THE AI INTERVIEW
T
he honeymoon period of AI adoption is over. That’ s the view of Shayan Mohanty, Chief Data and AI Officer at Thoughtworks, a global technology consultancy that advises major brands on software and digital strategy. With Thoughtworks’ s major clients ranging from PayPal to Spotify, Shayan’ s role involves helping global giants to build and run AI systems safely – without disrupting existing technology or culture.
He maintains that, as most organisations reach the culmination of their experimentation phase, what comes next is a series of complex structural questions.
“ A huge amount of the challenge isn’ t actually technical – it’ s organisational,” Shayan explains.“ AI doesn’ t just change software; it changes how teams collaborate, how decisions get made, how risk is managed and ultimately how work itself gets structured.”
At Thoughtworks, this translates into building AI resilience. In other words, ensuring systems remain observable, governable and accountable once the initial novelty fades.
Understanding complex systems It’ s fair to say Shayan’ s route into the field was unconventional. His first ever job was running his own company( one he started in high school), which bought and co-located servers.
Post-university Shayan joined Facebook and led the team responsible for handling advertising metrics, which meant dealing with real-time data at“ massive scale”.
He even enjoyed a stint as a guest scientist at Los Alamos National Laboratory, the US government research facility, focused on high-performance computing. Before joining Thoughtworks, Shayan co-founded Watchful, a San Francisco startup that automated what he calls“ one of the least glamorous but most important parts of machine learning”: data labelling.
“ One thing people outside the AI industry often underestimate is how much of successful AI comes down to data quality, observability and evaluation, not just the model itself,” he says.
“ Looking back, every stage of my career has really been about understanding increasingly complex systems, whether that’ s infrastructure, data, organisations or now AI agents.”
The trap of persona cloning Shayan is clear on one of the biggest mistakes companies are making as they continue to restructure teams around AI output – and calls it“ persona cloning”.
He continues:“ Companies look at a specific human role and think,‘ Great, let’ s just train an AI agent to replicate exactly what this person does’. But that’ s a dangerous trap because generative AI doesn’ t behave like traditional software.
22 August 2026