Beyond Chat: Creating Reusable AI Tools for Your Team

Make AI fast and effective for your workflows in minutes for consistent results.

Last Reminder! I'm hosting a free workshop in Northampton tomorrow, sign up here, and come by to network, and learn how to Build Business Value with AI!

Not able to make the workshop and ready to move beyond trial and error? Book a free 1-to-1 intro to AI session to chat about how AI can deliver value for your business, and the best place to start!

Last week, we explored why AI's built-in memory features often create more problems than they solve. Today, we're diving into something far more powerful: how to build reliable, shareable AI systems that actually work for your team.

The challenge isn't just getting AI to remember things—it's about creating consistent, reliable workflows that deliver value every time. This is where Custom GPTs and Claude Projects come in, and they're game-changers for scaling AI across your organisation.

Companies with the most successful adoption are creating 10s or 100s of Custom GPTs in some instances to handle all manner of tasks quickly and consistently.

What Can These Tools Actually Do?

Imagine having a specialist on your team who's available 24/7, perfectly briefed on your requirements, and can handle tasks like:

• Writing proposals that perfectly match your company's tone and include all your latest pricing
• Proofreading documents using your brand guidelines and house style
• Creating marketing content that's consistent with your messaging
• Answering customer service queries with your exact policies
• Analysing or generate reports using your specific metrics and KPIs

Instead of explaining these requirements every time or hoping the AI remembers them correctly, you build the instructions once and use them repeatedly.

When to Use Custom Tools vs Regular Chats

Think of it this way: You wouldn't write out detailed instructions for making a cup of tea every morning. You'd create a process, maybe even get a programmable kettle. The same principle applies to AI workflows. You need something more systematic than regular chat sessions when you find yourself:

• Repeatedly uploading the same company documents (like your brand guidelines or pricing sheets)
• Writing similar instructions over and over ("Please write this in our company tone of voice, in UK English...")
• Needing consistent outputs across your team (so everyone gets consistent results and quality)
• Wanting to share workflows with colleagues, clients, or partners (Be careful if sharing these publicly, they can leak your documents!)

The Tools:

What's Available and What You Need Before we dive in, let's be clear:

You'll need paid access to use these tools. Custom GPTs require a ChatGPT Plus subscription (￡20/month, a little more if you want a teams subscription to share the GPTs privately within your organisation), and Claude Projects need a Claude Pro account (￡20/month, a little more if you want to share projects within your organisation). Think of it as the difference between using a free email account and having a professional workspace setup. The paid plans offer better privacy and security, better models, features, and usage rates, if your business is on a free plan and you're serious about AI adoption, it's time to look at an upgrade.

Custom GPTs:

Your Shareable AI Specialist Think of Custom GPTs as creating your own version of ChatGPT that's pre-trained on your specific needs. The exciting part? They can do more than just chat:

• Connect to your calendar to schedule meetings or research an attendee
• Search the internet for up-to-date information
• Generate images based on your brand guidelines
• Analyse data from spreadsheets

You can turn these capabilities on or off depending on what you need. For example, if you're creating a GPT for internal document review, you might turn off internet access to ensure it only uses your approved content.

Building Better: Step-by-Step Setup

Let's get practical about how to actually build these tools. I'll walk you through setting up both systems, starting with Custom GPTs:

Custom GPTs: The Setup Process First, you'll need that ChatGPT Plus subscription (￡20/month). Once you have that:

1. Access the Builder

• Log into ChatGPT
• Click Explore Custom GPTs
• Look for "Create Custom GPT" in the top right
• Click "Create New"

2. Define Your Requirements

Think about what specific job you want this GPT to handle. For example, if you're building a proposal writer:

• What types of proposals will it write?
• What information must be included?
• What format should the output take?

There are two options to define this, a more conversational option that automatically writes the prompt as you chat with the tool, or in the configure tab, you can either edit the prewritten prompt, or write it yourself.

3. Choose Your Capabilities, You can enable/disable:

• Web browsing (great for keeping pricing up to date)
• Image generation (useful for creating mock-ups)
• Data analysis (for handling spreadsheets)
• Code interpretation (if you need technical outputs)

For that proposal writer, you might want web browsing to check competitor info, but disable image generation since it's not needed.

4 . Add Your Knowledge Base

Upload the documents your GPT needs:

• Pricing sheets
• Brand guidelines
• Past successful proposals
• Standard terms and conditions

Think of this as briefing materials for a new team member.

5 . Test, Test, Test

Testing is key, especially if this will be shared:

• Is it accurate enough? These can still make mistakes, even with documents attached
• Is the style in a format you like? Do you need to adjust the prompt or provide a template to be followed?
• Conduct extensive testing for public-facing tools to prevent errors that could harm your reputation.

Claude Projects:

Building Your Workflow with Claude Pro (￡20/month) takes a slightly different approach. Instead of creating a standalone tool, you're building a specialised workspace:

1. Project Setup

• Create a new project
• Name it clearly (e.g., "Client Proposals 2024")
• Write your description

2. Knowledge Base Similar to Custom GPTs

• Upload larger document sets
• Add internal policies
• Include example outputs
• Store templates

3. Custom Instructions

This is where you write the equivalent to the prompt in the configure box of a custom GPT.

• Step-by-step processes
• Decision trees
• Quality checks
• Review protocols
• What documents in the knowledge base it should reference, when, and how

4. Testing & Iteration

• Run the process
• Check the output
• Does it match your expectations?
• Do custom instructions need updating? Does it need a file about your business, or do the files need updating?

Beyond Setup: Making Your AI Tools Work Long-Term

The true measure of success isn't just building these tools—it's about creating systems that evolve and improve over time. While the initial setup might take a few hours, the real work lies in monitoring, refining, and adapting as both your needs and the technology change.

Monitoring Success and Failure

Start by tracking how your tools are actually being used. Are they saving time? Producing better results? Create simple feedback loops—even a quick rating system can help identify where adjustments are needed. When something goes wrong (and it will), document it. These "failure points" often reveal the most valuable insights for improvement.

Adapting to Model Updates

AI models are constantly evolving. Each update brings new capabilities—and occasionally, new quirks. When OpenAI or Anthropic announce major updates, take time to test your existing tools. Sometimes you'll find opportunities to leverage new features. Other times, you might need to adjust your instructions to maintain consistency.

The Feedback Loop

Encourage your team to share both successes and failures. Create a simple process:

• Document unexpected outputs
• Share particularly effective use cases
• Note when the tool misses context
• Identify patterns in successful interactions

This isn't about collecting data—it's about building institutional knowledge about what works for your specific needs.

Iteration Cycles

Set regular review periods—monthly or quarterly—to assess and update your tools. Ask:

• Which features are most used?
• Where do people still resort to manual processes?
• What new needs have emerged?
• How have model updates changed performance?

This Week's Experiment

Create a GPT or Project that can help you with one small task (reviewing emails, analysing your Linkedin analytics, writing report 1st drafts):

1. Check how it performs
2. Iterate on the prompt
3. Estimate how much time it saves you (and drop me a note to let me know!)

The goal isn't perfection—it's progress. Every iteration should make your tools more valuable for your team.

This Week's News

1. Eleven Labs' Smart Audio Evolution: The launch of ElevenReader's GenFM feature marks an interesting shift in AI audio. Having tested this extensively against other solutions, I find it particularly compelling. While Google's Notebook LM might win on pure voice quality, ElevenLabs' new offering delivers something more valuable: contextual understanding and reliable content adaptation. The ability to transform any document into a well-structured podcast, complete with relevant insights and natural transitions, could fundamentally change how we consume complex content. It's not just about text-to-speech anymore—it's about intelligent content transformation.
2. DeepMind's Socratic Learning - A Glimpse of Self-Improving AI: Google DeepMind's research into Socratic learning represents a significant shift in how we approach AI development. The framework introduces a practical method for AI systems to improve autonomously through structured "language games" without requiring external data. What makes this particularly noteworthy is its closed-loop approach. By creating environments where AI systems can generate their own training data and feedback, DeepMind may have uncovered a more scalable path to advanced AI capabilities. The implications for business AI implementations could be substantial—imagine systems that actively improve based on their interactions with your specific workflows.
3. OpenAI's Legal Challenges: The ongoing dispute between Elon Musk and OpenAI raises important questions about AI governance and commercialisation. While the legal drama makes headlines, the real story here is about the broader tension between profit-driven innovation and responsible AI development. For businesses relying on OpenAI's technology, this situation underscores the importance of having contingency plans and diversified AI strategies. It's a reminder that depending too heavily on any single AI provider carries inherent risks.

Transform How Your Business Uses AI

Beyond simple experimentation, Erictron AI workshops help you build AI systems that actually deliver value. Drawing from real implementations across many businesses, we've developed a practical approach that combines immediate results with long-term strategy.

What makes our workshops different?

?? Immediate Implementation

• Build working solutions during the session
• Leave with tools you can use tomorrow
• Focus on your specific business challenges

?? Strategic Development

• Identify high-impact opportunities
• Create scalable workflows
• Design processes that grow with your team

?? Practical Innovation

• Learn by doing, not just watching
• Test ideas in real-time
• Refine approaches based on actual results

?? Sustainable Growth

• Build skills that last beyond the workshop
• Create frameworks for ongoing development
• Establish clear metrics for success

Whether you're just starting with AI or looking to scale your existing implementation, our workshops help you build the right foundation for success. No more hoping AI picks up the right context—instead, create intentional systems that work for your whole organisation.

Besides workshops, we also support: hands-on implementation and solution development, AI policy support, AI Strategy, and product development.

Ready to move beyond trial and error? Book a discovery call and let's discuss how to build AI workflows that actually work for your business.