Beyond the Productivity Trap: rethinking work, choices and value creation with Agentic Coding

Code is no longer the scarce resource. When an agent writes the code, other capabilities become scarce: understanding what is really needed, knowing the whole system, knowing why the software works, deciding who approves what. Organisations that change their tools without changing how they work risk seeing the volume of code grow along with verification work and risk. Roles and professional profiles are changing too: developers spend more time ideating, specifying, designing and verifying, while those who lead information systems take on new skills and responsibilities. During the evening we will present a method for deciding, controlling and taking accountability for results when agents are part of the team, and for supporting people through this change. 

Control, governance and security. More agents in the team must not mean less control or less security. During the evening we will look at how to match autonomy and controls to the risk of each change, defining who decides, who verifies and who approves; how to update metrics, since with agents the number of tests and of passed tests can grow without quality improving; how to make quality, security and costs measurable; and how to keep track of what is produced and decide which models and data agents may use. Security is built into the method from the start, with automated checks on every change, least-privilege permissions for agents, data protection and human review of critical steps. 

The software you buy needs rethinking too. If building a feature keeps getting cheaper, the trade-off between buying, building and modernising changes. According to McKinsey, 32% of organisations that regularly use AI have already dropped at least one software purchase in favour of building it with Agentic Coding. Keeping software reliable over time, however, remains a separate cost: quality, security, maintenance, skills. This is what separates a fast project from an investment that lasts. 

Three decisions for those who manage the application portfolio: 

→ What to buy, what to build and what to assemble, and with which partners. 

→ Where and how to modernise existing applications, including by introducing agents that work under supervision. 

→ How to measure results beyond development speed. 

Our point of view. adesso.it is a technology-independent software factory specialised in application modernisation. Using agents, it delivers quality software, iterates quickly and helps clients evolve and govern their information systems, choosing for each task the AI model best suited in terms of quality, cost and risk. We help CIOs decide where Agentic Coding delivers measurable results, even when the best choice is to build nothing new or not to modernise at all.

POLIMI Graduate School of Management replaced the learning platform it used to buy as a service with Grow, a blended-learning platform built around its own needs, developed with Agentic Coding together with adesso.it and now running as a pilot. It is a concrete example of collaboration between client and software factory: the School made the decisions and governed the project, while adesso.it brought the method, the development and the verification. 

Luigi Gangitano describes the business domain, how the project was conceived and how the School oversaw and governed its development. Stefano Bruna presents the method for deciding, verifying and taking accountability for results when agents are part of the team. 

A virtuous partnership between client and software factory. If agents write the code, it is tempting to think you can build your own in-house software factory. What makes software reliable over time, however, is expertise in architecture, verification and security, a method proven across many projects, and the ability to keep it up to date while tools change every quarter. During the evening we will look at how to build a virtuous partnership: the client keeps the decisions, the governance of its information systems and knowledge that remains usable even when changing partner or tool; the software factory brings the ability to deliver quality software, iterate quickly and support the organisation as it evolves. 

A two-way evening. 

→ Each session leaves room for questions from the audience. 

→ During the evening, two short anonymous surveys on your phone: first your expectations and how you work today, then the indicators you use to evaluate code written with agents. Results are shown immediately and we discuss them together. 

→ The conclusions start from your questions and the survey results. 

What you will take home: 

→ Stefano Mainetti's presentation «When code stops being scarce». 

→ A one-page case summary, with the method and lessons learned. 

→ Insights on technical debt in the AI era, drawn from two recent articles by Stefano Mainetti and Massimo Ficagna.

CIOs, application managers and IT decision-makers at companies that are introducing Agentic Coding into their teams or assessing how their application portfolio is changing. No programming skills are required.