Citi Unveils AI Wealth Advisor ‘Citi Sky,’ Signaling a Shift From Relationship Banking to...

Citigroup is moving artificial intelligence from the back office to the front line of client engagement, introducing an AI-generated wealth advisor that could begin to redraw the contours of private banking.

The tool, branded "Citi Sky," is being positioned not as a novelty, but as an early expression of a different operating model-one built on continuous interaction, data depth, and scalable advice.

Set for a phased rollout this summer, Citi Sky will allow clients to query markets, portfolios, and financial scenarios through a human-like digital interface. Andy Sieg framed the initiative in expansive terms, saying it will "change the model of wealth management." That assertion underlines a deeper shift underway: advice is gradually being unbundled from the individual advisor and embedded into systems that can operate around the clock.

The economic logic is difficult to ignore. Traditional wealth management is constrained by time and headcount, with advisors typically managing a finite number of relationships. AI introduces a multiplier effect. By automating research, drafting recommendations, and maintaining client engagement between meetings, firms can expand coverage without proportionally increasing costs. This has direct implications for margins in an industry where profitability is closely tied to assets under management and advisor productivity.

However, Dipendra Malhotra pointed to a core technical constraint that continues to limit the deployment of AI in high-stakes advisory roles: memory.

"One is short-term memory: how long can you have this conversation before you start hallucinations?" he said, highlighting the instability that can emerge in extended interactions.

In a financial context, even minor inconsistencies can erode trust.

The more consequential challenge lies in persistence. "The second is the ability to have long-term memory, and that's pretty much all conversations: all the clicks, all the things which we know about our clients, transactions," Malhotra said.

Wealth management is cumulative by nature. Advisors build a layered understanding of clients over the years, incorporating behavioral patterns, risk tolerance, life events, and shifting priorities. Replicating that continuity in AI systems requires not just data storage but context-aware retrieval and interpretation.

Malhotra described the objective succinctly, saying: "That's the Nirvana." Systems capable of sustained, context-rich engagement would allow advisors to oversee larger books of business while maintaining a semblance of personalization.

The emphasis, he added, is on "productivity and scale," a formulation that captures the industry's direction of travel.

Citi's decision to maintain hiring plans for human advisors suggests the bank is proceeding cautiously. Rather than displacing relationship managers, the technology is intended to augment them-handling routine queries, surfacing insights, and preparing analysis, while leaving judgment and client trust anchored with humans. This hybrid model may prove durable, particularly as regulators scrutinize the use of AI in fiduciary roles.

The infrastructure underpinning Citi Sky points to the scale of investment required to compete in this space. The system is being developed in partnership with Google Cloud and Google DeepMind, indicating reliance on advanced large-model architectures and high-performance computing. More broadly, Citi is accelerating its technology spend. Chief executive Jane Fraser said earlier this year that generative AI tools have already saved developers 100,000 hours per week through automated code reviews, while the bank committed $2.3 billion to technology and communications in the first quarter of 2026.

Across the industry, similar deployments are taking shape. Bank of America has introduced AI systems that assist advisors before, during, and after client meetings, while other institutions are embedding generative models into research, compliance, and portfolio construction workflows. The common thread is a reallocation of human effort away from repetitive tasks toward higher-value advisory functions.

Still, the risks are material. AI-generated outputs introduce questions around accountability, particularly if recommendations influence investment decisions. Data governance is another pressure point, as these systems rely on extensive client information to function effectively. In cross-border contexts, regulatory regimes differ sharply, complicating deployment at scale.

There is also a competitive recalibration underway. As AI lowers the cost of delivering personalized financial insight, the traditional advantages of large institutions, distribution, brand, and balance sheet, may be challenged by more agile, technology-native entrants. At the same time, incumbents like Citi retain an edge in trust, regulatory experience, and access to proprietary client data.

Citi Sky sits at the intersection of these forces. It is both a tool and a test case: a measure of how far AI can be pushed into client-facing roles without undermining the foundations of wealth management.