Small Business Optimization

It needs the software, people, and processes it already has to behave like one system.

I start with how the work actually gets done: where information comes in, who touches it, what gets copied between systems, what lives in somebody's head, and what nobody can see until something goes wrong.

Then we decide what should be integrated, automated, made deterministic, handled by AI, or left to human judgment.

That usually covers:

• Systems & Data: Connect the CRM, accounting, project management, marketing, and files the business already runs on.
• Workflow Automation: Remove repetitive entry, follow-ups, handoffs, reporting, and scheduling a system can handle.
• AI Integration: Find the workflows where AI can research, draft, classify, summarize, or extract.
• AI Training: Teach the team to use AI consistently instead of starting from a blank chat window every time.
• Controls & SOPs: Put repeatable rules into workflows, permissions, templates, and checks instead of institutional memory.
• Business Visibility: Give the owner a useful view of the business without assembling it from several tools.
• Tool & Cost Cleanup: Drop redundant software and expensive tools solving problems the existing stack already handles.

AI is useful after the business makes sense.

If customer data is in one system, project status is in another, financials live somewhere else, and the real process sits in somebody's head, AI does not fix that. It gives you another place to ask questions.

Get the systems and workflows under control first. Then AI works against information you trust.

Give AI a specific job.

"Use AI more" is not a workflow. A useful implementation starts with a job: summarize incoming requests, draft a client update, classify documents, extract information, or prepare a first-pass analysis.

Then define what information it can use, what output is expected, and what needs to be checked.

What that looks like in practice

• Turn emailed or uploaded documents into structured information
• Draft recurring customer updates from trusted source data
• Classify and route incoming requests
• Generate meeting prep from CRM and project context
• Build repeatable reporting without manual copy and paste
• Replace ad hoc prompting with reusable workflows

How It Works

1. Follow the Work: Map how work actually moves, including the inboxes, spreadsheets, duplicate entry, and workarounds.
2. Decide What Belongs Where: Sort it into human judgment, deterministic rules, ordinary automation, and AI. Do not use AI where a simple system does better.
3. Build the System: Connect the tools, automate the repetitive work, add controls and visibility, and give AI a specific job.
4. Train the People: Teach the team how the system works and when a human still needs to make the call.

The team has to know how to use it.

Buying an AI tool is easy. Getting a team to use it consistently on real work is harder.

Training gets built around the jobs people already do: what to give the model, how to reuse context, how to check output, what to keep out of it, and when a workflow beats another prompt.

Keep the important rules outside the model.

If something matters every time, do not depend on a prompt to remember it. Permissions, approvals, calculations, routing, deadlines, and alerts belong in the systems around the model.

AI can help with judgment. The operating system should still enforce the rules.

Get a handle on the business.

If too much work still lives in inboxes, spreadsheets, disconnected tools, repeated prompts, or somebody's memory, that is a good place to start.