# Questions and answers

Plain answers to the questions business owners ask about AI, and about working with me.

## About working with me

### Who is Alek Lichucki?

I am Alek Lichucki, an AI engineer based in Los Angeles who builds machine learning and AI systems for small businesses, starting from the problem that is costing them time.

I have six years of experience across consulting and product companies and a master’s in data science from UC Berkeley.

See: [About me](https://lichucki.ai/about).

### What does Alek help businesses with?

Four things: machine learning models such as forecasts and scoring, AI assistants and agents that work inside a business’s own tools, chat-to-data and knowledge bases, and data strategy.

See: [Services](https://lichucki.ai/services).

### What kinds of problems do you solve?

The painful, everyday ones: staff repeating the same manual work, numbers nobody trusts, waiting days for a report, guessing at forecasts, and data too sensitive to send to an AI service.

I start from the problem and then pick the tool.

See: [Services](https://lichucki.ai/services).

### Why work with Alek instead of a larger firm?

You work directly with the engineer who builds your system.

My day job is rolling out AI across companies, and a small business gets the same experience and standards without a consultancy’s team or overhead. I also have no vendor to sell, and I recommend Anthropic’s tools, Google’s tools or open-source models according to the problem.

See: [Work experience](https://lichucki.ai/work), [About me](https://lichucki.ai/about).

### Who is a good fit?

A small business that has data or repetitive work, wants AI to help with it, and does not have its own AI team.

I take on small businesses only.

Remote clients anywhere are welcome.

See: [Services](https://lichucki.ai/services).

### Where is Alek based, and are remote clients welcome?

I am based in Los Angeles and work with clients remotely, anywhere.

### Which AI tools does Alek work with?

All three kinds: Anthropic’s managed tools, Google’s managed tools including Gemini Enterprise, and open-source models running on a business’s own hardware.

I recommend whichever fits the problem.

See: [Anthropic’s AI tools](https://lichucki.ai/services/anthropic-claude), [Google’s AI tools, including Gemini Enterprise](https://lichucki.ai/services/google-gemini-enterprise), [Open-source models and custom, locally hosted AI](https://lichucki.ai/services/open-source-local-ai).

### Do you work with open-source AI models?

Yes. I build with open-source models as well as the managed tools from Anthropic and Google.

My own private assistant answers every message with an open-source model on hardware I operate, and is in daily use.

See: [Open-source models and custom, locally hosted AI](https://lichucki.ai/services/open-source-local-ai), [The private assistant](https://lichucki.ai/projects/private-assistant).

### How do I get in touch?

Through the contact form on this site.

See: [Contact](https://lichucki.ai/contact).

## Choosing an approach

### What does an AI consultant do for a small business?

An AI consultant works out where AI will save time or improve decisions in your business, builds it using your own data and tools, and leaves you able to run it.

For a small business that usually means one focused project, not a programme.

See: [Services](https://lichucki.ai/services).

### Should we use Claude, Gemini, or open-source models we run ourselves?

It depends on where your data lives, how sensitive it is, and how much you will use it.

Managed tools from Anthropic or Google are the fastest to start with and need the least upkeep. 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 use a mix.

See: [Anthropic’s AI tools](https://lichucki.ai/services/anthropic-claude), [Google’s AI tools, including Gemini Enterprise](https://lichucki.ai/services/google-gemini-enterprise), [Open-source models and custom, locally hosted AI](https://lichucki.ai/services/open-source-local-ai).

### Can AI run privately on our own hardware?

Yes. Open-source models can run on a single machine you own, so your data never leaves the building and there are no per-message fees.

I run a private assistant on self-hosted open-source models, and it is in daily use. The trade-offs are an upfront hardware cost, some maintenance, and models that are smaller than the largest managed ones.

See: [Open-source models and custom, locally hosted AI](https://lichucki.ai/services/open-source-local-ai), [The private assistant](https://lichucki.ai/projects/private-assistant).

### What is Gemini Enterprise, and can you help us adopt it?

Gemini Enterprise is Google’s managed AI offering for organisations.

I work on enterprise-wide AI enablement on Google Cloud in my current role, and I help businesses plan and roll out Google’s AI tools.

See: [Google’s AI tools, including Gemini Enterprise](https://lichucki.ai/services/google-gemini-enterprise).

### Do we need a lot of data to use AI?

No. Assistants and chat-to-data tools mostly need your data to be organised and well described, not large.

Forecasting and scoring models need enough history to learn from, and the first step is checking whether you have it.

See: [Chat-to-data and knowledge bases](https://lichucki.ai/services/chat-to-data), [Machine learning builds](https://lichucki.ai/services/machine-learning).

### Can an AI assistant be trusted to take actions for us?

It can when it is built to ask first.

The assistants I build require a confirmation before anything that sends, spends or changes something, keep a record of every action, and never act on instructions found in an email or web page.

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

### How do we ask questions of our data in plain English?

By putting a knowledge base between your data and the AI that records what your tables and metrics mean.

Without it, AI writes queries that look right and are wrong for your business.

See: [Chat-to-data and knowledge bases](https://lichucki.ai/services/chat-to-data), [The knowledge system](https://lichucki.ai/projects/knowledge-system).

## Practical

### How do you know an AI system is working?

By testing it against cases with known answers before and after every change, and by recording what it does in use.

My own assistant project has about 1,200 evaluation cases, 60 of them taken from real use.

See: [The private assistant](https://lichucki.ai/projects/private-assistant).

## 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).
