Keynote, Conference Presentation
'The Future of Software With GenAI' by Marek Kalnik, Group Partner Theodo | RAISE Summit 2024
- Baseline for broader adoption is established by the sale of one million paid Copilot licenses, though the speaker notes this volume is "not that much."
- Global exposure to AI tooling is estimated to be widespread among developers, with a prediction that teams not using Gen AI daily are "lagging behind."
- AI is expected to deliver near-perfect correctness, approaching almost 100%, on specific tasks such as API calls, page generation, and data bootstrapping through prompt engineering.
- A milestone is predicted where AI can code software tickets with only "slight rework," supported by evidence of a 0.37 similarity score relative to human output.
- A new operational category is emerging where AI completes work under human supervision, a trend observed currently within the speaker's own company.
- Integration of AI into development workflows is anticipated to become autonomous, driven by sufficient success rates and trust, a shift expected to be widely adopted.
- The volume of human tasks is projected to decline slowly toward a plateau where the majority of tasks are handled by AI, though some human work will remain necessary.
- The development process is forecast to evolve into a "build platform" managed by a "product builder," where software teams focus only on difficult assignments or improving generated code.
- Platforms similar to v0.dev are expected to achieve success rates varying between 10%, 20%, or 80%, with the specific outcome presented as an open question.
- Most "AI engineering teams" for these platforms are predicted to be built in-house to address significant security investment and security issue concerns.
- The industry is expected to transition into a "two-tiered system" where AI collaborates with humans on distinct tasks, and companies treat "custom AIs" as special assets.
- A future possibility exists for businesses to achieve "personalized software" where platforms generate requests on demand, though this is framed as a potential outcome for "maybe one day."
- The industry is expected to undergo significant evolution over a period of "a few years," with the specific trajectory subject to future observation.