Week 7: Building a custom career coach for your team

First, an ask - if you like this newsletter, please share it with a friend or on Linkedin ❤️

So i’m still in sunny Florida this week before heading home, and it seems a lot of my conversations have ended up with US folks, always something I love - I think its the energy!

From ranting about the state of AI in HR conversations at transform in Vegas with a couple of folks who went, to learning about a fascinating tool for Performance - shoutout Incompass Labs.

Aside from this, I enjoyed a little Orlando Art fair, took a trip to Tampa (Hyde Park is beautiful) and looking forward to heading home and a weekend with my kids ❤️

Back to the reason you’re here… This week, I’m going to focus on:

• How to build a CustomGPT for career coaching

• Sharing how I use Lovable to build small products, and why you should too!

• News - the release of excellent native image creation in ChatGPT as well as the updates to Zapier using MCPs

Who Am I?

For new subscribers, I’m Matt Bradburn. I was VP People at Peakon, built and sold the DBR community to Talentful, built and sold People Collective to Scede and now working on several “small bets”.

I have been using AI in my work for the last 2 years, starting from nothing and now able to build agents and small apps, or provide deep research to clients to solve their challenges.

Now I want to share my learnings with you all to help you grow.

🎯 Deep Dive of the Week

The Challenge:

This week I have been diving back into writing ranting about career progression and levelling / career architecture on Linkedin…

Now I am going to show you how to build a CustomGPT to solve for one of the biggest challenges 💪

Having run over 75 of these projects for startups and scaleups of all sizes, from CharlieHR (a nice run down of what I built for them HERE) through to Pret a Manger, so I know what it takes to get it landed right!

However, a static framework and manager training, although important, is sometimes not enough to help people OWN their own careers.

So I wanted to get creative with ways we could improve this and I settled on building a customGPT.

Why?

• We can combine the details of our framework (hello FinEdge again), with our guides for managers, guides for employees.

• We also get a great common chat interface and coaching capabilities of chatgpt

• This allows us to build a co-pilot for employees to own their own career development, within the guardrails of our company.

I will breakdown the process of building this, and then give you the link to the customGPT and the materials. You can then re-create it and substitute your own internal content in (or if you need to work on that, you can always hire me 😉)

Roadmap Buildout

What do we need to make this work?

1. Our materials from inside the company - our progression framework, manager guides, employee guides, transcripts from training sessions.

   1. These need to be detailed, so I asked ChatGPT to help me set out exactly what I needed for this to be best in class.

2. Our processes - what do our promotion cycles look like, when do we review compensation, what is our compensation philosophy - remember, progression underpins all of this.

3. Instructions for our CustomGPT

   1. The tone of voice (e.g., supportive, direct, coach-like).

   2. The level of detail in responses.

   3. How the GPT should tailor responses based on employee level (e.g., early-career vs. senior).

Step 1: Collate your Materials

We need depth and details to ensure we’re operating within the guardrails, and I of course asked chatgpt

Please build a detailed table of the inputs required to build a best in class coaching focused customgpt for employees. I have a lot of materials relating to progression, levelling, compensation and performance. These include 4 level pathways with full frameworks, performance guides, transcripts, details on sub levels etc. Please set this out clearly and ensure you make it clear what format is best for building a customgpt.

For example, I downloaded each individual pathway:

1. Enablement

2. Professional

3. Manager

4. Executive

But I also made sure to instruct the GPT later that progression is based around company need, not something which is automatic, and that employees should not expect linear progression.

It’s these details which make all the difference.

So I had:

• 4x progression frameworks for each path, each with 4 competencies - exported as CSV files

• A manager guide to progression - c. 5000 words - notion to pdf

• An employee guide to progression - c. 5000 words - notion to pdf

• A guide on how sub levels work in the framework - notion to pdf

• A compensation philosophy - notion to pdf

• Our approach to promotions and reward - notion to pdf

• Full employee launch transcript - cleaned up using Google notebooklm

• Full employee guide to owning your career transcript - cleaned up using Google notebooklm

It’s a lot of content, in fact everything I build when running a project for a client.

Step 2: Define the instructions

Defining the GPT’s Tone & Capabilities

Now that we have the core inputs, I had to focus on:

• The tone of voice (e.g., supportive, direct, coach-like).

• The level of detail in responses.

• How the GPT should tailor responses based on employee level (e.g., early-career vs. senior).

I got deep into discussion with ChatGPT about the persona and detail of the approach - for example, I was inspired by a post from Natalie Lineton on linkedin, so I wanted it to focus 80/20 on strengths development. For the tone, I was interested to see what it came up with.

The persona should be that of a coach - focusing 80% of time developing strengths and 20% on weaknesses. It should encourage them to come to their own outcomes by asking questions. It should act as a co pilot for managers, helping employees own their career journey and their next career frontier in their specific function. I’ll let you define the tone, now please create a draft persona description.

Step 3: Bring it all together

First, I took my output from ChatGPT, which had put together a very detailed set of instructions. I had refined these first, before bringing them into the customGPT.

Then I added in my knowledge base, ensuring I clicked the code interpreter and data analysis checkbox to ensure it’s using my content.

Finally, if it were just for my company, I would make sure to select the right privacy settings - “only people with the link” before I published it.

Step 4: Off you go 📈

There we have it - building a customGPT end to end, combining our company knowledge as well as chatGPT’s capabilities with knowledge and coaching - check out the video below:

🤯 PS - I created all of the above in under an hour…

My 4 week, highly practical AI powered HR course to help you work like I do is in session, week one we focused on agentic approaches vs building an agent, week two, all about engagement and commercial data for action 📈

I want to help as many folks in the space build their skills as much as possible over the next year.

It’s critical to do so, as we’re at an inflection point. 20% of people will continue to grow because they use AI effectively, but 80% will stagnate.

I want to get as many people into the top 20% as possible - so they don’t get left behind.

🥳 April and May cohort sign ups open NOW - before I take a summer break

💡 Quick Win of the Week

So I had the chance to sit down with Ed Biden, a very well respected product leader (Wooga, Rocket Internet, Depop) to discuss how I built my little app to help people know what they should be paid.

I dive into my imposter syndrome, and the learnings I have found building with Lovable.dev in detail…

📰 News That Matters for HR

OpenAI Integrates Image Generation into ChatGPT 🎨

OpenAI has announced the integration of image generation capabilities directly into ChatGPT through its GPT-4o model.

GPT-4o enhances ChatGPT’s ability to create detailed and accurate images, including diagrams, infographics, and labeled visuals.

Notably, it excels in rendering text within images, a common challenge for previous AI models…. with many previous fails now fixed!

Why does this matter?

For HR professionals, this advancement offers a powerful tool to enhance internal communications, training materials, and employee engagement work.

The ability to generate clear and customised visuals can aid in creating more engaging presentations, instructional content, and organisational charts, thereby improving information retention and employee understanding.

Zapier Launches Model Context Protocol (MCP) Server for AI Agents 🤖

Zapier has introduced its MCP Server, enabling AI agents to connect programmatically to over 8,000 services integrated with Zapier without the need for individual service configurations.

This development allows for seamless integration of AI assistants with a vast array of tools, enhancing automation capabilities.

What is an MCP? I love this analogy…

Why does this matter?

HR departments can leverage Zapier’s MCP Server to automate routine tasks such as scheduling interviews, sending follow-up emails, and updating employee records across different platforms.

By integrating AI assistants with existing HR tools and services, teams can reduce manual workloads and focus more on strategic initiatives, happy days!

👋 Let's Connect!

Finding this valuable? There's more where this came from!

✨ Share this newsletter with your HR network

🔗 Connect with me on LinkedIn: https://www.linkedin.com/in/mattbradburn/ 

💌 Have questions? Just hit reply!

Remember: AI in HR doesn't have to be complicated. Let's figure it out together, one practical step at a time.