AI-Assisted Dependency Review
How I used Conductor and Claude to streamline my team's Dependabot review workflow — and how a colleague turned it into a GitHub Action
Frontend technical debt is devalued by startups. Early stage startups try to hire full-stack engineers. Frontend specialists only get hired later, after the foundations have been set and become difficult to change.
When you don't address technical debt for years, you cause hard-to-untangle problems with real business impact. For example, I once worked at a company where we discovered that the Chart page (a crucially important page to the product), took over 14s to load. We only found out via support tickets. That "cosmetic" issue likely cost us real revenue.
Frontend technical debt isn't just about developer happiness. It undermines user experience, creates legal liability, and quietly bleeds revenue. But because the costs are distributed and often invisible, they don't get the attention they deserve.
In most engineering organizations, there's an implicit hierarchy. Backend systems are "real engineering" because they handle data, scale, and keep the lights on. Frontend work is often seen as polish, something you do after the important stuff is built.
This mindset is dangerous because the frontend is your product from the user's perspective. They never see your database schema or your microservices architecture. They see your UI, and they judge your entire company based on that experience.
I've worked at multiple companies where the moat is an incredibly sophisticated database with powerful querying capabilities. Power users can surface critical insights through a deep understanding of how to instrument complex systems. But as these companies scale, they aren't selling to power users. They need to pivot to selling to enterprise customers. If they haven't invested in the user experience sufficiently, they risk losing crucial deals that they need if they want to move upmarket.
The problem with frontend technical debt is that it creates two kinds of pain: the obvious kind that slows down developers, and the insidious kind that impacts a company's bottom line.
A messy component library makes feature development slower—that's visible and gets prioritized. But a confusing user flow that increases time-to-value? That shows up as churn, support tickets, and mysterious drops in conversion metrics that are harder to trace back to specific technical decisions.
Performance is probably the most measurable way frontend debt kills businesses, but it's also the most ignored. We all know the statistics: Amazon found that every 100ms of latency cost them 1% in sales, Google discovered that increasing search results time by 400ms reduced daily searches by 0.74%. But somehow, these numbers feel abstract until you're staring at your own conversion dashboard wondering where the customers went.
I spent most of last year building observability tools, so I'm particularly sensitive to the gap between what we measure and what actually matters. Most application monitoring focuses on server-side metrics like response times, error rates, throughput. But your 50ms API response doesn't matter if your bundle size means users wait 3 seconds for anything to render.
The worst part is that performance degradation is gradual. You don't wake up one day to find your site is slow. It happens dependency by dependency, feature by feature. A marketing team adds a new tracking script. Someone imports the entire date library instead of just the function they need. A designer requests a custom font that looks great but takes forever to load. Each decision seems small, but the compound effect is massive.
Accessibility is an afterthought because businesses don't perceive or measure lost revenue based on it. Power users aren't complaining about accessibility, so the problem stays invisible until it becomes a blocker.
But the legal risk is just the tip of the iceberg. Poor accessibility doesn't just exclude users with disabilities—it makes everyone's experience worse. When you design an interface that works for someone using a screen reader, you're forced to create clear information hierarchy, logical navigation flow, and meaningful labels. Those improvements help everyone.
For example, imagine you are tasked with investigating a customer feedback form where completion rates were mysteriously low. The form may look fine visually, but when you test it with keyboard navigation, you discover the tab order is completely broken. You discover thta when users hit tab, their focus would jump randomly around the form, skipping required fields. This kind of issue doesn't only exclude users who navigate by keyboard. It creates a confusing experience for everyone who tries to tab through quickly.
The compound effect is massive. When your interface sends mixed signals about what's clickable, where focus should go, or what information is important, you increase cognitive load for all users. The person using a screen reader who can't figure out your navigation is having the same fundamental problem as the person using a mouse who can't tell which buttons actually do something.
Enterprise customers increasingly require accessibility compliance in their vendor evaluations. I've seen deals stall because procurement teams needed WCAG 2.1 AA compliance documentation that we didn't have. Accessibility went from nice-to-have to table stakes.
Inconsistent design patterns are like road signs that give contradictory directions. Users learn to expect certain behaviors from common UI elements. Buttons should look clickable, form validation should be consistent, navigation should be predictable. When your interface breaks these expectations, users lose confidence in your product.
This isn't about making things pretty. It's about building trust through predictability. If your primary button style means "submit" on one page but "cancel" on another, users have to think harder about every interaction. That cognitive overhead adds up, especially for complex products.
When users can't confidently navigate your interface, they start questioning whether your product is right for them. Professional appearance affects perceived product reliability. If your UI looks sloppy, users assume your software is sloppy too.
When communicating technical debt to business stakeholders, remember that they don't care that your build process is slow or your CSS is messy. They care about customer retention, revenue growth, and competitive advantage. If you want to get budget and prioritization for frontend platform work, you need to translate technical problems into business metrics.
Instead of saying "our component library is inconsistent," say "inconsistent UI patterns are increasing our support ticket volume by 15% and hurting user activation rates." Instead of "our bundle size is too big," try "slow page loads are costing us 8% of potential signups based on industry benchmarks, and here's how we measure our current performance."
I've found success with this basic framework: identify the business metric that's being hurt, quantify the impact, and propose a measurable solution. "Our accessibility issues exclude 15% of potential users and create legal risk" is more compelling than "we should fix our keyboard navigation."
The key is connecting frontend technical debt to metrics that already matter to your organization. If your company tracks conversion rates, show how performance improvements could increase conversions. If customer satisfaction is important, demonstrate how design consistency reduces user frustration. If you're trying to close enterprise deals, make the case for accessibility compliance.
Building a frontend platform approach means treating frontend infrastructure as seriously as backend infrastructure. You wouldn't ship a backend service without monitoring, error handling, and performance tracking. Your frontend deserves the same level of systematic attention because it directly impacts the metrics your business cares about.
Performance degradation, accessibility barriers, design inconsistency, unreliable testing are problems that don't announce themselves with dramatic outages. They quietly undermine your growth by making your product harder to use, harder to trust, and harder to improve. The companies that recognize and address these hidden costs will have a significant advantage over those that keep treating frontend work as an afterthought.
Your frontend isn't just the pretty face of your application. It's your entire user experience and a critical piece of business infrastructure. Start measuring it, monitoring it, and investing in it accordingly.
How I used Conductor and Claude to streamline my team's Dependabot review workflow — and how a colleague turned it into a GitHub Action
If everyone's shipping faster with AI, nobody has an advantage. The real edge is platform work.
How I'm using Claude Code to build a full-stack app while maintaining architectural understanding