And we’re live. Hello and welcome to our ChatGPT Basics webinar. Let me start with a statistic: £94 billion. McKinsey estimates that if SMEs became just 1% more efficient, it would add £94 billion to the economy within five years, and then again and again after that. That’s huge. We’re at a point where AI is one of the most powerful levers we have to improve efficiency, not just by 1%, but by 10, 20, even 40% in some areas. Yet we’re not really seeing those gains. Most people are only scratching the surface of what tools like ChatGPT can do. That’s the purpose of this session.

By the end of today, you should have a clearer understanding of the ChatGPT interface, what the different parts do, and how models, projects and custom GPTs fit together. You should leave with a more useful setup to get started, some practical time-saving tips, and the confidence to explore further. This is very much a first-steps session.

To introduce myself, I’m Crispin. I’ve been in tech a long time. When I started, Google didn’t exist and people still said “www” at the beginning of web addresses. Since then, I’ve seen multiple waves of change: the dot-com boom, Web 2.0 and user-generated content, the rise of social media, mobile-first, and the move to cloud computing. AI is probably the biggest shift we’ve experienced, but it’s still another evolution. We’ve adapted before, and largely those changes have been positive. Typically, we’ve been able to do more rather than facing industry-destroying catastrophes.

I run The Coders Guild, a training firm that’s industry-led and community-driven. All of our training is devised, delivered and reviewed by professionals who actually do these jobs. We focus on making a positive impact in our communities. I also sit on the Federation of Small Businesses technical committee. A real passion point for us is making sure tech capability reaches small businesses, not just larger organisations. We are a small business ourselves, and that’s who we’re here to help.

We’re currently working with Skills England and the Department for Education to develop a new Level 4 apprenticeship in AI and automation, which should launch early next year. During our research, we spoke to lots of people in businesses and communities to understand their needs. One consistent theme was that many people are already using tools like ChatGPT, Claude and Gemini, but only at surface level. There’s a big jump between someone using a free ChatGPT account casually and someone integrating AI into automated systems and leveraging machine learning. In the middle is a huge opportunity for efficiency gains.

You don’t need deep technical knowledge to take that next step. You don’t need to understand how large language models work or the implications of general AI. Instead, think about everyday use cases, not just writing LinkedIn posts or rewriting emails, but practical tasks. For example, the other day my family went out for dinner with two other families. When the bill arrived, we needed to split it. I opened Claude, although this would work just as well in ChatGPT, and used voice input. I said, “The bill is this much. We have three families. Family one is one adult and one child. Family two is two adults and three kids. Family three is two adults and two kids. Add a 12.5% tip and calculate the split.” It did the calculations instantly and showed the workings. We didn’t have to do any manual maths. That’s the kind of everyday efficiency we’re talking about.

Now I’m going to move into a live demo. You’re not supposed to do live demos in webinars, especially if you’re out of practice, but we’re going to go for it. I’ll demonstrate using ChatGPT, although most of these features are similar in Claude, DeepSeek and other tools. The first thing I want to talk about is custom instructions. If you go into the settings and choose “Customize ChatGPT,” you’ll see options like what ChatGPT should call you, what you do, what traits it should have, and anything else it should know about you. These are important. Don’t just accept the first version. Refine them.

If you’re not sure what to write, you can ask ChatGPT to help. For example, you could type, “I’m looking to create custom instructions. Ask me some questions to help draft a response.” It will then ask clarifying questions. You can respond using voice input if you prefer. You might say that you want help being concise in emails, drafting LinkedIn content, keeping the tone informal, or ensuring it doesn’t sound AI-generated. You can specify that you write in UK English. ChatGPT will then draft custom instructions that you can paste into the settings. Over time, you can refine them. For instance, mine clearly state that voice notes should always be transcribed in UK English. Custom instructions are key to getting consistent, high-quality responses.

At the top of the interface, you’ll see different model options. If you’re on a paid plan, you generally don’t need to change models. The latest model will handle most tasks well, and there’s rarely a strong reason to switch manually.

Next are projects. Projects allow you to group chats and documents together. You might create a project for a specific client, a marketing campaign, or your standard operating procedures. When you add documents to a project, every chat within that project references those documents automatically. Language models don’t truly remember things; they reference context. Projects provide structured context so you don’t have to repeat yourself. For example, in our marketing project, ChatGPT already knows how we make money, what our current campaigns are and how we position ourselves. That saves time.

Then there are custom GPTs. These are specialised versions designed for specific tasks, almost like mini apps. For example, there’s a Canva GPT that understands Canva’s functionality and helps you create slides or flyers before exporting them. You can search existing GPTs for tools you already use, search by use case, or create your own if you’re on a paid plan. We’ve built a few ourselves. One supports our sales team by understanding our sales process, objection handling and course schedules. Another helps standardise lesson plans and slides. Another acts as a copywriter trained on our brand tone of voice. Creating one follows a similar process to custom instructions: define the function, ask ChatGPT what it needs to know, upload relevant documents and refine.

When thinking about the difference between projects and custom GPTs, I see a project as a workspace where all the materials for a broader context live, such as a product launch or ongoing client relationship. A custom GPT is more like a specialised assistant for repeated, focused tasks. You can often achieve similar results with either. The important thing is to experiment. Remember that ChatGPT is there as your assistant. If you’re trying to do something, you can ask it for help in figuring out how to do that thing.

To close, there are three key takeaways. Tools only help if you know how to use them well. The first step isn’t big, and once you take it, the next step becomes easier. Often it’s the small improvements that generate the biggest returns. The goal isn’t to become an AI expert. It’s simply to become a bit better at using the tools that are already available to you.

We’re also running an AI and Automation Skills Bootcamp funded by Lincolnshire County Council. If you’re in the Lincolnshire or Rutland area, you can sign up via the QR code, and we’ll also share the link. The course covers ChatGPT, Claude, Perplexity, automation tools like Zapier and Make.com, and digital tools such as Notion. It’s about moving from being comfortable with these tools to understanding how they interact and how to use them for real productivity and efficiency gains within your team.

We’ll be running more sessions, including comparisons between different tools and models and exploring what each one tends to do best. This is hopefully the first of many. Thanks very much for your time and attention.