Custom GPTs: Why you need them, and how to create them

How to make a chatbot that leverages your unique data, using ChatGPT as a template

Whatever your personal feelings about the technology, GenAI is here to stay.

 

Globally, we’ve seen organisations across multiple sectors leverage it to cut admin, streamline workflows and boost productivity.

 

Sounds great, but what specific uses does GenAI represent for SMEs like yours?

 

If you rely on written copy, you’ve likely dabbled with OpenAI’s chatbot – ChatGPT.

But did you know you create your own personalised version of ChatGPT’s model?

 

Custom GPTs are affordable, accessible AI copywriting tools. More personalised than the average chatbot, they can be configured to work with your unique brand materials and reference data. Stuff that’s not included in the base LLM.

This allows them to mimic your tone of voice and refer to your internal data – creating copy that feels like it was drafted by your team.

 

Here, we’ll look at everything businesses need to know about custom GPTs.

 

 

What are custom GPTs?

If you’ve used ChatGPT, you’re already halfway there.

 

A GPT is a Generative Pre-trained Transformer – a type of AI model trained on huge amounts of data to generate text, images, code and more.

 

A custom GPT is a version tailored specifically to your business.
It augments its existing knowledge base with your own files. These may include brand guidelines, handbooks or blog posts. It uses these to generate content that fits your tone, reflects your values, and references your internal data.

At the time of writing, each custom GPT can hold up to 20 documents and 2,000 words of direct instructions.

 

What are the benefits of custom GPTs?

Custom GPTs are refer to your own materials, so they’re more relevant, accurate and brand-aligned.

 

A custom GPT can benefit your brand in a number of ways.

 

Reduced operational costs

As your business grows, custom GPTS can prevent your operational costs from growing with it. This enhances scalability without compromising your brand.

Custom GPTs can:

 

  • Answer customer queries.
  • Carry out basic sales responses.
  • Generate marketing copy.
  • Write product descriptions.
  • Summarise internal reports.

 

This saves businesses money by reducing the need to increase headcount in line with business growth.

 

What’s more, if you have ChatGPT Plus (currently £20 a month inc. VAT), you can create as many custom GPTs as you like without additional charges.

 

Time savings

As well as reducing operational costs, custom GPTs can also help businesses save time. They can lighten teams’ admin loads – leaving them with more time and energy for the work they enjoy.  

 

With custom GPTs, teams can easily redraft or personalise emails, ad copy and social media posts. They can help automate employee onboarding when they ingest your company handbooks and operational manuals. 

This allows GPTs to act as self-service resources for new hires.

 

Revenue growth

As well as reducing overheads, custom GPTs can also help to increase revenues.
They can:

 

  • Accelerate sales cycles by creating customised pitches.
  • Generate proposals based around the customers pain points you provide. 
  • Embedded in ecommerce sites, they can recommend products and loyalty perks to boost upselling.

 

Improved brand integrity  

Custom GPTs can be configured using your company’s unique data.

This may include:

 

  • Blog posts.
  • Style guides.
  • Proprietary research.
  • Product descriptions.
  • Guides and FAQs.
  • Service pages.
  • Information about your staff.

All this allows them to generate copy that feels authentic to your brand, and provides value for customers.

Common use cases for custom GPTs

Whatever your field, there’s a good chance you can integrate custom GPTs into your operations.

Let’s explore some examples:

 

Customer support and engagement

Custom GPTs are configured using internal data like support documents, product guides or FAQs. Therefore, they know your business, your products and your customers’ needs better than “out of the box” chatbots.

By automating administrative processes, custom GPTs can improve the customer experience, providing 24/7 support. By taking on the lion’s share of entry-level enquiries, they allow client-facing staff to spend more meaningful time with customers. Resolving more disputes and delivering more value.

 

Internal knowledge base management

Supplemented with your internal materials, they become always-on resources for onboarding, troubleshooting and policy queries.

 

You can augment your custom GPT with:

 

  • Employee handbooks
  • Internal operating procedures
  • Product catalogues
  • Technical documents

 

Sales support

Do your sales teams ever complain that so much of their time is spent on personal admin rather than selling?

 

Custom GPTs can help them automate their workflows – giving them more time to convert leads and smash their targets. 

 

By referencing your sales communications, a custom GPT can speak with the voice of your sales team. It can help with: 

 

  • Creating outreach emails.
  • Sending scheduled follow-ups. 
  • Responding to prospects’ objections.
  • Generating proposal drafts.

 

Marketing and content creation

By getting to know your brand’s ToV and communication style, custom GPTs can help you upscale your marketing copy.

 

They can help with everything from coming up with new ideas to redrafting copy to better fit its readership.

 

Training and onboarding

With myriad associated costs, taking on new staff is expensive, with average onboarding costs ranging from £3,600 to £5,000. By automating new hire processes, Custom GPTs can help employees hit the ground running.

 

By referring to company handbooks and other such resources, they can act as an interactive self-service resource for new hires.

 

Anything from dress codes to IT support contact details can be accessed by teams 24/7, wherever they are.

 

Legal and compliance support

Legal and compliance concerns can be tricky for small businesses with limited resources.

 

While custom GPTs can help with this, it’s important to remember the risk of hallucination. This is where LLMs generate content which is inaccurate or nonsensical. They may cite studies or legal cases that don’t exist. 

 

Nonetheless, there are several low-risk ways your business to use custom GPTs for legal support.

 

  • Summarising long legal documents in layman’s terms.
  • Breaking down policies for non-experts. 
  • Providing non-binding loose summaries of policies.
  • Automating form generation. 

 

 

How to create a custom GPT

Custom GPTs can be set up without any coding experience. If you already have a ChatGPT Plus account, it’s also free. 

Let’s take a look at how you can set one up right now.

 

Step One – Go to Chat GPT

  • Log into your ChatGPT account 
  • Check that you’re using ChatGPT Plus to access the latest version and Custom GPTs

 

Step Two – Click “Explore GPTs”

  • In the left sidebar, click “Explore GPTs”.
  • Click “Create” in the top-right corner.
  • Enter the Custom GPT Builder (essentially a guided form).

 

Step Three – Establish your GPT’s identity

  • Give your GPT a name (e.g. HR Bot, Sales Buddy, Copy Droid, Keith)
  • Describe its function in plain English (e.g. “You are a helpful assistant for an online meal kit subscription service. You explains product, ingredients and allergens and suggest recipes based on customers’ preferences”.)

 

Step Four – Establish your GPT’s core behaviour

Core behaviour refers to the way ChatGPT behaves ‘out of the box’. These behaviours can be replaced with 2,000 words of your custom instructions.

To do this:

 

  • Define the typical user: their goals, background, and expectations (e.g. “The user is a small business owner with limited technical knowledge looking for help with marketing ideas”)
  • Define its tone of voice and brand personality
  • Include vocabulary parameters to include or avoid (e.g. “Your tone should be warm, friendly and professional, with concise answers and clear examples. Avoid technical jargon and unexplained acronyms”).

 

Step Five – Upload resources

  • Upload the materials on which you want to train your GPT. These can be PDF, doc, or txt files. It can even recognise images.
  • Examples may include ToV style guides, previous blog posts, employee handbooks, product catalogues, product troubleshooting guides or anything else your GPT might need to do its job effectively.

 

Step Six – Add tools

You can add additional tools to expand your custom GPT’s abilities. 

Some examples include:

 

  • Python for interpreting code or analysing data
  • DALL·E for generating images  
  • A web browser tool for tracking information in real time (e.g. competitive analysis or trending topics)
  • File upload capabilities (enabling team members to upload their own files to the GPT to ingest). 

 


Step Seven – Test your custom GPT

  • Try testing your new GPT with a few sample queries to see how it responds
  • Tweak prompts or add new instructions if the tone or focus isn’t quite right
  • Keep adding new instructions or uploading supporting documents as needed
  • Enable memory and explicitly tell your GPT what you want it to remember. It will update and reconfigure based on what you tell it to remember.

 

It’s important to remember that even with access to your files and data, GPTs can still be prone to hallucinations and inaccuracies. All outputs should be fact checked against the files and data your custom GPT has ingested.

 

Are custom GPTs worth it?

Like all AI technologies, GPTs are evolving at breakneck speed. 

 

Each iteration brings a new host of capabilities. Chat GPT’s base model now includes built-in customisations, dynamic prompts and multimodal capabilities. In this context, do businesses like yours even need custom GPTs?

 

To use a tech analogy, nobody wants to buy a digital camera that will be superseded by the average smartphone in six months.

 

Opinion within the AI community is divided on this, but it’s unlikely that custom GPTs will be outpaced by ChatGPT’s standard model anytime soon.

 

This is because their value lies in your data. Still, there are some caveats. 

 

A custom GPT is not a “set it and forget it” affair. In order to get the best out of your GPT, it should be constantly tweaked and retrained. Ingesting new documents and receiving more system-level directives can help your GPT to evolve. 

 

The more of your own unique materials you can use to train your Custom GPT, the more fit for purpose it will be. 

 

Uploading proprietary data to ChatGPT also carries security caveats.

 

OpenAI encrypts data in transit and at rest, for business users it also offers retention controls including a zero-data option.

However, standard accounts may allow inputs to be stored and used for model training unless opted out. Even with protections, risks remain. Data could be exposed through misconfiguration, policy loopholes or cyber attacks like prompt injections.

For sensitive or regulated information, businesses should adopt a zero-trust approach.

 

Limit what you upload, check contractual terms, and ensure internal governance before relying on custom GPTs.

 

Let us help you breathe life into your AI content

Now you’ve created a custom GPT, it’s up and running and producing good content.

 

But… it still needs a little something.

 

That secret sauce that makes good content great.

 

That’s where our AI copy editing service comes in.

 

Working with experienced writers and editors, we provide AI-generated copy with that all-important human touch. Once you’re happy that we’ve got it just right, you can use our human-polished copy to further train your custom GPT.

 

Intrigued? 

 

We can help you master the balance between quality, scalability and affordability. 

So, why not give us a call?

< Back

We work with

Responsive website designed & developed byMadison Web Solutions logo