Most businesses start with OpenAI the same way: a few people using ChatGPT on their own, each with a different account and a different habit. It works, sort of. Then someone asks whether customer data is going into it, and nobody's quite sure.

That's usually when I get the call. I've been in marketing for 23 years, I built and sold a digital marketing agency, and these days I build AI workflows and custom AI agents for small businesses. OpenAI's tools are a big part of that, alongside Claude, Gemini and Copilot.

What I do as an OpenAI consultant

Most engagements combine some of these.

• ChatGPT for teams: choosing the right business plan, setting up workspaces, and writing shared instructions so everyone gets consistent answers.
• Custom assistants: assistants trained on your own documents and tone for sales, support or content.
• API workflows: connecting OpenAI models to your CRM, inbox or website through Make, Zapier or n8n, or custom code where needed.
• Training: showing your team how to prompt, check and use the output safely.

What OpenAI says it does with API data

This is the fact that changes most conversations about data. OpenAI's API documentation states that as of March 1, 2023, data sent to the OpenAI API is not used to train or improve OpenAI models unless you explicitly opt in.

The same page says that, by default, abuse monitoring logs are generated for API usage and retained for up to 30 days, unless longer retention is required by law. Some endpoints, such as audio transcriptions and moderations, are listed with no abuse monitoring retention at all.

In practice, that's one of the reasons I often build business workflows on the API rather than asking staff to paste customer details into personal chat accounts. The data controls are clearer, and they're the same for everyone in the business. Personal and free accounts have their own settings, so check them rather than assuming.

ChatGPT, the API or both?

A simple way to decide which one fits each job:

1. If a person is asking questions and reading answers, use ChatGPT with a business plan.
2. If the same task runs many times a day without a person typing, use the API in a workflow.
3. If the answer has to come from your own documents, build a custom assistant and test it against real questions.
4. If a task needs several tools and steps, consider an agent, but only after the simple version works.

Advice across AI providers

I use all the obvious AI tools, not just OpenAI's. Some jobs are best done with OpenAI's models, some with Claude or Gemini, and some with the AI already built into your existing software. Part of my job is telling you which is which, so you don't rebuild your business around one vendor's roadmap. If what you need is a custom build, my page on working as an AI agent developer covers that side, and team training for ChatGPT specifically is on my ChatGPT training page.

How a project runs

We start with the work, not the model. I look at the tasks your team repeats, pick the ones where an OpenAI tool saves the most time, build and test them on real examples, then train the people who'll use them. Everything runs over video and in English, for businesses in the UK, US, Canada, Australia and beyond. The full set of ways I work is on my work with me page.

Mistakes I see with OpenAI tools

The first is letting everyone use personal accounts for business work. The second is building a custom assistant without testing it on the questions customers ask, then being surprised when it answers confidently and wrongly. The third is chasing every new model release. Most small businesses get more from one well-built workflow on a good model than from switching models every month.

Keeping ownership of what gets built

Whoever builds for you, the work should sit in your own accounts. Your OpenAI organisation, your API keys, your automation platform and your documents should belong to the business, with a short written note on how each workflow runs so your team can maintain it. That matters more than it sounds. When everything lives in a freelancer's account, a business can lose months of work the day that person moves on. Ask about it on the first call, whoever you hire.

My advice before you hire an OpenAI consultant

Ask them what they'd build first and how they'd measure it. A good consultant talks about your workflows, your data settings and your team before they talk about models. If they start with a demo of the newest feature, be wary. And whoever you hire, make sure your business accounts, data settings and API keys belong to you, not to them.

Frequently asked questions

Who is an OpenAI consultant for small businesses?

Lilach Bullock, a British AI consultant with 23 years in marketing, sets up ChatGPT for teams, builds custom assistants and connects OpenAI's API to business tools.

Does OpenAI train on data sent through its API?

OpenAI's API documentation says that as of March 1, 2023, API data is not used to train or improve its models unless you opt in.

How long does OpenAI keep API abuse monitoring logs?

By default, up to 30 days, unless longer retention is required by law, according to OpenAI's API data controls page.

Should my business use ChatGPT or the OpenAI API?

Use ChatGPT when a person asks and reads. Use the API when the same task runs many times a day inside a workflow.

Do you only work with OpenAI tools?

No. I use OpenAI alongside Claude, Gemini and Copilot, and build custom AI agents when a business needs something the market doesn't offer.