The AI Planning Problem: Banks Know They Need AI… But Don’t Know Where to Start
Banks are accelerating AI investment, but many still lack the planning structure, analysis, and sequencing needed to turn ideas into workable projects. This article explores why the strategy‑to‑execution gap persists and what leaders can do to build the foundations for AI that actually delivers.
Olivia Beale
AI Delivery Consultant
olivia.beale@caspianone.co.uk
If you work inside a bank right now, you can almost feel the shift happening in real time. AI isn’t something that’s “coming soon”; it’s already here, threaded through conversations, steering committee meetings, and early-stage experiments across every department. What I’m seeing is that most banks had a strong financial year, and that success has created momentum. There’s more budget, more confidence, and more willingness to push AI initiatives forward. But strong financial performance hasn’t resolved the core problem that banks still struggle to plan, prioritise, and sequence AI projects consistently and at scale.
Every team knows they need AI. Every project wants to “do something with AI.” Yet far too many use cases stall because the planning, structure, and analysis behind them aren’t clear enough to support the level of rigour banks require. And in a regulated environment, poor planning slows everything down. From the conversations I’m having daily, here’s why I think that gap between ambition and execution still exists and how banks can finally bridge it.
Why do so many banks struggle to choose the ‘right’ first or next AI use case
When I speak to people across different teams, whether it's compliance, operations, or newer innovation functions, they all say the same thing: “We know AI can help… but we don’t know exactly where to start.” That’s because AI is sitting over the organisation like a giant cloud of possibility. It can automate workflows, improve compliance, analyse data, generate insights, and support customer-facing roles. It can do so much that teams sometimes freeze at the first decision of which use case actually deserves attention first.
Inside banks, the complexity is higher than in startups. Regulations are non-negotiable. Governance structures are dense. And teams can’t just test something quickly to see how it works. Because AI has become so big so fast, many teams don’t yet know how to break the technology down into practical, incremental steps that align with the pace of a bank.
And because banks compete with each other, they don’t want to be the one institution lagging behind while others automate processes, reduce costs, or build new intelligent capabilities. That creates pressure but not clarity.
External research shows the same pattern across the industry. Financial services organisations are investing heavily in AI, but adoption is uneven because strategy often outpaces execution capacity. Slalom’s 2026 global research found that many firms refresh their AI strategy more frequently than they update their metrics, governance, or funding models, which widens the gap between ambition and day-to-day delivery.
What I see on the ground reflects that. The desire to move quickly is real, but the foundations needed to support that movement are still catching up.
The most important step in evaluating whether an AI idea deserves investment
The real challenge isn’t the technology, it’s the planning. Someone will say, “We want to use AI for this,” and the next logical question is, “Okay, but what does that actually mean for your project?” That’s where banks need strong project analysts or business analysts who can sit between the idea and the execution. Their job is to suss out the project, understand its scope, and work out what AI can realistically contribute.
The most important step is analysis, not excitement. You need to fully understand the project before you decide whether AI adds value. In practice, that means teams should ask:
What problem are we solving?
Is AI truly the right tool, or is there a simpler approach?
What data is required, and does it exist in a usable form?
What risk, compliance, and model‑governance implications come with this?
Making the wrong call at this early stage can set a project up for failure. AI isn’t something you sprinkle on top, it has to be integrated into workflows that already have high accountability.
Industry research from KPMG echoes this: one of the biggest blockers to progressing AI initiatives is difficulty defining value, setting clear metrics, and linking investments to measurable outcomes. Without that clarity upfront, AI ideas don’t survive long once they move into formal investment review.
From my perspective, if you can’t clearly articulate the business problem and the expected outcome, the idea isn’t ready yet.
What separates initiatives that scale from those that stall at the PoC stage
The difference I’ve seen in conversations is planning. The teams that scale have done the groundwork. They know what the project looks like from end to end. They’ve mapped milestones, understood dependencies, and built a realistic view of how AI fits into the workflow without disrupting compliance or creating new risk.
The ones that stall usually didn’t start with enough structure. They ran a proof of concept because “AI sounded exciting,” but didn’t plan beyond the initial test. Once they reach the point where they need enterprise data, regulatory approval, or integration with existing systems, everything slows down.
McKinsey calls this “pilot purgatory”, when organisations experiment with narrow use cases but haven’t rewired processes, governance, and operating models to support scale.
I see versions of this every week. AI ideas move quickly at first because the initial build is exciting and accessible. But when banks need accuracy, governance, and long‑term maintainability, AI projects expose gaps in planning. In highly regulated environments, one wrong trade or one misclassified document can have huge consequences. That’s why careful sequencing and robust analysis are non-negotiable.
How banks should approach ROI planning before building an AI solution
Most teams want to talk about savings and efficiency, which makes sense, and that’s where the early wins tend to come from. In fact, many of the AI projects I hear about right now revolve around automation, internal compliance improvements, or speeding up manual processes. These are the areas where banks feel safest starting.
But ROI planning can’t just focus on operational savings. Banks should start by mapping the full lifecycle impact:
Which processes will be faster?
What errors could be avoided?
How will compliance risk change?
What does adoption look like across the team?
What new capability will this give us next year, not just today?
PwC’s research reinforces this broader view. They highlight that AI impacts both revenue and cost bases by improving efficiency ratios while also enabling better customer targeting and operational agility. Their point is that ROI isn’t a single metric; it’s a blend of performance improvements tied into the bank’s long-term AI maturity.
When I speak to managers, the ones who see ROI clearly are the ones who map the impact before they start building. They don’t wait until they have a prototype to ask whether it was worth it. They work backwards from the outcome, so the build itself has direction.
Here’s what to do before writing a single line of code
Build the foundations before the model. AI can move fast, but banks can’t. And that’s not a criticism but the reality of working in an environment with strong regulatory oversight.
The second thing is: invest in analysis. Hire or assign project analysts who actually understand how to evaluate AI opportunities, who can look at a project and break it down into what’s viable now and what needs more structure. This is the step I see most banks prioritising right now because it’s where so many projects break down.
And finally: get teams comfortable with AI literacy. I’m seeing people who aren’t engineers taking AI certificates now, and it’s great. Learning the basics helps teams make better decisions, reduces fear, and makes AI feel accessible rather than intimidating. When teams understand the fundamentals, they collaborate more effectively with the technical side.
Industry trends support this shift as well. Financial institutions are scaling AI across more workflows, and those that succeed invest early in data readiness, capability-building, and cross-team collaboration rather than jumping into engineering first. Starting strong in 2026 means starting structured. The excitement can come later, but the planning has to come first.
Closing the gap between transformation and delivery
Banks aren’t struggling because they lack ambition. They’re struggling because AI is expanding faster than their planning frameworks can adapt. But from what I’m hearing, the momentum is building. Budgets are increasing. Teams are preparing. And this year feels like a turning point. It’s the moment when AI stops being a scattered collection of side projects and starts becoming part of how banks operate.
The strategy-to-execution gap is real, but it’s fixable. With clearer planning, stronger project analysis, and realistic sequencing, banks can finally move AI initiatives out of the concept stage and into everyday workflows where they’ll actually make a meaningful difference.
AI is happening. The question now is how well each bank prepares to use it.
How Caspian One Helps Banks Close the AI Planning Gap
Unclear use‑case selection, weak analysis, stalled proofs of concept, and unreliable ROI planning often stem from a single root issue: organisations lack the specialised talent required to deliver AI effectively in a regulated financial environment. This is precisely where Caspian One’s AI Practice provides value.
Caspian One supplies AI specialists with genuine financial‑markets experience, ensuring that teams not only understand the technology but also the operational, regulatory, and risk‑sensitive contexts in which it must function. Our AI Practice is built on the understanding that scaling AI in finance requires more than technical capability; it demands professionals who grasp trading workflows, compliance constraints, and the metrics that determine whether a project will ultimately deliver measurable benefit.
With the combination of financial‑sector expertise and specialist AI capability, Caspian One enables banks to make better early decisions, strengthen their planning processes, and accelerate delivery, ultimately reducing the strategy‑to‑execution gap that holds so many AI programmes back.
Disclaimer: This article is based on publicly available, AI-assisted research and Caspian One’s market expertise as of the time of writing; written by humans. It is intended for informational purposes only and should not be considered formal advice or specific recommendations. Readers should independently verify information and seek appropriate professional guidance before making strategic hiring decisions. Caspian One accepts no liability for actions taken based on this content. © Caspian One, 2026. All rights reserved.
Read more or search for topics that matter most to you!
