s intent and uses its result to formulate a response.

Intent→ Tool selection→ Tool result→ AI response
04

Multi-step agent

+

Combines subflows and Switch logic to coordinate a sequence of workflow steps.

Trigger→ Switch logic→ Sub-workflow→ Result
07 / TECHNOLOGIES & INTEGRATIONS

Technologies I Work With.

A practical toolkit for connected systems, alongside concepts I’m developing through hands-on work.

AI & agents

LLMsAI AgentsTool CallingStructured OutputsPrompt EngineeringAgent OrchestrationMulti-Agent ArchitecturesRAG concepts

Automation

n8nWebhooksSubflowsHTTP APIsWorkflow DesignRouting Logic

Commerce

Shopify

Communication

WhatsAppEvolution APIEmailGmailTelegramSendGridGoogle Calendar

Data

PostgreSQLRedisGoogle Sheets

Infrastructure

Docker

Development

PythonJSONGitVS CodeAPIs

Control & learning

Human-in-the-loop systemsRoboticsArtificial IntelligenceIntelligent Systems
08 / A DELIBERATE APPROACH

How I Build AI Automation.

Start with the business process. Design the intelligence and controls around it.

STEP 01

Understand the Workflow

Understand the current business process, repetitive tasks and existing software.

STEP 02

Design the AI Architecture

Decide what the AI should understand, which tools it can use and which actions require control.

STEP 03

Connect Business Systems

Integrate APIs, databases, ecommerce platforms and communication channels.

STEP 04

Add Guardrails

Protect important actions using validation rules and human approval.

STEP 05

Test Real Scenarios

Test normal requests, edge cases, ambiguous messages and failures.

STEP 06

Iterate & Improve

Improve prompts, routing, tools, memory and operational reliability.

The builder’s journey & long-term directionEducation, experiments, and what comes next +
06 / THE JOURNEY

Learning by building. Moving with intention.

2025

The foundation

Started BS Robotics and Intelligence Systems at Bahria University Islamabad.

2026

From workflows to tool-using agents

Focused on AI automation and n8n. Built tool-using agents, integrated WhatsApp through Evolution API, and connected Shopify to AI workflows.

2026

A complete support system

Built Ecommerce AI Customer Support V1, bringing policy retrieval, order lookup, and human escalation together.

NOW

Building deeper systems

Learning PostgreSQL, Docker, and persistent AI memory while developing V2 with order actions, guardrails, and human approval. Long-term direction: combining AI automation with robotics.

05 / WHAT I’M BUILDING NEXT

One ecosystem.
Many intelligent agents.

Concept / In development

AI Ecommerce Operating System

A larger multi-agent vision: specialized agents cooperating across ecommerce operations, coordinated by an orchestrator and governed by human approval.

Potential integrations include Shopify, Meta Ads, TikTok, databases, WhatsApp, email, Google Sheets, APIs, and LLMs.

Orchestrator
Customer supportOrder operationsProduct researchSupplierMarketingAnalyticsInventoryHuman approval
EXPLORING A COORDINATED OPERATING SYSTEM
07 / THE PHILOSOPHY
“I’m not interested in building AI that only talks. I want to build AI that understands, uses tools, makes controlled decisions and performs real work.”
ABUBAKAR KHALID
LET’S BUILD SOMETHING USEFUL

Have a Business Process
You Want to Automate?

Tell me what your team currently does manually. I’ll help explore how AI agents, workflow automation and integrations could turn it into a smarter system.

Discuss Your Project on WhatsAppConnect on LinkedIn
abubakarkhalid0300@gmail.com
Available for selected AI automation & agentic system projects.