Suggestions
Greg Freeman
Executive Technology Leader | Experienced CTO | 2 Successful Exits
Greg Freeman is the Vice President of Engineering and Product Development at Teamworks, a role he holds while based in Sunshine Coast, Queensland, Australia. His professional background includes significant experience in engineering and leadership within the tech industry.
At Teamworks, Freeman is responsible for overseeing engineering initiatives and product development strategies, driving innovation and efficiency within the organization. His expertise contributes to enhancing the company's technological capabilities and product offerings, aligning them with market demands and user needs.24
Freeman is also active on LinkedIn under the username gfreeau, where he shares insights related to leadership and technology.13
Highlights
How Gen AI Helped Me Upgrade My Long-Standing Open Source Project
A year ago, I started working with large language models (LLMs) on a daily basis. One of my open source projects is a Pomodoro app, which I started 12 years ago. This app has been essential to me for time management, helping me create focus blocks and get the most valuable things done, even as an executive.
Here's what happened when I decided to bring it up to speed with the current standards:
A 12-Year-Old Codebase Gets a Second Life:
• My Pomodoro app, a favourite among some Linux distributions, had become quite popular, with contributions from 14 developers and translations into 16 languages (everything from Catalan to Hungarian).
• Built on an obscure JavaScript runtime that predates Node.js, it was in dire need of an update to ES6 standards.
Enter Gen AI:
• Over the past year, I've become quite proficient with various Gen AI tools—OpenAI, Claude, Llama, and even some mini LLMs running on my local machine.
• I thought, why not employ these AI models to refactor my code as a way to understand their real-world applications?
The Process:
• The LLMs provided me with accurate documentation, helping me quickly get back up to speed with the app's event-driven structure.
• They updated the code to use ES6 classes without errors, transitioning from the original prototype inheritance seamlessly.
• Despite some context size limitations with certain models, Amazon Bedrock came to the rescue, allowing me to handle everything in one go.
User Requests and Overcoming Procrastination:
• Addressing long-standing user requests (like auto-starting the timer on login) became a breeze. I was able to take open feature requests, ask the LLM how it would implement them, and it was mostly right.
• One of the surprising benefits of using LLMs was overcoming procrastination. Even if the initial suggestions weren't perfect, they sparked ideas and helped me get into a flow state quicker.
Reflections and Takeaways:
• My programming journey spans 20 years, but even with less hands-on coding in recent years, the LLMs helped me get back on track swiftly. I felt like I didn't miss a beat.
• In recent years, I've been mainly coding in Python and R for data science and analytics to support my executive business cases and make informed decisions. LLMs helped me quickly translate that knowledge into other languages I used more in the past.
• I estimate that my efficiency increased by 30-50% using LLMs.
Final Thoughts:
• I don't feel LLMs will replace engineers; I had to be in control and tweak things. I ran into issues with the context window where I knew the LLM output I was getting was wrong. Coding knowledge was still important. But once I got the setup right, I was able to produce code at a much higher clip than I would have otherwise.
• Every engineering organisation should make these tools available to their teams. There are obvious benefits.
Want to avoid bad AI investments? Start with strategy.
The hype around AI is deafening, and I can't help but think about all the dollars being burned right now to avoid FOMO.
But let's not let FOMO lead us into making impulsive investments. The key to success with AI lies in strategic adoption.
It starts at the top. Company leadership must define a clear AI strategy that directly supports broader business goals. Once you have that, you can identify where you want to play on the AI field such as:
• Employees Powered Up by Copilots: Tools that increase efficiency and improve decision-making. Example 1: Making sales and support more efficient by serving up technical product answers on the fly without needing to involve tech or sales engineers. Example 2: Engineers using coding co-pilots. Example 3: Meeting summaries and notes.
• AI-Fueled Insights: Harnessing AI's power to analyse data for deeper business understanding driving business analytics. This has been around for a long time, but it's now much more efficient, cheaper, and accessible due to the availability of LLMs. You can gain insights with minimal training, delay, and without needing extensive machine learning expertise on staff. Example: Customer sentiment analysis on large amounts of reviews, surveys, or other unstructured data you have as often as you like, whereas before these insights were stale or costly.
• AI-Powered Products: Creating products or enhancements with AI at their core – but avoid building just for the sake of it. Ensure a real market need exists.
Pitfall: It's easy to fall into the trap of "build it and they will come". Remember, AI-powered products need careful planning and should be subject to the same (or more) rigorous customer discovery process as any product release. Contingencies you need to account for include graceful degradation if the model is unavailable or changes due to upgrades or retirement, the output changes, data sovereignty guarantees, and the pricing you can absorb. It's also critical to log all model input and responses for debugging as needed. All of these things need to be considered on your roadmap, often you need multiple models available as backups, and these all require testing.
Recommendation: For most companies, focusing on employee-focused copilot tools is a great starting point, provided it aligns with your strategy. Is your customer support team struggling with efficiency? A co-pilot solution can help achieve clear goals in that area. Next, prioritise an AI insights initiative – identify one impactful area of your business where insights can unlock tangible results. Leave AI-powered product initiatives until you've thoroughly validated customer needs and the unique value you can provide. Validate early product ideas in AI playgrounds to avoid costs.
Focus is essential. Pick the area that best aligns with your strategy (you have one right?) and gain momentum with tangible results.
#leadership #techleaders #AI #strategy
