Coinspaid Dev’s Alexey Tulia on Who Answers for AI Agents in Production

As AI agents move from drafting code to acting inside live systems, responsibility for their decisions is turning into a leadership question. Alexey Tulia, Executive Leader at Coinspaid Dev, believes that responsibility still belongs to people.

As reported by DEV Community, Tulia shared his view during the AI Impact in Engineering panel at Tech Race Summit 2026 in Warsaw, where speakers examined how artificial intelligence is reshaping the work of engineers and CTOs. His remarks focused on the next phase of adoption, when AI systems gain the ability to take actions through a company’s own infrastructure instead of only suggesting them.

Most businesses today treat AI as an assistant that drafts documents, summarizes information and speeds up analysis. The shift Tulia describes goes further. Agents connected to production environments could reach sensitive data, trigger deployment pipelines and change live services, and that level of access forces organizations to decide who authorizes each action and who answers for the result. “The more authority we give machines, the more important accountability becomes,” Tulia said. He pointed to a practical scenario: an agent capable of preparing a code change and shipping it to production on its own. Should it be allowed to release without human sign-off? And if the release breaks something, whose responsibility is that? In Tulia’s view, a company should have a set of safeguards in place before it grants such access:

• permission controls that define exactly which systems the agent can touch;
• audit logs that record every action it performs;
• a reliable way to halt the agent while it is working;
• recovery procedures that allow a failed deployment to be rolled back.

The more freedom an agent has in production, the more clearly its authority and the human accountability behind it need to be defined.

Tulia also addressed where engineering budgets should go. He urged CTOs to link AI spending to a specific organizational need and to avoid buying tools for their own sake. The foundations he prioritizes are well-designed APIs, dependable data, automated testing, observability, security and flexible architecture, because together they let a company bring in new technology without putting its operations at risk. Spare capacity matters as well. When a roadmap fills every sprint, a team has no room to trial an emerging tool or react when priorities change. Investments in architecture and in reducing vendor lock-in rarely show up in short-term revenue, yet they lower the cost of swapping a provider or reworking a system once earlier assumptions stop holding. “I don’t need to predict the future perfectly. I need to make being wrong cheap,” Tulia said.

The role of the individual engineer is changing too. AI already accelerates coding and prototyping, and Tulia sees that as a chance for engineers to spend more time understanding the business problem behind a task and to stay involved until their work is running in production. Leaders can encourage this by sharing business context with their teams and setting a clear expected outcome for each piece of work. Productivity then becomes a matter of correctness, maintainability, security and operational performance, while the sheer volume of code written stops being a useful measure.

Looking ahead to 2029, Tulia expects smaller engineering teams to take ownership of wider areas of responsibility. He also anticipates that AI will write most production code, which will make verification and technical judgment more valuable than they are now. The CTO position, he argued, will keep demanding deep technical expertise combined with a solid grasp of the business, especially as easier software creation brings more vendors and AI-generated systems into every organization. “I think technical judgment becomes even more important,” Tulia said. For engineering leaders, the practical conclusion is to settle safeguards and ownership before AI agents are given access to critical production systems, since fixing those gaps after an incident is far more expensive.

Tulia’s perspective is shaped by the work of Coinspaid Dev, an independently owned and operated software engineering company specializing in blockchain infrastructure development. The company employs more than 120 engineers and has over 11 years of industry experience. Its software engineering, infrastructure, security and R&D teams have built distributed systems and blockchain infrastructure that operate across more than 20 blockchain networks, an environment where reliability and clear accountability are part of everyday engineering.