Gautham Pallapa, Scotiabank’s Principal Director of AI and Innovation, asks teams to define value and measurable success before deciding whether a project needs AI. On Practical AI with Chris Benson, he describes banking’s risk constraints through a personal boundary: he would not trust an agent with access to his own financial accounts and authority to transact. Customer assistance, compliance support, and helping advisers offer more bounded possibilities. His most concrete organizational example is a redesigned internship program. Stable teams tackle business problems with a technical lead, design lead, and product owner, then pitch executives for internal funding. When someone proposes an agent, Pallapa asks what it contributes; ordinary automation remains eligible. He also places developer experience upstream of code generation, in user stories, design, rapid prototypes, and customer feedback. That expands what a productivity claim would need to measure. His multi-agent sketch gives humans responsibility for the overall design and specifications, with specialist agents, shared memory, an orchestrator, and an evaluation jury checking progress. He describes experiments and explicitly acknowledges that he lacks the data to prescribe a definitive development method. The episode’s tension is between his concern for employee dignity and his claim that coding, documentation, and tests have become lower-value work. Generated artifacts still need review, and he concedes that generated code may lack enterprise quality. People-centered adoption needs evidence about accepted work, rework, and employees’ actual experience, alongside the business outcome. The interview offers a practitioner’s framework and experiments, not comparative proof of productivity gains.
“Just because you identify what the value is does not mean that you know what success means.” — Gautham Pallapa, passage at 10:50
“And, yeah, I don’t have the right data to say this is the way to do it, but we definitely are having a lot of fun.” — Gautham Pallapa, passage at 34:58
Based on the full publisher transcript; timestamps locate the quoted passages.
Listen to Practical AI: Building Human-Centered AI · Read the publisher transcript