AI Bots in Service Organisations, Part 2: Banking & Finance
From Today's Chatbots to Tomorrow's Autonomous Agents
By Sandra De Zoysa, Founder & Managing Director, CEx Consulting
This is Part 2 of a four-part series looking at how AI bots are reshaping service organisations. Part 1 laid out the framework; here I focus on banking and finance, with telecommunications and insurance to follow. In each part, I'll map the bots already at work in the industry against the bots still on the horizon, and what that means for CX and boardroom strategy.
Banking has more automation infrastructure in place than almost any other service industry, decades of RPA, established core banking APIs, and mature fraud-monitoring systems. Yet that head start hasn't fully translated into agentic AI at scale. Most of what's live today still sits at the conversational layer; the shift toward autonomous, multi-step decision-making is only beginning to move out of pilot and into production.
The Bots Already Working in Banking
Chatbots remain the most visible layer, answering balance, product and FAQ queries on web and app. What's changed is capability: modern conversational banking now understands intent, not just keywords, and hands off routine servicing, payment initiation, card blocking, dispute filing, without a queue. Singapore's Trust Bank cut human support workload by 50% and complaints by 40% with a GenAI chatbot, redeploying staff into AI-oversight roles rather than cutting them.
RPA bots run quietly behind the scenes, reconciling accounts, processing KYC documents, copying data between core banking and compliance systems. RPA isn't being replaced by AI agents; it's becoming the reliable foundation those agents sit on. The hybrid model, RPA for the repeatable core, AI agents for the exceptions and unstructured data, is where most banks are finding ROI today.
Trading bots have been the banking industry's longest-standing algorithmic actor, executing rules-based buy/sell decisions on stocks, FX and, increasingly, crypto, at speeds no human desk can match. These remain tightly governed given their systemic-risk implications.
Web scrapers/crawl bots feed the risk and compliance function, pulling public filings, sanctions lists and adverse media for KYC and fraud checks, while social bots increasingly support brand monitoring and complaint triage on social channels, flagging reputational risk before it reaches a call centre.
The Bots on the Horizon
Autonomous agent swarms are the clearest near-term shift. Rather than one bot trying to do everything, banks are moving toward specialised agents working together, a triage agent, a fraud-investigation agent, a lending-decision agent, coordinating on a shared customer data model. This is what "agentic AI" means in practice, and it's forecast to grow from a $2.1 billion market to $81 billion by 2034. AI-orchestrated loan origination is already compressing decision cycles from days to hours while keeping every decision auditable for regulators.
Empathic AI companions are emerging in wealth and retail banking as "always-on" relationship agents, negotiating personalised products in real time, reading emotional cues in a conversation, and knowing when to step back and route to a human. The research is consistent: 91% of customers say AI isn't acceptable for certain issues, so the winning model isn't full automation, it's empathy-aware handoff.
Decentralised, blockchain-based bots sit further out but are worth watching, autonomous agents that could execute cross-border settlement or asset transfer directly on distributed ledgers. Regulators are already flagging the systemic risk of autonomous financial agents moving capital instantly between institutions on tiny pricing differences. This is a governance conversation the industry needs to have before the technology matures, not after.
Physical service robots have the least near-term relevance to retail banking, though they're worth watching in flagship branches and back-office document handling as a longer-horizon possibility.
What This Means for Boards
Three questions banks should be asking themselves:
● Do we have the core infrastructure, open APIs, a unified customer data model, to support agent swarms, or are we asking a lot to compensate for a legacy platform?
● Who owns accountability when an autonomous agent makes a decision that affects a customer's money?
● Are we re-skilling frontline staff into AI-oversight and exception-handling roles, or simply cutting headcount?
The institutions that will lead the next decade of service industries aren't the ones with the flashiest bot. They're the ones that got unified data, governance and explainability right first, and never lost sight of the human moments that still need a human.
Next in this series - Part 3: AI Bots in Telecommunications.
