Making Better Decisions with Less Information

As software engineers, we often make some of our most important engineering decisions when we have the least information. We choose which problems to investigate, which approaches to compare, and which assumptions we can reasonably proceed with.

Those choices are shaped partly by the cost of learning. When exploring an alternative takes weeks, narrowing our options early can be a reasonable compromise.

When AI reduces the effort needed to research a problem or build a prototype, we can afford to investigate more alternatives. But that depends on how organizations use the time it frees up. If every hour saved becomes another delivery commitment, little time remains for investigation.

Solving the Right Problem Before Building the Solution

Consider a wallet provider whose double-entry ledger keeps accurate financial records. Yet its operations team still has to search several systems to investigate a disputed transfer. The team asks for a better dashboard.

That might be the right solution. But perhaps transaction references are inconsistent across systems. Or perhaps the information is there, but the exception-handling workflow makes it difficult to use. Each explanation points to different engineering work.

Before committing to the dashboard, engineers could use AI to prototype alternative workflows and develop test scenarios from representative cases supplied by the operations team. Together, they could test which approach helps resolve those cases with fewer steps and less uncertainty.

These experiments would help inform the decision, not establish that a solution is ready for production.

The result might be a dashboard, a smaller integration change, or a simpler process. The value comes from learning what addresses the problem before committing to a solution.

Streamlined Expense Tracking and Reconciliation

Beyond security, virtual cards simplify the entire accounting process. Every transaction is automatically tied to a specific card, which can be linked to an employee, department, or travel request. This removes the need for manual expense claims and reduces the back-and-forth between employees and finance teams.

At the end of each cycle, reconciliation becomes significantly faster. Finance teams can instantly see who spent what, where, and for what purpose. This not only improves accuracy but also gives leadership real-time insights into travel spending patterns and cost efficiency.

Scalable Infrastructure for Growing Teams

As organizations grow, managing travel expenses manually becomes unsustainable. Virtual card systems are built to scale with demand, allowing companies to issue hundreds or thousands of cards instantly without operational bottlenecks.

This makes them especially valuable for businesses with distributed teams, frequent travelers, or contractor-heavy operations. Instead of relying on centralized approvals for every transaction, teams can move faster while still staying within predefined financial controls.

Value Engineering at Vrtx

At Vrtx, we build financial infrastructure for wallets, cards, and payments, with a focus on immediate, practical value for our clients.

We start with the outcome they need and work backward to the right solution. That might mean helping them launch a new service or making an existing one easier to operate.

This is where I see an opportunity to put more emphasis on value engineering: starting with the outcomes we want, then deciding what is worth building, improving, or simplifying to achieve them.

Those outcomes have always mattered. AI can give us more capacity to explore how best to reach them.

From Experimentation to Better Engineering Decisions

In practice, product and engineering teams should agree on the outcome they want and the uncertainty worth resolving, then set aside time to compare approaches.

The evidence from those experiments should guide whether they proceed, change direction, or stop.

Organizations should recognize this work too. Finding that a requested feature would not solve the problem is a useful result, even when no new software ships.

More experimentation is not the goal. Better-informed decisions are.

The Future of Software Engineering Is About Value

AI gives us an opportunity to spend less time on repetitive implementation and more time understanding problems, exploring alternatives, and deepening our expertise.

Making use of that opportunity requires a deliberate choice about how we spend the time we save.

To me, that's a far more interesting future for software engineering than simply completing our backlogs faster.