Aadhib
ENع

Service / 01

AI agent development and integration

Aadhib Nasser · AI Engineer and AI Consultant

Most AI agents look finished in a demo and come apart in the first week they touch real systems. The work I do is the part after the demo: connecting an agent to the ERP, the procurement platform, the ticketing system or the internal database it needs, deciding what it is allowed to do on its own, and making every action it takes traceable.

DAMMAM, SAEnglish · Arabic
Service
AI agents & integrations
Based in
Dammam
Saudi Arabia
Clients in
Saudi Arabia, the UAE and Ireland
Engagements
Direct, under my own name

Overview

I have built the enterprise systems that agents now have to work with, including visitor and access platforms, maintenance management, internal IT support and a SAP Ariba cXML connector. The integration is the part I start from, not the part left for later.

Who this is for

  • 01Companies with an ERP, procurement platform, ticketing system or internal database an agent would need to work with
  • 02Teams with an agent that works in a demo but has not yet been trusted with real systems
  • 03Anyone who needs every action an agent takes to be traceable

What the work covers

  • 01Deciding which tasks an agent should take on, and which it should not
  • 02Tool and API design: the functions an agent calls, each with narrow permissions
  • 03Integration with existing systems, including ERP, procurement platforms such as SAP Ariba, ticketing and internal databases
  • 04Approval steps for consequential actions, enforced in the application rather than the prompt
  • 05Logging what the agent actually executed, not what it reports it did
  • 06Evaluation on your own tasks before launch and after every model change

How I approach it

Permissions first

An agent inherits your permissions model. Read-only by default, write access granted per tool and per target, and a human approval for anything with a financial or legal consequence.

Validate before you send

Enterprise protocols do not forgive. In the SAP Ariba connector, validation happens before transmission and every push is recorded, so a failure can be diagnosed after the fact rather than guessed at. Agents get the same treatment.

Run it where it can be watched

I run my own AI agents every day on a Mac Studio. What breaks, what is slow and what maintenance actually costs only shows up in daily use, not in a benchmark.

Related work

Case studies and projects

01Running AI agents locally for business workloads02Integrating SAP Ariba using strict cXML validation03Local AI & agent infrastructure04SAP Ariba cXML connector05Enterprise CMMS: maintenance management system

Further reading

Notes on this subject

01AI agents are easy to demo. Production agents are different.02Why enterprise integrations are harder than the API documentation suggests03cXML isn't REST: lessons from building a SAP Ariba integration04MCP went stateless: what the 2026-07-28 spec changes if you run agents

Questions

Asked before an engagement

Which models do you build agents on?

Whichever fits the task, the data and the budget: hosted frontier models, open-weight models on your own hardware, or a mix of the two. Models change every few weeks. The integration and the permissions model are what last, so that is where the design effort goes.

Can an agent work with our existing ERP or procurement system?

That is usually the actual job. I have built integrations against enterprise systems including SAP Ariba over cXML, and the approach carries over: learn the protocol's rules, validate before sending, and keep a record of every exchange.

Start a conversation

I usually reply within a day

Send a few sentences about the problem: what the process is today, where the data lives, and what would change if it worked.

WhatsApp +966 575 741 941hi@aadhib.aiContact form

Other services

All services →
02Local & on-premise AIPrivate models on hardware you control, and hybrid local and cloud designs.03AI strategy & advisoryWhere AI fits, which model and vendor, and what a pilot has to prove.04Arabic-first AI productsBilingual, right-to-left AI products, and AI that cites Arabic sources.