A company is delivering breakthrough advances in artificial intelligence (AI) by moving to create prompts and consume APIs without the need for AI science expertise thanks to large-scale language models. I did. To improve the developer experience and create applications and tools, they defined and established principles around simplicity, instant access, security and quality, and cost efficiency.
Romain Kuzniak spoke at FlowCon France 2024 about enhancing the developer experience for creating AI applications.
Kuzniak said there was a huge gap in scaling initial AI applications to meet the needs of millions of users. This transition required Data to hire scientists, develop a dedicated technology stack, and overcome a number of areas in which it previously lacked experience.
Given our high costs and time-to-market, and our status as a startup, we had to carefully evaluate our priorities. There were many other opportunities that offered a higher return on investment. As a result, we have decided to pause this effort.
Kuzniak said the breakthrough in AI came with the advent of large-scale language models (LLMs) like ChatGPT, which changed the approach to leveraging AI. The main change brought about by LLM is that it significantly reduces the cost and complexity of implementation.
LLM reduces the need for data scientists, data cleansing, model training, and specific technical infrastructure. Now, creating meaningful engagements is as easy as creating prompts and leveraging APIs. No AI science expertise required.
Kuzniak said improving the developer experience is just as important as improving the user experience. Their goal is to eliminate any obstacles in the implementation process and ensure a seamless and efficient development flow. They envisioned their ideal developer experience with a focus on simplicity and efficiency.
We have established important principles when it comes to implementing AI.
- Simplicity: Implementation can be done in just one line of code.
- Instant accessibility: Allows real-time access to prompts with no introduction required.
- Security and Quality: Integrate security and quality control by design.
- Cost efficiency: Design cost controls and thresholds that are built into the system by default.
Kuzniak said organizational structures are evolving in the face of the technology landscape. A traditional cross-functional team of product managers, designers, and developers, while still appropriate, is not necessarily the best setup for an AI project. He explained:
Alternative organizational models should be considered. For example, the way information is structured and its impact on the quality of results potentially highlights the need for new team structures. For example, it may become more common to envision a team that includes an AI product manager, a content designer, and a prompt engineer.
Kuzniak advised applying the same level of dedication and best practices to improve the internal user experience as you would with your external customers. He said the mindset should shift to one in which team members consider their ideal user experience and actively contribute to its creation. We concluded that this approach not only increases efficiency and productivity, but also significantly increases employee satisfaction and retention.
InfoQ interviewed Romain Kuzniak about developing AI applications.
InfoQ: What does an AI application look like?
Roman Kuzniak: Our AI applications are diverse and focused on internal use, especially given our nature as an online school that generates substantive content. We prioritize making AI tools easily available across the company, especially by integrating them into familiar platforms like Slack. This approach allows staff to seamlessly leverage AI in their daily work.
Additionally, we have developed a prompt catalog. This initiative encourages employees to leverage their existing work and fosters an environment of collective intelligence and continuous improvement.
Externally, we have extended the benefits of AI to users, for example through the introduction of Student AI Companion. This tool is designed to enhance the learning experience by providing personalized support and guidance and help students navigate their courses more effectively.
InfoQ: What challenges are you currently facing with AI applications and how are you addressing them?
Kuzniak: Among the challenges faced in AI applications, the most important is resisting the temptation to implement AI for its own sake, especially when it adds little value to the product. Integrating AI capabilities just because they're trendy or technically feasible can take focus away from what's really important: the value these capabilities bring to your customers. We've all encountered products announcing new AI features, but how many of these truly improve the user experience or provide real value? Or?
Our approach to this challenge is rooted in fundamental product management principles. We constantly ask ourselves what value we want to deliver to our customers and whether AI is the best way to achieve that goal. If AI enhances our services in a meaningful way, we'll embrace it. However, if a different approach better suits your needs, we'll accept it as well.
