Written by: Frank Caccio
Everyone in financial services is talking about AI right now. Boards are asking about it. CFOs are budgeting for it. And the pressure to do something is real. KPMG's Global AI Pulse found that 74% of leaders will keep AI a top investment priority even if a recession hits, with 32% already deploying and scaling AI agents. In the UK, EY's financial services research found 34% of firms say they have fully integrated AI into their operations. Budgets are going up. Expectations are high.
And yet most firms are stuck. In a 2026 survey of 150 financial services firms run by Coastal Cloud with Oxford Economics:
● Two thirds cite data as a leading area where AI initiatives stall
● 71% report data accuracy or availability problems still hurting AI performance after launch
● 61% say AI fell short of the return they expected for the cost and effort invested
The gap between ambition and execution is wide, and I hear about it in nearly every conversation I have across the industry.
I think I know why.
Most firms are asking the wrong question.
They are asking: should we build AI ourselves or buy a tool that does it for us?
The better question is: what does our operational foundation actually look like right now, and is AI going to be able to do anything useful with it?
Artificial intelligence is only as intelligent as the data it has access to. If your workflows live in spreadsheets and your institutional knowledge walks out the door every time a senior person leaves, AI has nothing meaningful to work with.
The firms that are going to win are not the ones that rush to buy or build the most sophisticated AI tools. They are the ones that get their operational house in order first.
But can't AI just build the foundation for you?
A lot of firms think so. There is a real belief right now, fueled by the rise of vibe coding, that AI can solve the foundation problem itself. Vibe coding means building software by describing what you want in plain language and letting AI write the code. The idea is appealing. The evidence is pushing back hard on it.
Veracode's 2026 GenAI Code Security Report tested more than 100 models and found the average security pass rate stuck at 56%, with 44% of code-generation tasks producing code that carried a risky vulnerability. On speed, the picture is mixed: a controlled Microsoft and GitHub experiment found developers completed a bounded, well-defined task around 56% faster with an AI assistant, while METR's randomised trial of experienced developers working in their own large repositories found they were 19% slower with AI. What the optimistic studies measure is speed to a working first version. Not speed to something that is production-safe, audit-ready, and regulation-compliant.
The more honest framing is this: AI lowers the cost of writing code. It does not lower the cost of understanding the problem. Someone could probably use AI to build something that looks like an operational platform in a short amount of time. What they cannot do is replicate thirty years of embedded business logic from someone who actually sat in the chair and ran these operations, combined with a decade of real-world refinement from the COOs, CFOs, CCOs, and heads of operations who shaped the product through their daily use and feedback. That is not code. That is institutional knowledge. And it cannot be prompted into existence.
What the foundation should look like.
● **Clean workflows. **Every process your team runs, monthly NAV close, quarterly compliance reviews, investor and client onboarding, should be structured, assigned, and tracked somewhere that is not inside someone's head or a shared folder nobody updates.
● **Clear accountability. **Every task needs one owner. One person who did it and one person accountable for it. When something is late, you should know before anyone has to tell you.
● **Documented evidence. **Regulators and investors are not asking whether your controls existed. They are asking whether you can prove it, quickly, and in a format they can actually review. Evidence that cannot be retrieved on demand is evidence that might as well not exist.
● **A plan for when things go wrong. **High risk events are not rare. NAV errors, contract trigger breaches, regulatory inquiries, cybersecurity incidents, key person departures, these happen to every firm. The ones that handle them well are the ones that had already thought through the response. If your team is figuring it out in real time on the day it happens, you are already behind.
The industry is not waiting.
According to Alpha FMC, asset management firms are moving past AI experimentation and into large-scale implementation, productivity, risk management, investment decision making. The firms pulling ahead are the ones that see operational discipline and AI adoption as the same strategy, not two separate projects.
If your foundation is not ready, getting there is going to cost you more time and more money than it should.
● Two thirds cite data as a leading area where AI initiatives stall
I think I know why.
Most firms are asking the wrong question.
But can't AI just build the foundation for you?
What the foundation should look like.
The industry is not waiting.
This is a solvable problem.
The good news is that none of this requires a massive technology overhaul or months of painful implementation. It requires structure, discipline, and the right platform to hold it all together. Most firms that get this right do not start by overhauling everything at once. They start with one team, one workflow, one recurring process, and build from there. The foundation goes up faster than most people expect, and the value shows up quickly.
This is exactly what OpsCheck solves.
After 25 years running operations at firms like Tiger Management and Highbridge Capital, the challenge I kept running into was the same: no clear and precise way to oversee the full operation. No single place where you could see what was done, what was not done, and what was at risk. That gap is what I built OpsCheck to close.
OpsCheck is the operational foundation that the rest of this article has been describing. One place where every workflow, control, task, and accountability lives. The result is an operation that greatly reduces the chances of a major operational failure, which can have serious consequences, and one that can demonstrate to investors, clients, regulators, and auditors that the firm has full control over its operational infrastructure.
The platform did not stop at my own experience. It has been shaped continuously by the feedback of the COOs, CFOs, compliance officers, controllers, and heads of operations who use it every day. That collective knowledge is built into every part of it.
We serve the full investment management ecosystem, alternative and traditional asset managers, fund administrators, family offices, allocators, and the service providers that work alongside them. OpsCheck is SOC 2 certified, used across three global regions, and built from the kind of COO-level operational experience that no vibe-coded competitor can replicate.
We also have three AI capabilities in active development, each sitting directly on top of this foundation: conversational access to your own operational data, a significantly easier way to build and structure workflows, and a growing playbook library covering high risk events and routine best practices across every function of the firm.
