Agentic AI set to transform tech roles, McKinsey warns
Agentic AI is poised to shift software engineering from manual coding to high-level orchestration, which is set to transform tech roles according to a McKinsey executive.
Speaking at Viva Tech Paris on Thursday, Stéphane Bout, the firm’s France leader and senior partner at QuantumBlack, AI – McKinsey’s data and analytics branch – said the software industry is on the cusp of a fundamental transformation.
“Agentic AI can plan, execute, test and debug code with minimal human intervention”, he said. “We’re moving from assistive tools to collaborative agents”.
Agentic AI refers to autonomous software systems that can reason and make decisions, acting independently of traditional software development lifecycles. This makes them far beyond today’s code autocomplete tools.
These systems are capable of goal-oriented behaviour; they can take a user’s high level objective and automatically break it down, generate code and fix it, test it, and so on.
Often, this is even done without direct step by step instructions.
A shift in the engineer’s role
This evolution is happening at a rapid pace, said Eiso Kant, co-founder and chief tech officer of Poolside, at the event.
“We’re entering an era where writing code might no longer be the primary task of a software engineer”, he said. “In the near future, engineers will help design goals and set constraints, as well as supervise intelligent agents that handle the execution”.
Bout added that this shift will generate a high demand for new skills: “The engineer of tomorrow is more of a system designer and strategic thinker than someone focued on syntax or boilerplate code”.
The result could be significant productivity gains. “Agentic AI opens the door to tenfold engineering productivity”, said Brout, who assured that it is not about replacing engineers, but removing the bottlenecks that slow them down.
Devine Pasta, chief executive of software at Siemens, pointed to real world applications already taking shape in infrastructure and industrial environments.
“In mobility and infrastructure, agentic AI could reduce downtime proactively fixing code before issues arise”, she said. “It’s not just about speed, its about reliability”.
But the speakers cautioned that agency must also come with oversight, pointing to a vast difference between code completion and decision making autonomy.
“If agents are acting independently, we need to ensure they are operating within trusted guardrails”, added Kant.
The panel concluded that the future of software lies in delegation – not of control, but of complexity.