Speak, the language learning app, is rethinking its software development with AI agents that collaborate to accelerate developers’ work.
For several years, artificial intelligence has gradually become part of developers’ daily lives: autocompletion, code generation, documentation assistance, etc. These tools have become real co-pilots. Today, AI goes further: it can perform complete tasks and collaborate in teams of AI agents, profoundly transforming software design.
At Speak, a company specializing in language learning using artificial intelligence, we observe this transformation every day. What was still an experiment a year ago is gradually becoming a new way of organizing the work of technical teams.
From the co-pilot to the team of agents
The first AI tools for developers have already enabled significant productivity gains, often of the order of 20 to 30%. The new generation of systems no longer just suggests code: they can analyze, code, test and document functionality. Several specialized agents can now work together on the same project, in coordination, which modifies the traditional role of the engineer: he becomes the conductor who organizes the agents and structures the context of their work.
The emergence of “agentic engineering”
This approach, which some call agent engineering, transforms several aspects of software development. The speed of experimentation increases: it becomes possible to test new ideas and produce prototypes in a few hours. Preparing codebases and documentation becomes essential, as agents must understand the project architecture and its conventions to be effective.
The boundaries between professions are also evolving. Product or design profiles can create prototypes or fix simple issues with the help of agents, while engineers focus on complex architectures and overall oversight.
A transformation still being explored
Despite this rapid progress, this is still an experimental phase. Tools are evolving at high speed and good practices are gradually being built. We organized a week dedicated to these new methods, to observe how teams can collaborate with agents. This initiative identified potential gains, but also challenges: the need for secure environments to run agents, the importance of a well-prepared code base, and the human supervision essential to guarantee quality.
Towards a new profession of developer
Software development is entering a new phase. The role of engineers is not disappearing, but it is evolving: key skills now go beyond mastery of a language or framework. It’s about knowing how to structure a problem, guide intelligent agents and supervise increasingly autonomous systems.
For engineering teams, the era of AI agents may just be beginning.
