# AI services for small businesses

Start with the problem, not the technology.

I build machine learning and AI systems for small businesses that do not have their own AI or machine learning team. Based in Los Angeles. I work with clients remotely, anywhere.

## Who this is for

Owners and operations leads whose teams lose hours to repetitive work, and who have data they are not getting much out of. I take on small businesses only.

## What I fix, and how

### [Machine learning builds](https://lichucki.ai/services/machine-learning)

“We are guessing.” Sales forecasts come from gut feel, the team does not know which leads to call first, and problems are noticed after they have cost money.

Models that answer a specific question the business asks every week, built on the data you already keep. Each one ships with a simple baseline to beat, an honest measure of how good it is, and a way to roll it back.

A question it answers: Which accounts are most likely to buy this quarter, and how sure are we?

Delivered on Google’s tools (Vertex AI and BigQuery) or with open-source tools on your own hardware.

Why me for this: I built sales forecasting and lead and account scoring models at a real-estate software company. ([Work experience](https://lichucki.ai/work))

### [AI assistants and agents](https://lichucki.ai/services/ai-assistants-and-agents)

“My team answers the same questions and repeats the same manual steps every day.” Skilled people spend their time on routine work.

An assistant that uses your calendar, email, documents and systems to get routine work done. It confirms before it sends, spends or changes anything, and every turn is logged so it can be checked and improved.

A question it answers: Can something handle the routine requests and check with a person before it sends anything?

Delivered on Anthropic’s tools, on Google’s tools including Gemini Enterprise, or on open-source models hosted on your own hardware.

Why me for this: I built a private assistant with about 150 tools across 30 tool servers, confirmations before any send or purchase, and an automated quality review every day. ([Private assistant project](https://lichucki.ai/projects/private-assistant))

### [Chat-to-data and knowledge bases](https://lichucki.ai/services/chat-to-data)

“Nobody can get a number without waiting on the one person who knows the data, and two teams report two different figures for the same thing.”

A knowledge base that records what your tables mean, how they join and how you define your metrics, so an AI assistant writes the right query. Captured once, and kept correct as the business changes.

A question it answers: What was revenue last month, by our definition of revenue?

Delivered on Anthropic’s tools, on Google’s tools including Gemini Enterprise and BigQuery, or on open-source models hosted on your own hardware.

Why me for this: I built chat-to-data products and knowledge bases that ground AI, at a real-estate software company. ([Work experience](https://lichucki.ai/work), [Knowledge system project](https://lichucki.ai/projects/knowledge-system))

### [Data strategy and foundations](https://lichucki.ai/services/data-strategy)

“We know AI could help and don’t know where to start.” Data is scattered, no one agrees on the numbers, and money is about to be spent without a plan.

Deciding what to measure, setting up experiments that give a clear answer, and building the reporting people will use. Often the first step, and sometimes the only one a business needs.

A question it answers: Did the campaign work, or would those customers have bought anyway?

Usually delivered on Google’s data tools, BigQuery included.

Why me for this: I built a large-scale A/B testing framework and reporting from scratch, and left roadmaps for continued data work, for a major streaming service through a consultancy. ([Work experience](https://lichucki.ai/work))

## Choosing the right tooling

Managed tools from Anthropic or Google are the fastest way to start and need the least upkeep; you pay as you use them and your data is handled under the provider’s terms. Running open-source models on your own hardware keeps data in-house and removes usage fees, in exchange for an upfront cost and some maintenance. Many businesses end up with a mix. I work in all three and will tell you which fits.

- [Anthropic’s tools](https://lichucki.ai/services/anthropic-claude): Claude models and the Model Context Protocol, for assistants and agents that work inside your tools.
- [Google’s tools](https://lichucki.ai/services/google-gemini-enterprise): Gemini Enterprise, Vertex AI, BigQuery and Google Cloud, for businesses that already run on Google.
- [Open-source models, locally hosted](https://lichucki.ai/services/open-source-local-ai): Open-source models and custom AI on hardware you own, so your data stays in-house and there are no per-message fees.

## How an engagement works

1. **A conversation.** We talk about the problem: what the business does, the job that keeps repeating or the question you want answered, and what data and tools you already have.
2. **A scoped plan.** I follow the conversation with a scoped plan.

## Tell me what’s painful in your business.

I’ll tell you plainly whether AI can fix it and what it would take. [Work with me](https://lichucki.ai/contact).
