BLOG

Your AI strategy runs on wiring you haven't modernized

Author | Laura Rodriguez
Laura Rodriguez
November 20, 2026
Share this article

Introduction:

Your bank has probably bought the appliances. AI tools, a modern data platform, a digital banking upgrade. The question for the next budget cycle is whether the wiring can run them. 

In most banks, it can't yet. AI needs current, trusted data from the core and a safe way to act on it. That data still moves through point-to-point integrations, nightly batch files and business rules nobody has written down. Until that changes, AI stays in pilots and the investment doesn't pay back. 

The good news: modernizing the wiring doesn't mean betting the core. It can be done in phases, each one paying for itself. 

AI strategy for banks: four questions before the next investment 

1. Can our AI see today's data, or yesterday's? 

If customer, lending and card data reach your analytics platform once a night, AI decisions run a day behind. Ask which data feeds an AI use case needs, and how current they are. 

2. Who can explain how a result was produced? 

Regulators, auditors and your board will ask. If nobody can trace the data behind an output, or show who reviewed it, the use case won't survive its first exam question. Governance has to be designed in, not added after launch. 

3. What is our team not doing while they keep old systems running? 

Every hour spent keeping integrations alive is an hour not spent on the roadmap. This is the hidden cost behind most stalled AI plans: not the technology, but capacity. 

4. Does the first phase fit a budget we've already approved? 

Large, multi-year programs are hard to approve and harder to steer. A first phase of 8 to 12 weeks, scoped to one use case and the systems it depends on, keeps the decision with the CIO and gives the board a result to look at. 

Where the value actually shows up 

For executives, the case for this work is in money and risk, not architecture: 

  • Added capacity without hiring. More of your current team works on what's next. 
  • Lower cost to run what you already own. Fewer manual workarounds and one-off fixes. 
  • Full value from platforms you already license. Snowflake, Databricks, Q2, Alkami and FedNow do more when the data reaches them. 
  • Fewer exam surprises. Documentation and controls come out of the work, instead of being rebuilt before every exam. 
  • AI that reaches customers. Use cases like lending document intake, contact center agent assist and fraud alert triage move from pilot to production.

Safety and control first, then speed

Banks are right to be careful with AI. The answer isn't to slow down. It's to build the controls into how the work is done:

•        Every AI output is reviewed by an engineer, security-scanned and tested.

•        A person on your side is always accountable for what reaches production.

•        Audit trails exist from day one.

•        Your vendor review happens before anyone gets access, under the 2023 Interagency Guidance on Third-Party Relationships.

What this looked like at one bank

For a bank with $55B in assets, we modernized 115 applications, cut the time to map the architecture by 90%, and delivered engineering work 60% faster. The starting point wasn't AI. It was understanding how the systems around the core connect, which is what any AI use case depends on.

How Devsu works

Devsu is a software modernization company headquartered in Orlando, FL and Quito, Ecuador. We help banks modernize the systems around their core and build new products, with AI built into how we deliver. Our engineers work in your time zone, with Velx, our AI-driven delivery approach, built into every engagement. We're ISO 27001 certified and SOC 2 Type II Compliant, with right to audit.

Start with the wiring

Before the next AI budget discussion, find out what your core and the systems around it can support today. We'll help you scope a first phase that fits a budget you've already approved. Learn more about how Devsu can help modernize the systems your AI depends on →