FEATURE Industrial AI
THE RISE OF AGENTIC AI: AUTONOMY IN AUTOMATION
THE RISE OF AGENTIC AI: AUTONOMY IN AUTOMATION
Dijam Panigrahi, Co-founder and COO of GridRaster, says we are entering the next phase of robotic automation, where robots no longer simply complete repetitive tasks, but can begin to take on some of the decision-making
T
he industrial robotics landscape is undergoing a fundamental shift from robots that simply follow instructions to those that decide what to do next. For the past decade, collaborative robots (cobots) have been viewed as tireless helpers for scripted, repetitive tasks, while humans retained all meaningful decision rights. capped the potential of automation because every exception required human intervention. Agentic AI is now eroding that ceiling by granting robots bounded autonomy over micro-decisions.
How robots learn to think This evolution is driven by two primary advances: learning from video and language. • Video Learning: Robots are trained by watching high-skilled operators. Vision models map human motions and outcomes into machine-understandable patterns. Rather than just replaying a trajectory, correlate with the correct next action. • Language Learning: Large Language Models (LLMs) and Vision Language Models (VLMs) ingest the same 200-page manuals and work procedures used by technicians. The AI layer consumes this documentation directly to infer rules, such as acceptable tolerances and defect taxonomies.
By combining these, robots become
grounded in both how humans actually work and how a process is supposed to work on paper.
The autonomous inspection loop autonomy is in inspection, a sector that
18 March 2026 | Automation
is data-rich and historically under-automated. In complex robots can now: • Capture high-resolution visual and depth data.
• Classify defects against standards encoded from manuals and human judgment.
• Decide if a non-conformance is acceptable, reworkable, or scrap. • Close the loop by autonomously generating and inserting task orders into repair queues.
For instance, if a weld on an aircraft frame is out of tolerance, the robot doesn’t a digital work order for a technician or downstream cell. This transforms inspection from a passive gate into an time yields and more stable schedules.
The human boundary Despite these gains, humans remain essential for complex process decisions. Expert welders still synthesise subtle cues – like the sound of an arc or the feel of heat – that have never been fully documented or labelled for AI training. Current systems also struggle with novel scenarios, documentation. The near-term equilibrium is a
well-bounded domains, while humans
Executive priorities for ROI To capitalise on agentic robots, leaders must view this as a decision-rights transformation rather than a hardware refresh. Key priorities include: 1. Building a Digital Backbone: Autonomy depends on access to 3D models and progress. 2. Capturing Expert Knowledge: Systematically record expert decisions in video and data form to provide ground truth for future models. 3. Redesigning Roles: Human work must shift toward oversight and exception handling, with KPIs focused on quality stability and faster recovery.
Executives who move early will not just the decision fabric of their operations. In an era where resilience, quality, and speed are most consequential automation upgrade of the next decade.
GridRaster
www.gridraster.com
automationmagazine.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40
