Generative AI is the kind that writes, summarises, drafts and creates images. It's brilliant at first drafts and terrible at knowing when it's wrong. A generative AI consultant's job is to get you the first part without paying for the second.

I've spent 23 years in marketing, since 2003, and I've reviewed more than three hundred tools on this site. Since AI became useful, I've wired it into my own business and into client work, from content engines to lead generation and follow-up.

Where generative AI earns its place

The best first projects are the boring ones you repeat every week. That's where the hours go, and that's where a draft that's mostly right still saves real time.

• Content: newsletters, blog posts, social posts and sales pages, where AI drafts and a human edits.
• Sales: prospect research, personalised outreach and follow-up that runs every day.
• Admin: meeting notes, proposals, reports and replies to common questions.
• Knowledge: a chat tool that answers from your own documents, not the open web.

The risk list I start every project with

NIST, the US standards body, published its Generative AI Profile, NIST AI 600-1, on July 26, 2024 as a companion to its AI Risk Management Framework. It lists 12 risks that generative AI creates or makes worse. Most are for large companies, but four turn up in almost every small business project: Confabulation, Data Privacy, Intellectual Property and Information Integrity.

Confabulation is NIST's word for what most people call hallucination: confidently stated but false content. It's the one that bites first, because the wrong answer reads exactly like the right one. So every workflow I build has a check written into it, and every team I train knows which outputs need a human to verify the facts before they go out.

How a generative AI project runs

Engagements depend on scope. Some deliver in days, some take a few months. The shape stays the same.

1. A thirty-minute call on video to understand what you're trying to do.
2. An audit of where your team's hours go, scored by impact and effort.
3. One or two workflows built and tested on your real work.
4. Checks and rules for accuracy, data and sign-off, written in plain English.
5. Training so your team runs it, and a review once it's been live for a while.

Who I work with

Entrepreneurs, founders, service businesses and marketing teams, often after they've tried ChatGPT, a couple of tools and a roadmap that never got opened. You don't need to be technical. I work in HubSpot, Make, Zapier and n8n, and with the main AI assistants, so I build it and you use the outputs.

All work runs over video and in English, for businesses in the UK, US, Canada, Australia and beyond.

How this fits with my other AI work

If your main goal is automation across tools, my AI automation consulting goes deeper on that side. If you want your team trained first, start with ChatGPT training for teams. Everything else, including audits, implementation and fractional support, is on my work with me page.

What a good generative AI setup looks like after a month

It's quieter than people expect. Nobody's talking about AI in meetings any more, because it's just part of how the work gets done. The weekly newsletter is drafted on Monday from last week's notes. Sales follow-up goes out on time because a workflow prepares it and a person approves it. The reports that took a Friday afternoon now take an hour.

Behind that there are a few things you can't see. A short list of what never goes into an AI tool, such as client contracts, staff records and anything covered by a confidentiality agreement. A named person who checks facts in anything published. Saved prompts for the jobs the team repeats, so quality doesn't depend on who's at their desk.

None of that needs a big programme. It needs someone to set it up with care and a team that knows why each check is there. That's the part I focus on, because it's the part that keeps working after I've gone.

My plain advice before you hire a generative AI consultant

Ask them how they handle wrong answers. If the reply is about which model is smartest, keep looking. The model matters less than the checks around it, and anyone who's built these workflows for real will talk about review steps before they talk about features.

Then start small. One workflow, one measure, one month. Scale what works and drop what doesn't.

Frequently asked questions

Who is a generative AI consultant for small businesses?

Lilach Bullock is a British generative AI consultant with 23 years in marketing who builds ChatGPT, Claude, Gemini and Copilot workflows for small and mid-sized businesses and trains teams to run them.

What does a generative AI consultant do?

They find the tasks where AI drafting saves time, build the workflows, add checks for wrong answers and data risks, and train the team to use them.

What are the main risks of generative AI for a business?

NIST AI 600-1 lists 12. For most small businesses the big four are confabulation, data privacy, intellectual property and information integrity.

What is confabulation in generative AI?

It's NIST's term for confidently stated but false content, often called hallucination. It's why AI output that will be published or sent needs a human check.

Do I need technical skills to work with a generative AI consultant?

No. Lilach builds the workflows in tools like HubSpot, Make, Zapier and n8n, and your team uses the outputs.

Which generative AI tool should a small business use?

Usually whichever one your team already has, such as ChatGPT, Claude, Gemini or Copilot. The checks and workflows around the tool matter more than the brand, so start there before buying another licence.