Chukwuka Fredrick UzohoNigeria · Working with teams globally
Delivery experience across
100+
businesses supported across consulting, digital operations, systems and business enablement
Understands business problems like a consultant. Designs solutions like an architect. Builds systems like an engineer. Thinks about AI like a product builder.
My work spans the boundaries that often separate business analysis, systems architecture, implementation, and applied AI — which is why the design, the integration and the intelligence layer stay consistent with one another.
The through-line
Operations → Systems → Automation → Intelligence. Every section of this site is a step along it.
01The operating model
AI belongs on top of a system, not in place of one
Five layers. Each one is the reason the next one works — you cannot automate a process you have not mapped, and you cannot apply AI to data no system was ever designed to produce.
Layer 01Operations
People, roles, handoffs, exceptions — the operational reality, including the work that never made it into the manual.
Layer 02Process & requirements
Process mapping, requirements, business rules, ownership and governance. Written down before a tool is chosen.
Layer 03Systems & integration
ERP, CRM, databases and line-of-business applications connected through APIs into one source of truth.
Layer 04Automation
Workflow engines, webhooks, scheduled processes and business rules that carry work forward without anyone chasing it.
Layer 05Intelligence
AI assistants and agents, RAG, machine learning, prediction and decision support — grounded in the four layers below.
Most failed AI projects are really layer-two failures. The model is asked to compensate for a process nobody designed and data no system was built to produce.
02Selected capabilities
Four disciplines that have to work together
01 · Business systems architecture
Design the system before buying the software
Process mapping, requirements engineering, workflow design, CRM and ERP process architecture, system integration and operational improvement.
SAPZohoHubSpotMonday.comClickUpAirtable
02 · Automation engineering
Make the workflow carry itself
API integrations, workflow automation, webhooks and event-driven handoffs, business process automation across disconnected tools.
PythonFastAPIn8nMakeZapierREST APIs
03 · Operational AI agents
Agents that act inside business systems
Agents that read operational context, call business tools, follow rules, request approvals and take controlled action — with RAG for policy and procedure, not for numbers.
Embeddings and vector similarity, semantic search, NLP concepts and data processing pipelines — the difference between retrieving the right answer and sounding confident while wrong.
PythonPyTorchScikit-learnSentence Transformers
03Featured work
Four pieces of work that show the whole stack
Each one carries a visible maturity label. Nothing here is presented as deployed enterprise software unless it is.
The same discipline at two very different scales — inside global operators running SAP, and alongside founders and SMEs building their first real system.
100+businesses supported across consulting, digital operations, systems and business enablement
Since 2017working in business process, solutions architecture and enablement
Shell & Totalenterprise delivery inside global energy operators
Jul 2024 — PresentTotalEnergiesNigeria · On-site
SAP Enterprise Asset Management Consultant
Partners with maintenance, operations and business users to improve SAP-enabled asset management processes — work orders, notifications, functional locations and the master data that decides whether an operation can be measured at all.
Jan 2021 — PresentVirgocreativoBusiness systems & AI solutions
Solutions Architect
Leads end-to-end business systems design for founders, SMEs and organizations navigating growth — discovery and process mapping through configuration, integration, automation and handover.
Oct 2023 — Oct 2024ShellBusiness process & governance
Senior Business Process & Solutions Architect
Coordinated reporting, governance and follow-up workflows — designing how information moved and who acted on it, rather than reporting on it after the fact.
Where results are not published or measured, the outcome is stated qualitatively rather than invented.
Client project
Connected business operations
A growing operation running sales, delivery and finance in separate tools. Each function held part of the truth, so answering a simple question about a job meant asking three people and trusting the most recent spreadsheet.
No appetite to replace the existing tools, and no team to run a migration. The redesign had to work with what was already bought and paid for.
One record as the system of truth, with each tool writing to it rather than around it. Process mapped end to end first, ownership and status transitions defined, then integration points chosen to match the process rather than the other way round.
Workflow rebuilt around the single record, tools connected through APIs and webhooks, status changes made automatic where the rule was unambiguous and left manual where judgement was required.
One pipeline everyone can see, status that updates itself as work moves, and reporting that comes out of the operation rather than being reconstructed from it afterwards.
Technical prototype
AI lead qualification with voice intake
Enquiries arriving outside working hours and being qualified by whoever was free — inconsistently, and often late enough that the lead had already moved on.
Qualification criteria had to stay defensible and identical every time, and nothing could be written into the CRM that a human had not been able to review.
Voice intake into an intent and qualification step, tool calls to look up and validate against existing records, and a structured lead handed into the CRM workflow with a follow-up attached.
Built with voice AI APIs, a Python service for reasoning and tool calling, and CRM integration for record creation. Every call ends as structured data rather than a recording someone has to listen to.
Every enquiry is answered and qualified the same way, with a follow-up already assigned before anyone opens a laptop. Running as a prototype — not deployed as a production customer-facing system.
Product development
Jobkard
Service businesses running real operations across paper, WhatsApp and spreadsheets, with no reliable job history and no clear view of what is owed against which job.
The people using it are on site, often on a phone, often in a hurry. If capturing the data costs more time than it saves, it will not be captured at all.
One job record carrying its own history through the lifecycle, so structured data is produced as a side effect of doing the work rather than as an extra admin task.
Job management platform in active development, with an AI roadmap deliberately held back until the operational data underneath it is trustworthy enough to reason over.
In development. The current target is a dependable job management platform; the longer-term vision is a connected operating system for service operations.
06The Process Virtuoso
Writing on process, systems and AI readiness
Arguments for operators and leaders deciding where technology actually fits.
AI readinessAI Won't Fix Your Broken Processes
Automation applied to a bad process produces bad outcomes faster. What has to be true before the model is worth adding.
DataYour AI Agent Is Only As Good As Your Data
Retrieval quality is a data quality problem wearing a different hat. Where the failures actually originate.
Systems thinkingFive Tools. Still No Business System.
Buying software is not designing a system. The difference between a stack and an operating model.
Operational designWhy Businesses Need Operating Systems
The case for one connected record of how work moves, and what it changes about how a business is run.
AI agentsThe Future of Intelligent Workflows
What changes when the workflow can reason — and which parts of the process should stay deliberately human.
