There is no shortage of AI hype aimed at small businesses. Most of it is vague, theoretical, or trying to sell you a platform. These are four ways I use generative AI in my own work, every day, that have hugely sped up my workflow — making me productive as both an employee and freelancer.
1. Research and Requirements Gathering
If your business involves any kind of research — market analysis, competitor reviews, scoping a new product, evaluating vendors — generative AI is transformative here.
I recently led a complex requirements gathering project for Thinkbox, working across 12+ stakeholders from ITV, Sky, and Channel 4. Using Claude Code as a knowledge work tool, I was able to process lengthy interview transcripts, extract themes, identify areas of agreement and disagreement, and maintain a living requirements document — work that would traditionally need a larger team.
Where to start: Record your next stakeholder call with a tool like Fathom, feed the transcript into Claude with a structured prompt asking it to extract requirements, flag contradictions, and map dependencies. Chain multiple transcripts together across sessions and you have a living analysis that builds on itself — far more useful than a summary.
2. Code Pairing and Prototyping
You don’t need to be a developer to benefit here. AI coding assistants can help you build internal tools, automate spreadsheets, create simple web applications, or prototype an idea before committing budget to a full build.
I use Claude Code daily as a pair programmer — it is embedded in my workflow for everything from data pipelines to building this website. For an early-stage fashion AI startup, I used it to accelerate the development of a deep learning prototype that helped them secure their first round of investment.
Where to start: Pick a process you currently do manually — a data transformation, a report, a client deliverable — and try building it with an AI coding assistant like Claude Code or GitHub Copilot. You will be surprised how quickly you can go from “I have an idea” to a working prototype, even with limited coding experience. And if you are technical, the productivity gain is substantial — this website, for example, was built and deployed using Claude Code as a pair programmer.
3. Automation of Repetitive Workflows
Every business has processes that eat time — reformatting reports, preparing recurring analyses, pulling data from one system into another. Generative AI is excellent at taking a manual workflow and helping you systematise it.
At ITV, I have used AI-assisted automation to transform measurement workflows — turning manual, repetitive analysis processes into streamlined pipelines. The same approach scales down. If you are spending hours each week on a task that follows a predictable pattern, there is almost certainly a way to automate a significant chunk of it.
Where to start: Map out a workflow you run regularly — the inputs, the transformations, the outputs. Then build it as a pipeline: use an AI assistant to write the glue code, schedule it, and have it output directly into wherever you consume the results (a dashboard, a Slack channel, an email). The goal is not to remove yourself entirely — it is to eliminate the manual steps so you only engage at the decision points.
4. Agentic Workflows Grounded in Your Own Data
This is where things get genuinely exciting. It is now possible to build AI workflows that are grounded in your own business data — your CRM, your analytics, your internal documentation. The AI is not just generating generic responses but working with your actual information.
I have built agentic workflows grounded in BigQuery data and internal sites, where the AI can query real data, reason across it, and produce outputs that are specific to the business context. This is not science fiction — the tooling exists today and is accessible to businesses of any size, provided the data is reasonably well organised.
Where to start: If your data lives in a warehouse like BigQuery, Snowflake, or even a well-structured PostgreSQL database, you can connect an AI agent to it today using tools like Claude with MCP or Agent Development Kit (ADK). Start with a single, well-defined question your team asks repeatedly — the kind that currently requires someone to write a query and interpret the results. Build an agent that can answer it on demand, then expand from there.
If you are not sure where to start with AI — or you have tried and it has not landed — get in touch. We offer a free 30 min consultation on how we might be able to help you.