Tax season is that time of year that puts everyone on high alert, and for banks, it goes way beyond the usual tax obligations.
This is exactly when financial institutions’ AI systems face their biggest test: transaction volumes skyrocketing, regulatory pressure through the roof, and a level of data complexity that doesn’t forgive mistakes.
And you know what happens when an artificial intelligence system doesn’t have a solid foundation to support all of that?
The cracks show up, and fast. 😬
That’s exactly what Mark Blake, financial services industry practice lead at Stibo Systems, talked about on FinAi News’ The Buzz podcast.
According to him, tax season works as a real stress test for the AI models in use at banks, exposing problems that go unnoticed the rest of the year — especially those related to data quality and information governance.
In this article, we take a deep dive into that conversation and understand why getting data foundations right can be the difference between a smooth tax season and a full-blown operational nightmare. 🚀
The current landscape: banks are betting big on AI
Before getting into the problems, it’s worth putting into context just how big of a bet financial institutions are making on artificial intelligence. In his conversation with FinAi News, Blake explained that over the past decade, banks have invested heavily in modernization — migrating to the cloud, building robust data platforms, and developing analytical layers on top of those structures. The result, according to him, is that roughly two-thirds of financial organizations have already deployed AI in some capacity within their operations.
These systems are present in areas like fraud monitoring, credit underwriting, document classification, digital assistants, and risk modeling. We’re not talking about pilot projects tucked away in some corner of the company. These are applications that are already part of the daily operations of these institutions, influencing decisions that affect millions of customers and billions in transactions. And precisely because they’re so deeply integrated into how banks operate, any failure in these models takes on massive proportions during high-pressure periods like tax season.
Blake also made an important point: while many of these AI implementations are working, a good number of them are still more tactical than strategic. In other words, they were adopted to solve specific problems within isolated departments, without an integrated vision connecting the entire value chain of the organization. This kind of fragmented approach works up to a point, but it doesn’t hold up under real pressure — when all systems need to work together in a coordinated and reliable way.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Why tax season exposes the flaws in AI systems
During tax season, banks deal with a significant increase in transaction volumes, customer inquiries, and reporting obligations to regulators. This spike in activity pushes AI systems to a much more intense pace than usual, and that’s when foundational problems start showing up more clearly. When the data feeding these models is outdated, inconsistent, or poorly organized, the system starts delivering inaccurate responses, wrong recommendations, and in more serious cases, failures that can trigger serious regulatory consequences for the institution.
Mark Blake made it very clear on the podcast that this scenario isn’t the exception — it’s the norm across a large portion of financial institutions that haven’t invested strategically in data quality. The problem, according to him, is that during normal operating periods, these systems manage to mask the inconsistencies because the volume of requests is lower and the margins of error go unnoticed. But when demand spikes dramatically, as it does during tax time, the system simply can’t hide what was broken underneath anymore. It’s like a bridge that seems solid on a regular day but shakes when a heavy truck rolls across it.
Blake pointed out that tax season is a deadline-driven event. Everything needs to be completed within a specific window of time. That means there’s no room to start over or adjust processes midstream. Banks need their AI models and data pipelines to work correctly the first time around, which makes any failure in the information chain far more costly than it would be at any other time of year.
Another relevant point Blake raised is that banks frequently underestimate the interdependence between different data sources within their own infrastructure. An incorrect piece of data in a customer record can propagate across multiple AI systems simultaneously, generating cascading errors that range from offer personalization all the way to tax reporting. And fixing those errors retroactively in the middle of the season is an expensive, time-consuming process full of operational risks.
The warning signs that show up during tax season
Blake was quite specific in describing the red flags that typically emerge at financial institutions during tax season. The first and most common one is record inconsistency across different systems. When customer identifiers don’t match, when data is missing from essential fields, or when different versions of the same information coexist across separate platforms, AI models simply can’t deliver reliable results.
The second major warning sign involves traceability and auditability. Financial regulators increasingly require banks to be able to explain how an automated decision was made. This is what the industry calls explainability of AI models. If the bank can’t demonstrate the lineage of the data that fed a decision — where it came from, how it was processed, and why it was considered trustworthy — it becomes exposed to regulatory scrutiny that can result in sanctions.
The third point is the operational symptoms that emerge when systems fail. Blake mentioned that one of the clearest indicators that something is wrong is an increase in manual corrections. When teams need to step in repeatedly to fix errors that should have been resolved automatically, it signals that data quality at the foundation level is compromised. Conflicting information about values, customer profiles, or transaction statuses are concrete examples of these problems that eat up time and resources during the tax period.
Data quality: the foundation nobody sees but everyone feels
Data quality is one of those topics that seems too technical to make it onto the agenda of strategic meetings, but in practice it defines the success or failure of any AI initiative inside banks. Blake took a pretty direct approach: you can have the most sophisticated model on the market, but if the data feeding it is bad, the output will be bad too. Worse than that, AI can end up amplifying problems that already existed, delivering wrong answers at a speed and scale that manual processes could never reach.
In the context of tax season, this becomes even more critical because the information needs to be accurate, traceable, and available in real time. Banks need to cross-reference transaction data, customer information, tax records, and financial histories quickly and precisely. Any noise in that chain can result in incorrect filings, inconsistencies with IRS records, and even fines. And as AI systems become increasingly present in this process of data consolidation and analysis, the responsibility for the quality of that information falls directly on the institution’s data infrastructure.
Blake explained that the concept of the Golden Record is essential in this context. It means having a single, consolidated, and trustworthy version of each customer’s data and each transaction, with clear validation rules applied consistently. When that record exists and is properly maintained, AI models can work on a solid foundation, generating results that can be audited, explained, and defended before regulators.
Stibo Systems, the company where Blake works, specializes in Master Data Management solutions that help financial institutions centralize, standardize, and ensure the reliability of information flowing through their systems. In Blake’s view, institutions that invest in this layer of governance before tax season arrive much better prepared for the challenges of the period — with more stable AI systems, more accurate decisions, and significantly less regulatory exposure risk.
The importance of real-time data during tax season
One of the points Blake emphasized quite clearly during the conversation was the need for access to real-time data, especially during tax season. When systems operate with outdated information or when professionals need to access multiple platforms to build a complete picture of a customer or transaction, the risk of error increases significantly — and the time needed to complete each task multiplies.
According to Blake, a well-implemented Master Data Management platform solves exactly that bottleneck, offering a single, controlled environment where all relevant information is accessible, up to date, and validated. This eliminates the need for manual lookups across different systems, drastically reduces the chance of errors, and frees teams up to focus on higher-value activities instead of getting stuck in data reconciliation processes.
For banks, this real-time availability isn’t just a matter of operational efficiency. It’s a matter of regulatory compliance. During tax season, deadlines are tight and regulators expect fast, accurate responses. Institutions that rely on manual processes to consolidate data are at a clear disadvantage — both in terms of speed and reliability of the information they report.
Data governance: the strategy that separates prepared banks from those at risk
If data quality is the foundation, governance is the structural system that ensures that foundation stays solid over time. Within banks, data governance involves clear policies about how information is collected, stored, updated, and used by AI systems. It’s a set of processes, responsibilities, and controls that defines who has access to which data, how it’s validated, and what the criteria are for considering a piece of information reliable enough to be used in decision-making.
Blake pointed out that the lack of governance is one of the biggest challenges he encounters at the financial institutions he works with. Many banks still operate with data silos, where each department maintains its own databases with its own quality standards, and without any unified view of the customer or transactions. When tax season arrives and AI systems need to consolidate that information to generate reports, the mess reveals itself in all its glory. The data doesn’t match, versions conflict with each other, and the effort to manually reconcile everything eats up time and money that could be spent far more productively.
The good news, according to Blake, is that there’s a clear path to solving this problem, and it starts with adopting a governance strategy that treats data as a strategic asset of the organization — not as an operational byproduct. That means creating a trustworthy data layer with continuous validation processes, well-defined responsibilities, and an integrated view that allows AI systems to access accurate information regardless of its source. Banks that do this tend to perform better during tax season, with fewer incidents, less rework, and a much greater ability to respond quickly to regulatory demands.
Enterprise adoption: AI needs to be an organization-wide strategy
One of the topics Blake addressed with the most emphasis was the need for financial institutions to treat AI as a company-wide strategy, not as an isolated project from a technology department. He observed that many banks rushed to adopt artificial intelligence but did so in a tactical and fragmented way. The result is a series of disconnected initiatives that work well individually but don’t communicate with each other and don’t offer an integrated view of the business.
For AI to truly deliver value during high-pressure moments like tax season, there needs to be a coherent enterprise roadmap connecting data, models, and processes end to end. Blake mentioned that the institutions seeing the most success are those investing not only in technology but in education, awareness, and cultural change within the organization. When every department understands the role of data in the value chain and takes responsibility for its quality, the result is an operation that’s far more resilient and ready for market challenges.
Ongoing training and team development also came up in the conversation as decisive factors. Blake reinforced that people are an essential part of this equation. It’s not enough to have the right technology if the professionals using it don’t understand how to get the most out of it. And in the financial sector, where precision and compliance are non-negotiable, that understanding becomes even more critical.
What banks can learn from this annual stress test
One of the most interesting perspectives Blake brought to the conversation is the idea of viewing tax season not just as a period of pressure, but as a diagnostic opportunity. Every failure that surfaces during this period is a clear signal of where the weak points are in the bank’s AI and data quality infrastructure. Ignoring those signals and just putting out fires in the moment is a strategy that guarantees the same problems will repeat next year — possibly on an even larger scale.
Blake also noted that tax season isn’t the only high-pressure moment of the year. There are regulatory and reporting obligations throughout the entire calendar, and each one represents a similar opportunity to test and validate the robustness of data and AI systems. The difference is that tax season concentrates a particularly high volume of simultaneous demands, making problems more visible and urgent.
Banks that take a more proactive stance use the lessons from tax season to review their governance policies, identify the data sources that need attention, and fine-tune their AI models to better handle demand spikes. This continuous improvement mindset, driven by real business data, is what separates institutions that grow with technology from those that merely survive it. And with regulatory pressure increasing every year, that difference keeps becoming more and more relevant in the financial market.
What client conversations are revealing
Blake also shared a bit of what he’s been hearing in conversations with Stibo Systems clients in the financial sector. According to him, professionals working in areas like risk, pricing, and credit are coming to the table with a common theme: a good chunk of the work done so far in AI and data still isn’t delivering the expected results. And the growing perception among these teams is that the problem isn’t with the technology itself, but with the lack of a solid data foundation to build on.
Many of these institutions are now going through what Blake described as a backfilling process — going back to implement the master data management that should have been put in place before the AI layers. It’s an acknowledgment that the race to adopt artificial intelligence was, in many cases, done without adequate groundwork. And the ones making that course correction now tend to see better and more sustainable results in the medium term.
AI isn’t magic: results depend on the environment
The conversation with Mark Blake on The Buzz podcast serves as an important reminder for the industry: AI isn’t magic, and its results depend directly on the quality of the environment it operates in. Advanced models, sophisticated algorithms, and well-designed pipelines lose much of their potential when the data feeding them isn’t reliable. And in a sector as sensitive as finance, where the consequences of an error can go far beyond the operational, investing in data quality and governance isn’t optional. It’s the prerequisite for everything that comes after.
As Blake put it pretty directly: financial institutions don’t have a technology problem. Technology is the one thing this sector has plenty of. What’s missing in many cases is solving the data question. Once the data is correct, organized, and governed, everything built on that foundation — including AI models — starts working with far more efficiency and reliability. And tax season, with all its pressure and complexity, is the perfect moment to find out whether that foundation is truly solid or still needs attention. 💡
