AI-assisted workflows
Use LLMs for extraction, classification, summarization, drafting or structured interpretation inside an existing process.
AI Automation for Business Operations
Automate the workflow, not the chaos.
I help businesses redesign operational workflows, connect their systems and introduce AI where it creates measurable operational leverage — not simply because AI can be added.
AI is not the first step. A working operation is.
The operational problem
AI cannot repair a workflow that nobody has defined.
Before adding an agent or automation, I identify where information originates, how decisions are made, which rules are deterministic, where exceptions occur and which actions still need human oversight.
Manual information handlingTeams repeatedly read, classify, extract, summarize or route the same kinds of information.
Disconnected systemsAutomation breaks because the data and actions live across tools that do not share a reliable workflow.
Unclear decision pointsAn AI agent cannot safely act when the business has not defined what should happen under different conditions.
Automation without controlA fast workflow is not useful if nobody can see what happened, intervene in exceptions or audit important decisions.
What I build
AI that is part of the system — not a demo beside it.
I combine APIs, workflow orchestration, custom logic and LLM capabilities so AI can operate inside a controlled business process.
Operational agents
Connect agents to tools, APIs and business rules so they can support real actions rather than only answer questions.
Document intelligence
Capture information from email, forms, PDFs and documents and route it to the right workflow.
Workflow orchestration
Coordinate deterministic steps with n8n, Make, Apps Script, APIs, webhooks and custom code.
Human-in-the-loop systems
Keep review and approval at the points where judgment, risk or accountability still belongs with a person.
AI-ready operations
Restructure the underlying workflow and data so future AI capabilities can be added without increasing operational fragility.
When I’m usually brought in
You know AI could help. The harder question is where it belongs.
I am most useful when the business does not just need an AI feature. It needs the workflow, data, integrations and decision boundaries designed so AI can operate reliably inside the process.
People spend hours interpreting information.
Email, PDFs, forms or notes repeatedly need to be read, classified, summarized or turned into structured data.
Automation exists, but it is brittle.
Several tools are connected, yet exceptions and unclear business rules still force people to repair the workflow manually.
An agent needs real operational context.
The goal is not a chatbot. The agent needs access to the right systems, rules, permissions and human review points.
Operations-first architecture
Workflow → System → Automation → AI
I do not begin by choosing a tool. I first understand what the operation needs to control, what can be deterministic, what needs judgment and where technology can remove friction without creating another fragile layer.
01Workflow
Map how work actually moves, including rules, exceptions and decisions.
02System
Create the right structure for data, interfaces, permissions and business logic.
03Automation
Connect systems and remove repetitive work that should not require manual effort.
04AI
Add intelligence where classification, extraction, reasoning or assistance creates real leverage.
Proof in practice
AI embedded in the operation, not added beside it.
Selected system · AI operations
Operations Command Center
The system combines operational intake, recurring work, documents and financial tasks with AI-assisted actions. AI is introduced inside a controlled workflow rather than used as a separate conversational layer.
Result: a clearer operating environment where automation and AI support the work without hiding the underlying process.
Questions
Before building another tool.
What business processes are good candidates for AI automation?
Processes with repeated information handling, clear outcomes and identifiable decision points are strong candidates. The workflow still needs to be mapped before selecting the AI component.
Do you build AI agents?
Yes, when an agent is the appropriate architecture. In other cases a deterministic workflow with one or two AI steps is simpler, safer and easier to maintain.
Do I need to replace my current software?
Usually not. I prefer to connect useful systems through APIs and automation before recommending replacement.
Which tools do you use?
Depending on the workflow: custom JavaScript or Python, Google Apps Script, REST APIs, webhooks, n8n, Make, LLMs and databases such as Firebase or Firestore.
How do you decide where a human should stay involved?
I look at business risk, ambiguity, reversibility and accountability. High-impact exceptions or uncertain decisions can remain reviewable instead of being fully autonomous.
Start with the operation
I don’t start with the tool.
I start with the workflow.
Tell me where the operation is becoming difficult to control. You do not need a technical specification — the current workflow is enough to start.
Discuss your operation →