Built in Copilot Studio, Claude or ChatGPT. My job is the part after the demo: getting the agent past security review and keeping the bill where finance expects it. I'm Abdulrahman Ibrahim Tudu, in Warsaw, working remotely.
One thing people ask me every week: an Azure budget does not cap Copilot Studio spend, it only sends alerts. The hard stops are in the Power Platform admin center. I wrote down where they are.
Tools: Copilot Studio, Claude and Claude Code, ChatGPT and Codex, MCP, Power Automate, Power Apps, Dataverse, Power BI, Azure AI, PAC CLI.
A fixed-price review of your use cases against Copilot, Claude and ChatGPT: which tool fits which job, and what each option will cost under your licences and data rules. You get a written recommendation you can take to procurement.
Who may build agents, where their data may go, and how an agent gets from a maker's sandbox to production. I write the policy, set up the environments, DLP and author groups, and run the same exercise for Claude Code and Codex when developers want them.
Agents and flows delivered as managed solutions, tested, with a runbook your team can follow without me.
After the assessment: a build project, an advisory retainer, or a workshop for your admins.
Client names are left out. Numbers follow once the clients agree to them.
Supplier PDFs arrived in a finance mailbox and had to reach SAP's invoice process as structured e-invoices. I built a queued pipeline: each PDF becomes a tracked record, Azure AI extracts the fields, the flow generates and validates PEPPOL UBL XML and hands it to the SAP mailbox. It shipped as a versioned managed solution with a UAT guide and an import runbook that matches the running system.
Business units in several countries wanted to build agents. Central IT needed rules for who builds, who approves and who owns what. The result was one authoritative guide plus short role guides for makers, admins and business owners, with a plan for publishing each to its audience.
A central admin team bills each business unit for its Power Platform and Copilot Studio usage. Azure budgets only alert, so the limits had to live elsewhere. I designed a four-stage control model (Azure, billing plan, environment overage, per-agent limits) and a Power BI chargeback model fed by Cost Management exports and the licensing API, generated as code and refreshed in a test tenant.
Several agents on Claude and OpenAI models do my own operations work. One administers Power Platform through the PAC CLI, Dataverse and MCP servers, under the identity and audit rules I would expect a client to enforce. I use what I recommend.
A school-management platform for Nigerian schools that I founded and built: admissions, results, fees, parent access and computer-based tests. It is live at akadaa.com.
I start by reading the tenant, the licences and the bill. What is true today goes in writing before anything is designed. The design itself is one document: the decision, the options I rejected and why, and the cost. Then I build in your environments as managed solutions and test against real cases. At hand-over your team gets a runbook and a walkthrough, and owns it from then on.
The AI in AI Tudu is my initials: Abdulrahman Ibrahim Tudu.
I'm a Power Platform solution architect and administrator, and I administer and use Claude, ChatGPT, Claude Code and Codex every day. I grew up in Gusau in northern Nigeria, trained as a scientist (MSc Applied Biotechnology), and have lived in Warsaw since 2020, working with several clients at a time.
Outside client work I build Akadaa, a school-management platform for Nigerian schools, and I co-founded the Northspire Education Foundation, which supports students in northern Nigeria.
Tell me what you're building and what worries you about it. I reply within two working days with a few questions, or a short call if that's quicker.