01 / THE CHALLENGE
Understanding the problem.
Clients needed AI workflows connected to their existing business processes and systems. The work combined understanding the process, connecting the required services and refining how the workflow behaved in use.
- 01Client process
- 02APIs + controlled AI workflow
- 03Integrated deployment
02 / MY INVESTIGATION
Following the evidence.
I translated client business processes into automated workflows, then worked directly with clients to refine the behaviour.
Integration work spanned Xero, Google Workspace, Microsoft Graph/Microsoft 365, SprintSuite, XPlan, Monday.com and Zernio. The relevant integration depended on the client's process; these were not all part of one deployment.
03 / IMPLEMENTATION
Putting the work into practice.
- 01
Integrated APIs and Model Context Protocol servers to connect business systems and productivity platforms.
- 02
Configured OpenAI workflows with a Kimi fallback and Gemini image processing, including workflow guardrails and safeguards.
- 03
Integrated Firecrawl for web research and information gathering in financial advisory workflows.
- 04
Iterated directly with clients to refine deployments around their working requirements.
04 / OUTCOME & EVIDENCE
What the work delivered.
Delivered AI deployments for 5–6 clients, integrating business systems and refining workflow behaviour directly with clients.
Client delivery
AI deployments delivered for 5–6 clients.
Integration breadth
Connected business, productivity and research systems through APIs and MCP servers.
Workflow controls
Configured fallback behaviour and safeguards alongside the integrations.
05 / ENGINEERING PERSPECTIVE
What I take forward.
Useful AI delivery starts with the work someone needs to complete. The integrations, safeguards and client feedback are as important as the model configuration.
Organisation names are generalised. These accounts describe my work without publishing client systems, internal logs or proprietary source code.