Close Menu
The Financial News 247The Financial News 247
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
What's On
How Managing AI Costs Starts With Where It Is Used

How Managing AI Costs Starts With Where It Is Used

July 10, 2026
BTS’s Latest No. 1 Album Passes One Of The Band’s Biggest Releases

BTS’s Latest No. 1 Album Passes One Of The Band’s Biggest Releases

July 10, 2026
Comcast eyeing acquisition of cable rival Charter

Comcast eyeing acquisition of cable rival Charter

July 10, 2026
How Banks Can Deploy AI Properly

How Banks Can Deploy AI Properly

July 10, 2026
Trump Won’t Sign Bipartisan Housing Bill In Protest Of Voter ID Law Stalemate

Trump Won’t Sign Bipartisan Housing Bill In Protest Of Voter ID Law Stalemate

July 10, 2026
Facebook X (Twitter) Instagram
The Financial News 247The Financial News 247
Demo
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
The Financial News 247The Financial News 247
Home » How Banks Can Deploy AI Properly

How Banks Can Deploy AI Properly

By News RoomJuly 10, 2026No Comments5 Mins Read
Facebook Twitter Pinterest LinkedIn WhatsApp Telegram Reddit Email Tumblr
How Banks Can Deploy AI Properly
Share
Facebook Twitter LinkedIn Pinterest Email

Naveen Jayaram is a Senior Technology Leader at UniCredit driving global product and technology innovation.

Banks share a growing desire to use AI to become faster and more innovative. Yet, for many organizations, the real challenge isn’t knowing why to use AI; it is knowing how. Too often, institutions struggle to get AI projects past the testing phase.

Senior technology leaders face a relentless need to launch faster and outpace the competition, all while relying on legacy technology. However, treating legacy stability and AI innovation as an either-or choice is a mistake.

Instead of beginning a risky, multiyear project to replace older systems, executives should adopt a “coexist” mindset. In my experience, it’s best to add a smart layer of AI on top of your existing processes to connect and orchestrate your systems safely.​​

The Integration Challenge

Organizations need to find the specific integration points where AI delivers measurable value without sacrificing existing controls, auditability or operational resilience.​

To do this, you need to embrace layered coexistence. Older, rule-based (deterministic) systems should continue handling strict math and logic, while AI can start managing reading, writing, exception handling and data orchestration. AI is not a universal fix; it should only be deployed where it is proven to work best.​​

Building The Coexistence Stack

When tackling operational overhead, searching for places to force AI into your workflows is a trap. Instead, try asking, “What core operational problem are we actually trying to solve?” Foundational systems do not require reinvention; they simply need capable neighbors.

Almost every bottleneck boils down to a distinct action owned by a specific layer:

• Strict Decisions (Rules Engines And Risk Models): Credit approvals, sanctions screening and capital calculations require mathematically consistent, auditable outcomes. AI should be kept out of making these decisions. Instead, AI can translate complex alerts from these systems into simple English so humans can review them faster.

• Predictive Analytics And Forecasting (Traditional Machine Learning): Traditional machine learning remains the best tool for predicting historical patterns like fraud probability, liquidity needs or customer behavior.

• Reading Documents (OCR And Document AI): Basic scanning tools break when document formats change. Pairing them with generative AI allows the system to read, understand and organize messy documents so older systems can process them easily.

• Routine Automation (RPA): Basic bots are highly effective at repetitive tasks. Adding agentic AI as the “brain” can help these bots when something unexpected happens by analyzing exceptions and dynamically instructing the bots on how to proceed when a process deviates.

• Connecting Workflows (Agentic AI): To solve the issue of employees switching between multiple screens, AI agents can be used to gather information across different departments and organize it automatically, bypassing the need for massive data overhauls.

• Drafting Documents (Generative AI): AI can efficiently write standard summaries, memos or reports. However, a strict boundary must be enforced: AI can write the draft, but a human must always review and approve it.​

The Model In Practice

Take a standard anti-money laundering (AML) investigation, for example. The legacy rules engine still triggers the alert. That deterministic process does not change.

After the alert, however, an AI agent can take over. It can pull data from various internal systems and write a quick summary before a human even opens the file. This can compress hours of preparation into minutes. The employee can focus on making a judgment rather than hunting for data. This same pattern can apply to customer complaints, loan approvals and background checks.​

Architectural Mandates

To safely connect AI with older systems, organizations must upgrade how their software communicates. Senior tech leaders should enforce these key architectural mandates to scale safely:

• Event-Driven Architecture (EDA): Instead of systems constantly polling for updates (which wastes compute power), organizations should use an EDA. This means a system simply publishes an event when it finishes a task, and the AI reacts instantly.

• The Model Context Protocol (MCP): MCP should be adopted as a standard plug-in for AI. This can prevent vendor lock-in and allow institutions to swap AI models easily without breaking legacy connections.

• Smart Failsafe (HITL): If an AI model hits a low-confidence threshold, it should not guess. The system should be set to flag the issue and route it straight to a human expert for a decision.

• The Emergency Kill Switch: Maintaining system control is nonnegotiable. Architects must build a central kill switch that instantly cuts off the AI if an anomaly occurs. The system should automatically undo any partial mistakes and route tasks back to manual queues to keep data safe.​

The Coexistence Imperative

Rather than replacing core systems, leaders should add AI directly on top of existing infrastructure. I’ve found that this overlay approach is generally safer, faster and more cost-effective than a massive technology overhaul.

Executives should focus on matching the right tool to the job: Let legacy systems handle strict rules, while AI handles reading, writing and organizing. These layers should be connected smoothly using modern methods like EDA and MCP.

Establishing clear human-in-the-loop pathways and embedding an emergency kill switch can help ensure that your organization can pause AI operations instantly and protect data integrity when anomalies occur.​​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Naveen Jayaram
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email

Related News

How Managing AI Costs Starts With Where It Is Used

How Managing AI Costs Starts With Where It Is Used

July 10, 2026
The No. 1 Habit That Destroys Deep Focus, By A Psychologist

The No. 1 Habit That Destroys Deep Focus, By A Psychologist

July 10, 2026
iOS 27 Locks In September Schedule

iOS 27 Locks In September Schedule

July 10, 2026
Do Your AI Models Know What Day It Is?

Do Your AI Models Know What Day It Is?

July 10, 2026
Aurora Possible In 20 States This Weekend

Aurora Possible In 20 States This Weekend

July 10, 2026
‘Escape From Tarkov’ Launches New Expansions Hub

‘Escape From Tarkov’ Launches New Expansions Hub

July 10, 2026
Add A Comment
Leave A Reply Cancel Reply

Don't Miss
BTS’s Latest No. 1 Album Passes One Of The Band’s Biggest Releases

BTS’s Latest No. 1 Album Passes One Of The Band’s Biggest Releases

News July 10, 2026

Since its release in late March, Arirang, BTS’s comeback album, has been a massive commercial…

Comcast eyeing acquisition of cable rival Charter

Comcast eyeing acquisition of cable rival Charter

July 10, 2026
How Banks Can Deploy AI Properly

How Banks Can Deploy AI Properly

July 10, 2026
Trump Won’t Sign Bipartisan Housing Bill In Protest Of Voter ID Law Stalemate

Trump Won’t Sign Bipartisan Housing Bill In Protest Of Voter ID Law Stalemate

July 10, 2026
Stay In Touch
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Vimeo
Our Picks
Delta CEO warns higher airfare costs could last even as oil prices fall

Delta CEO warns higher airfare costs could last even as oil prices fall

July 10, 2026
The No. 1 Habit That Destroys Deep Focus, By A Psychologist

The No. 1 Habit That Destroys Deep Focus, By A Psychologist

July 10, 2026
Christopher Nolan Addresses ‘The Odyssey’ Online Backlash

Christopher Nolan Addresses ‘The Odyssey’ Online Backlash

July 10, 2026
iOS 27 Locks In September Schedule

iOS 27 Locks In September Schedule

July 10, 2026
The Financial News 247
Facebook X (Twitter) Instagram Pinterest
  • Privacy Policy
  • Terms of use
  • Advertise
  • Contact us
© 2026 The Financial 247. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.