Agents Write Code. How Do We Know Development Has Become More Effective? | Code of Leadership (#AI4SDLC)
In 20 minutes, at 17:00 Moscow time, we will meet Vladimir Dolgov on the Code of Leadership livestream. We will discuss developer productivity and how development changes when AI agents join the team. Suppose an agent writes the code quickly. Then someone reviews it, asks for changes, integrates it into the system, and ships it to users. I want to account for that entire journey: how much time and effort it took, what we got out of it, and where the speedup actually happened. A token counter will not answer those questions on its own :)
Vladimir Dolgov is a technology and product leader with more than 20 years of experience managing teams and major projects at the CxO level in finance and telecommunications. At Sber, he led the AI transformation of the development process, introduced AI assistants, and developed Developer Experience products: GigaCode, GigaIDE, GigaView, and API/Function Studio. Previously, he built a CRM/MDM practice from scratch, managed distributed teams of up to 430 people, and implemented some of the largest Oracle Siebel CRM systems. A mathematician by training, he graduated from Saint Petersburg State University.
Vladimir and I will explore several big topics: 🔸 How development tools are evolving IDEs, CLIs, background agents — how working habits change, what is worth building in-house, and how to decide whether a new tool deserves a place in the team. 🔸 What we mean by developer productivity An individual engineer’s speed, work moving through the team, quality, value for the business — how to connect these and understand what actually improved. 🔸 How agentic development works From defining a task and preparing context to testing, reviewing, and shipping. What can we delegate to an agent, where do we need a human, and how much does checking the result cost? 🔸 How to measure AI’s impact What to compare before and after adoption, how to account for rework, and how to distinguish a tool’s contribution from changes in team composition or task complexity. And what to do if PRs appear faster but releases still happen at the same rate.
There will also be a DevOptify (https://devoptify.ru/) demo — the platform Vladimir is building to assess developer productivity. We will look at how engineering data leads to conclusions and management decisions. We will also discuss the limits of metrics: what they can show, where they lack context, and how a developer can challenge an assessment of their work.
Join us if you are adopting agents, developing tools, or responsible for your team’s results. Bring your questions and stories about what happened to your team and product after you learned to write code faster.
#AI #Agents #Engineering #AI4SDLC #Software #Knowledge #Data #DevEx #DeveloperProductivity #Management #Leadership