AI and automation: Urgent solutions to productivity
Automation has been on the rise for a long time, and the AI boom has brought it to the attention of customers as well. It’s clear to all ISVs that the most common business model — selling developers by the hour — won’t be sustainable for much longer. Customers are no longer willing to pay for coding as they used to, and soon, automation will be a necessity just to break even. While long-term profitability will require rethinking the business model, the short term requires us first to fix the immediate productivity problem.
As coding is quickly becoming a commodity, many are turning to AI-powered coding assistants like Copilot or ChatGPT to reduce development time and increase productivity. These tools provide immediate, short-term gains by automating individual coding tasks. However, a more profound solution to productivity lies in going beyond the individual tasks, and automating the development process itself. As ISVs evaluate how to respond to falling prices for manual coding, understanding the productivity benefits of both AI and development automation is key. Both approaches offer significant productivity increases, but they differ in scope and long-term impact.
Generative AI for quick wins
Tools like Copilot and ChatGPT can boost developer productivity on code generation and documentation. These allow developers to focus on higher-level problem-solving, but their impact is mostly limited to coding challenges. Generative AI is highly effective for quick wins in short-term gains, but it doesn’t transform the entire development pipeline.
Automation for deep impact
For a more significant boost in productivity, you automate the entire system development process, minimizing the time spent on writing boilerplate and implementing features repeating in every business application. Automation ensures consistency, scalability and defect-free implementations across projects. By automating the development process itself, ISVs can not only bring more value to customer, but also speed up delivery, reduce technical debt and future-proof their applications, driving even greater long-term productivity and sustainability.
Comparison
| Topic | AI (Copilots, ChatGPT, etc.) | Development process automation (LeBLANC) |
|---|---|---|
| Scope and complexity of output | Assists with individual tasks: Produces code snippets or suggestions; automates coding tasks, varying in complexity. | Automates the entire system development process. Handles system creation end-to-end with consistency across projects. |
| Level of abstraction | Works with high-level instructions and abstract concepts. Suggests code based on natural language or simple prompts, but requires the user to handle the low-level details. | Operates at a more advanced abstraction level by automating entire workflows and system generation based on models. Reflects the rising abstraction in software engineering, from machine instructions to models capturing business logic. |
| Consistency of output | Exhibits noticeable variance when responding to different instantiations of the same question. | Produces consistent results; the same task is solved in the same standardized way. |
| Error handling | Makes errors, requires human review and correction. | Code generation is deterministic, flawless transformation from models to output. |
| User involvement | Requires human interaction and guidance. | Largely hands-off once set up. |
| Speed of adoption | Quick to adopt as integrates with existing manual conventions and provides immediate suggestions or code generation. Easier to accept as it feels like a personal solution, since AI generates code similar to what users might write | Slower to adopt, requiring a learning curve for utilizing full-scale automation. |
| Future-proofing and maintenance | User has to manually update and integrate new practices | "Evergreen" applications, automatically updating to current industry best practices, reducing technical debt by automatically applying updates and improvements on regeneration. |
| Improvements | Improvements based on new, potentially better LLM models. | Improvements based on emerging industry best practices, patterns and conventions in creating business applications. |
| Productivity increase | Can increase productivity by 20-30%, mostly for smaller or specific tasks. | Over 80% of any line-of-business application can be generated from models, leading to a 50-70% productivity increase by automating most of the development process. |
The recipe for productivity
Utilize both!
LeBLANC automated software production line produces roughly 80% of your application code. Use this to enhance your development process efficiency, and then use AI tools for the final 20% customization phase - to implement the part that makes your application unique.
Best of both worlds.
Afterthoughts
"The entire history of software engineering is one of rising levels of abstraction"
Grady Booch (X, 4.7.2024)
From machine code to high-level languages, and now to software production lines
and model-driven development.
Our development automation embodies this shift, allowing ISVs to move beyond just writing code and
into generating entire systems from models. This approach not only accelerates development but
ensures consistency, reduces technical debt, and keeps applications "evergreen" by automating
updates and adhering to best practices. In contrast to AI, which provides individual coding solutions,
automation leverages these higher abstraction levels to transform the entire development process,
making it the more sustainable answer to ISV productivity dilemma.