Technological legacy as a barrier to AI, data, and operational efficiency

You have likely experienced this scene: the same question posed to two different areas returns with two different numbers. Each area trusts its report, each report comes from a distinct database, and the meeting that should decide the future spends the first half-hour determining which number is the true one. When this happens frequently, the institution does not have a reporting issue. It has a data governance problem. And it costs more than it seems.

Why does the financial market have the most valuable and fragmented data?

Few sectors generate such rich data. Each transaction, each credit proposal, each claim, and each customer interaction carries information about behavior, risk, and opportunity. In theory, no sector should know its clients better than the financial sector.

However, this data structure was born fragmented and for historical reasons. Institutions grew in layers: the central system from decades ago, product systems that came later, digital channel platforms, point solutions acquired to address emergencies, and databases inherited from mergers and acquisitions. Each layer keeps a piece of the customer. Add to this the regulatory diligence, which over the years has encouraged silos as a form of control; thus, the data exists, but it doesn’t

Why is data dispersion a governance problem?

Because fragmentation is not just an operational inconvenience. It undermines the three things for which boards and management are accountable: decision quality, risk control, and the integrity of what is reported to the regulator.

When each area has its version of the data, the institution loses its single source of truth. Credit decisions, pricing, and capital allocation come to depend on manual reconciliations, intermediate spreadsheets, and the knowledge of those who “know where the right number is.” This is fragile, slow, and, in a regulated sector, risky: LGPD, the Central Bank, and Susep don’t ask from which system the data came; they ask if the institution is accountable for it.

Defining who owns each piece of data, who can access it, what the official source is, and how quality is measured are management decisions with a direct impact on results and risk. Technology implements it, but governance belongs to the business.

How to build a culture where data leads to decision-making?

The most challenging part is not technical. Institutions invest in modern data platforms and continue to make decisions based on intuition because culture cannot be purchased along with software licenses.

Three actions make a difference. The first is to assign ownership of the data: each domain of information needs a business owner with the authority to define the official source and ensure quality. Without an owner, all data belongs to everyone, and the quality belongs to no one. The second is to bring data to where decisions are made, in the committee, in credit, on the product desk, in business language and not in programming language. Data that requires translation arrives late. The third is for leadership to lead by example: when the executive committee rejects discussions based on numbers without a clear origin, the entire company quickly learns what the new standard is.

What is the role of data in scaling AI sustainably?

Here, the conversation becomes truly financial. The question I hear most about Artificial Intelligence is how much to invest. The right question comes first: on what base will this investment operate?

AI applied to credit, fraud, service, and efficiency only generates returns at scale when it finds integral, governed, and accessible data. On a fragmented base, each AI initiative must first build its own data infrastructure, which multiplies costs, prolongs timelines, and produces that pattern which boards are beginning to recognize: many pilots, little recurring results. An AI pilot based on poor data is not an investment; it is an expense with a good appearance.

Scaling sustainably is the opposite: solving the data foundation once and reusing it for each new use case, decreasing the marginal cost of each AI application rather than increasing it. This is how the investment curve in AI begins to behave like an asset curve, which appreciates with use, and not like a collection of projects competing for budget in every cycle.

A concrete case: governance as a foundation in a credit union

A large credit union recently experienced a version of this dilemma. About 50 critical processes, from account openings to credit lines, operated on a legacy automation without structured governance: without careful management of who accesses which processes and which data.

Instead of swapping one tool for another, the credit union built, alongside Objective, a new orchestration layer with governance designed from the outset, including careful management of access permissions to processes and data, and with the internal team empowered to develop autonomously. Governance as a foundation, not as bureaucracy added later.

In just four months, the first critical processes were already running in production on the new platform, with more security, faster development speed, and smooth integration with existing automation. Learn about the results of this success story.

How much is the data your institution already has, but still cannot use worth? Objective supports financial institutions in building governed, auditable data foundations ready for AI, combining consulting, data engineering, and internal team training. Talk to OGI and transform fragmented data into a composing asset.

Frequently asked questions on the topic