Digital Pulse

From Engineer to Editor - The New Role of Developers in the Age of AI

Rafa Jimenez

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 22:13

Send us Fan Mail

In this episode of Digital Pulse, we explore how AI is collapsing the cost and timelines of software development and what that means for businesses, developers, and leaders. As code generation becomes nearly instantaneous, execution is no longer the bottleneck. The real challenge shifts to judgment: deciding what is worth building and why. The conversation introduces the rise of the “D‑user,” where end users shape software directly through natural language, and draws parallels to the prosumer revolution in digital media. We discuss the risks of speed without oversight, the evolution of developers from coders to editors, and why empathy, context, and product clarity are now critical technical skills. From living, self‑evolving software to redefining trust and human involvement, this episode is a deep dive into the mindset shift required to thrive in the AI‑driven future of software.

This episode is co-written by Marc Petralito, Head of Industry Business Line REM at ELCA

Links: 

ELCA website 

ELCA on LinkedIn 

Digital Pulse newsletter on LinkedIn  

Rafa Jimenez on LinkedIn 

ELCA on YouTube 

SPEAKER_00

Welcome to Digital Pulse. I'm Rafa, producer and host of this series where we explore how emerging technologies like digital assets, blockchain, and AI drive real innovation in business. Each episode is crafted using a fully AI-enabled workflow. I'll now hand it over to my AI co-hosts to take you through today's episode.

SPEAKER_02

Thanks, Rafa. Welcome everyone to another episode of Digital Pulse.

SPEAKER_00

It is great to be here.

SPEAKER_02

So I want you to imagine something for a second. Picture an RFP landing on your desk for a massive public sector infrastructure project.

SPEAKER_01

Aaron Powell Oh, yeah, those are usually monsters.

SPEAKER_02

Right. Historically, answering that tender means you're building this sprawling hundred-page slide deck. You've got three months of alignment meetings, and you're burning through consulting fees just to propose a theoretical architecture. Trevor Burrus, Jr.

SPEAKER_01

Just theoretical, yeah. Not even writing real code yet.

SPEAKER_02

Trevor Burrus Exactly. But now imagine a competitor answering that exact same tender in 72 hours. And not with a slide deck, but with a fully functional, customized working prototype.

SPEAKER_01

Aaron Powell Something the procurement officer can actually just log into and test right then and there.

SPEAKER_02

Aaron Powell Yes. We're no longer waiting for the future of software development, right? We are actively dealing with the fallout of the cost of execution fundamentally dropping to zero.

SPEAKER_01

Aaron Powell And that collapse of the timeline from months down to mere hours, it entirely shifts the gravity of the technology sector. I mean, we are navigating a transition that's easily as disruptive as the advent of the internet itself.

SPEAKER_02

Aaron Powell It really feels that massive.

SPEAKER_01

It is. Because for the last four decades, the primary constraint for any business was execution. Board-level conversations were all about, you know, feasibility, resource allocation, timeline, bottlenecks.

SPEAKER_02

Trevor Burrus, Jr.: Right. The how do we build this and can we even build this question.

SPEAKER_01

Exactly. But AI abstracting away the heavy lifting of code generation, that completely eradicates that specific constraint. The strategic culture for anyone leading technology today is no longer about the mechanics of building.

SPEAKER_02

So what is it about now?

SPEAKER_01

Aaron Powell It's exclusively about evaluating what is actually worth building. In a world where you can build almost anything instantly, you have to figure out what actually matters.

SPEAKER_02

Aaron Powell The implication there is just staggering. I mean, when that barrier to entry vanishes, it doesn't merely make your senior engineering team faster. It entirely redistributes who can create software across the whole enterprise.

SPEAKER_01

Yeah, it completely redefines the fundamental act of software engineering.

SPEAKER_02

Okay, let's unpack this. Because our mission for this episode of Digital Pulse is to dissect the mechanics of that exact shift.

SPEAKER_01

Aaron Powell Which is so needed right now.

SPEAKER_02

Aaron Ross Powell It is. We need to completely reframe the role of the developer for the next 20 years. We're going to examine the risks, like the catastrophic security and organizational risks of shipping what we call AI slop.

SPEAKER_00

Oh, that's a huge one.

SPEAKER_02

And we'll map out how teams actually need to restructure to survive a landscape where the code base is effectively written by the end user.

SPEAKER_01

Aaron Powell Right. And to frame this properly, I think we really need a historical parallel. Sociologists often talk about this thing called the prosumer shift.

SPEAKER_02

Aaron Powell The prosumer shift. Okay, walk us through that.

SPEAKER_01

So we saw this play out really violently in the early 2000s with the media landscape. Before blogging infrastructure took off, the cost of content distribution was a massive barrier.

SPEAKER_02

Oh sure. You needed literal printing presses, editorial hierarchies, delivery trucks.

SPEAKER_01

Exactly. The physical logistics alone were a gatekeeper. But then platforms like Blogger and WordPress came along and drove the marginal cost of publishing to zero.

SPEAKER_02

Aaron Powell Right. And the result wasn't just that legacy journalists typed their articles faster.

SPEAKER_01

Aaron Powell No, not at all. The readers, the people who are strictly consumers of the media, they mutated into producers. The boundary between consumer and producer completely dissolved. They became prosumers.

SPEAKER_02

I love that analogy because AI is doing that exact prosumer shift right now, but to software development.

SPEAKER_01

Literally the exact same thing. The tooling has evolved way past just being a sophisticated autocomplete for senior engineers.

SPEAKER_02

It's not just a fancy spell check for code anymore. Trevor Burrus, Jr.

SPEAKER_01

Right. It's lowering the foundational floor of creation so aggressively that the end user is absorbing the role of the architect.

SPEAKER_02

Aaron Powell And that introduces a new persona we really need to explicitly define for our listeners right now. We call it the dear user. The de user development plus user.

SPEAKER_01

Aaron Powell Exactly. A de user is an end user who actively directs shapes and fundamentally alters the application they consume, but they do it using strictly natural language. Right. They aren't writing Python or React. They aren't managing Kubernetes clusters.

SPEAKER_02

Aaron Powell No, not at all. The AI layer acts as this bridge between their raw intent, just what they want the software to do, and the actual implementation of the logic.

SPEAKER_01

Aaron Powell So the code itself kind of becomes this invisible byproduct. The divuser just interacts with their enterprise tools conversationally, and the architecture statefully reshapes itself around what they need to do right then.

SPEAKER_02

Aaron Powell Let's ground this for a second. For the business leaders listening who might not be digging into a code base every day, let's look at basic design tools.

SPEAKER_01

Aaron Powell Oh, like Canva or PowerPoint.

SPEAKER_02

Yeah, exactly. Think about how easy it is today for a totally non-technical, non-designer to open Canva, type in a few words, and the AI agent generates a balanced professional layout.

SPEAKER_01

Aaron Powell It just understands the design heuristics automatically.

SPEAKER_02

Right. Now I want you to transplant that exact dynamic into complex, data-heavy enterprise workflows.

SPEAKER_01

Aaron Powell That's where it gets really interesting. Let's take a real world scenario. Imagine a claims manager at a tier one insurance firm.

SPEAKER_02

Okay. Pretty standard corporate role.

SPEAKER_01

Right. In a traditional workflow, if that manager spots a systemic inefficiency, say they want to correlate local weather data with roof damage claims on a single dashboard to spot fraud, what do they have to do?

SPEAKER_02

Well, they have to submit an IT ticket.

SPEAKER_01

Exactly. And that request goes into this massive backlog. It has to battle for prioritization in the next quarter.

SPEAKER_02

Aaron Powell A product manager has to scope it out, a developer has to build the API integration, QA tests it.

SPEAKER_01

Right. And maybe five months later, the dashboard actually updates. Trevor Burrus, Jr.

SPEAKER_02

Which is just five months of lost operational efficiency.

SPEAKER_01

Aaron Powell But under the Divuser model, that claims manager essentially edits their own environment. They just prompt their AI assistant, I need to visualize local weather data APIs overlaid on this week's roof claims.

SPEAKER_02

Just in plain English.

SPEAKER_01

And under the hood, the AI agent interpret the intent, queries the internal enterprise data lake, all while respecting row level security and active directory permissions, by the way.

SPEAKER_02

Aaron Powell Oh, that's crucial. It's not bypassing security.

SPEAKER_01

Nope. It writes the localized query, fetches the external weather data, and dynamically generates the front-end components to render that specific visual.

SPEAKER_02

Aaron Powell So a totally new application state is just instantiated in what, four minutes?

SPEAKER_01

Literally minutes. Zero IT tickets, zero sprint planning.

SPEAKER_02

Aaron Ross Powell Wow. Okay, but here's where it gets really interesting. And I know every CTO listening is going to push back on this.

SPEAKER_01

Aaron Powell Security and architectural integrity.

SPEAKER_02

Aaron Ross Powell Exactly. If everyone can build software in the afternoon, aren't we just going to drown in a sea of terrible disconnected apps? If a field salesperson can just prompt their CRM to say, show me stagnant deals over 30 days and auto-remind me every Friday, what's the risk of them writing infinite loops that crash the database?

SPEAKER_01

Aaron Powell or exposing personally identifiable information, right? You're hitting on the core architectural pivot here. You do not give the AI agent root access to the production database.

SPEAKER_02

Aaron Powell Okay, so how does it work safely?

SPEAKER_01

You build a semantic layer. The AI acts as an object relational mapper on the fly, but it writes the logic within a heavily sandboxed environment.

SPEAKER_02

So it's contained.

SPEAKER_01

Highly contained. If a manufacturing floor manager spins up five what if dashboards, they aren't altering the master application for the whole company. Those are localized, containerized instances that only execute within that user-specific permissions.

SPEAKER_02

That effectively obliterates the traditional product roadmap, doesn't it?

SPEAKER_01

Completely. The sequential single-lane pipeline where you prioritize feature A over feature B. That's dead.

SPEAKER_02

Because if the gap between an idea and a functional product is tending to zero, you can just move to a portfolio mindset. Product teams can do ABCDE testing instead of just AD testing.

SPEAKER_01

Yes. You deploy entirely distinct product architectures concurrently against real user cohorts.

SPEAKER_02

But wait, that velocity, that speed introduces a massive vulnerability.

SPEAKER_01

Right. The risk of infinite speed without human judgment.

SPEAKER_02

Yeah, it's shadow IT on steroids. If the marketing intern and the logistics manager are all spinning up custom logic loops, the organization is going to ship AI slop until the technical debt just crushes the system.

SPEAKER_01

And validating that fear is so paramount. The bottleneck in technology has definitively shifted. It's no longer about execution, it's about human evaluation.

SPEAKER_02

Because if you just automate the output without elevating your evaluation capacity, you just generate chaos faster.

SPEAKER_01

Exactly. And this is why the role of the professional software developer is not being eradicated by AI, it is being aggressively elevated to counter this exact threat.

SPEAKER_02

Okay, so let's reframe that role. I love this concept. Think of your software teams less like traditional manufacturing factories and more like editorial boards.

SPEAKER_01

The transition from engineer to editor. That is the next decade of software talent in a nutshell.

SPEAKER_02

Imagine your best developer spends less time typing code and more time editing what the AI builds.

SPEAKER_01

The editor analogy maps so perfectly to modern software development. Think about an editor-in-chief at a major newspaper. They don't physically type every sentence of every article right now. Their value is in maintaining the coherence of the publication, enforcing truth, calibrating the narrative, quality control. As AI assumes the raw generation of syntax, your developers have to sit in that editorial seat.

SPEAKER_02

Let's break down the granular math of a developer's day because this is the operational reality. Pre-AI, a classic senior developer, probably spent about 65% of their time actively writing code.

SPEAKER_01

Translating human logic into machine syntax, yeah. And maybe 50% on planning, and the rest on meetings and debugging.

SPEAKER_02

Post AI, that profile completely flips. Writing code drops from 65% down to roughly 10%.

SPEAKER_01

Which is a massive shock to the system. But that vacuum, that 55% reduction in typing, is filled by judgment and evaluation.

SPEAKER_02

So what are they doing with that time?

SPEAKER_01

Well, 40% of their time is now spent evaluating those parallel AI-generated architectures we talked about, stress testing edge cases, exercising deep systemic judgment. And the other part about 30% is dedicated to actively directing the AI, writing highly specific prompts, orchestrating different models. The craft is shifting from syntax to specification.

SPEAKER_02

I really want to pause here and speak directly to the senior software developers listening right now. If you've spent 15 years honing your coding craft, mastering complex syntax, it is entirely normal to feel threatened or anxious when an AI generates working code in three seconds.

SPEAKER_01

The professional shock is very real. But the reframe is so critical for your survival in this industry. Your fluency isn't being discarded.

SPEAKER_02

Right. It's not obsolete.

SPEAKER_01

Not at all. It's being abstracted upward. You are moving from the person who types the code to the person who knows why one solution is better than another.

SPEAKER_02

Because an AI can generate flawless code, but it doesn't have the institutional memory.

SPEAKER_01

Exactly. The AI has no idea why that legacy mainframe in the basement was configured with weird data constraints after an audit in 2015.

SPEAKER_02

It doesn't know the politics of your compliance department.

SPEAKER_01

Right. And it cannot feel the subtle emotional friction the user experiences during a clunky onboarding flow. AI fundamentally lacks systemic context. A developer's supreme value is now their context and their judgment.

SPEAKER_02

And this shift actually elevates an artifact that we used to treat as a totally boring formality. The product requirements document, the PRD.

SPEAKER_01

Now the PRD. For years, those were just bloated PDFs that product managers wrote and engineers barely skimmed before opening their IDEs.

SPEAKER_02

But now the PRD is paramount. It is the single most consequential artifact. Because if the AI is executing the code, your instruction set is the actual application.

SPEAKER_01

The prompt is the compiler target. If you get the PRD right, AI amplifies your clarity at speed.

SPEAKER_02

But if you get it wrong, it compounds your error at scale. It just perfectly builds the wrong thing.

SPEAKER_01

Exactly. The first draft that matters isn't the code anymore, it's the brief, which means the soft skills are now the hardcore technical skills.

SPEAKER_02

Customer empathy, organizational translation, lateral systems thinking. If your team doesn't have those, you fall into a massive trap.

SPEAKER_01

The ultimate trap. Industry veterans call it shipping your org chart.

SPEAKER_02

Yes, Conway's Law in Hyperdrive. Shipping your org chart happens when a company builds software that unconsciously mirrors its own internal departmental silos instead of the actual user's journey.

SPEAKER_01

It's so common. Think about a checkout flow for enterprise software. For the buyer, it should be a two-click process.

SPEAKER_02

But internally, the vendor needs finance approval, an InfoSec risk assessment, and legal review.

SPEAKER_01

Right. And because those departments don't talk to each other, the company uses AI to rapidly build a checkout flow that forces the user through three complex interfaces that perfectly mirror the internal bureaucracy.

SPEAKER_02

So what does this all mean? It's like putting a rocket engine on a car. If the steering wheel is locked in the wrong direction, you don't reach your destination faster, you just drive off a cliff at mock speed. AI accelerates the risk of building the wrong thing.

SPEAKER_01

And you can't just fix that with more tech. People try to use adversarial AI, having one AI build the app and another AI play the role of an angry user to critique it.

SPEAKER_02

Which sounds cool in theory.

SPEAKER_01

It's great for load testing. But AI cannot diagnose why a design is emotionally misaligned. It doesn't feel frustration when a legal disclaimer interrupts a purchase. Human empathy is the only reliable counter to that org chart trap.

SPEAKER_02

That brings up a massive misconception about user experience right now. Because of Chat GPT, everyone assumes the future of all software is just a blank chat box.

SPEAKER_01

Yeah, that graphical user interfaces are dead, but conversational UI is just a transitional layer.

SPEAKER_02

Because human cognitive speed hasn't changed. Reading text is sometimes way too slow.

SPEAKER_01

Exactly. Structured visuals, tables, buttons, sliders are already re-emerging inside chat interfaces. If I need to see pipeline growth across four regions, I don't want the AI reading me three paragraphs. I want a dynamic bar chart.

SPEAKER_02

And because building is so cheap now, we are entering the era of true hyper-personalization. A single product can have multiple distinct interfaces.

SPEAKER_01

Right. A power user gets a dense AI adaptive flow, while an AI skeptical user gets a reassuring traditional widget interface. The UI is just a personalized translation layer.

SPEAKER_02

Which leads us to a concept we really need to define for everyone: autopoietic software.

SPEAKER_01

It's a great term.

SPEAKER_02

Yeah, autopoietic essentially means self-evolving or self-creating software. It's a living application that exists in a continuous loop of feedback, evaluation, and code base adaptation without waiting for manual human release cycles.

SPEAKER_01

The idea of a finished project is becoming totally obsolete. Let's trace the mechanics of this. It's a continuous feedback loop. Phase one, the user gives feedback either by typing a complaint or just rage clicking a broken module.

SPEAKER_02

Phase two, the AI agent interprets that telemetry and synthesizes it.

SPEAKER_01

Phase three, the AI actually edits the code base to fix the friction.

SPEAKER_02

And phase four is the rigorous sandboxing, right?

SPEAKER_01

Crucial step. The agent deploys the new code into a shadow environment, runs automated regression tests to ensure nothing broke, and if it passes, it deploys it to production.

SPEAKER_02

So phase five, the user logs in an hour later and the app has natively adapted. Take a global logistics platform, for example. Say a truck driver keeps missing a turn because the reroute alert on their tablet is too low on the screen.

SPEAKER_01

Under the old model, the driver logs a complaint and it sits in a database for months until a product manager looks at it for a version 2.0 release.

SPEAKER_02

And meanwhile, the company bleeds millions in fuel efficiency. But with autopoyadic software, the AI identifies the anomaly instantly, rewrites the display logic to move the alert up, runs the tests, and updates the dashboard for the whole fleet on the fly.

SPEAKER_01

If your software is treated as a static asset, you will be outpaced by products that behave like living organisms.

SPEAKER_02

But there's a cultural risk here, isn't there? Current AI tools are heavily optimized for isolation. They are strictly single player.

SPEAKER_01

The AI tunnel risk, yeah. They optimize for a solo developer wearing headphones in a dark room, generating massive logic blocks alone.

SPEAKER_02

Which means you lose the collective creativity, the lateral thinking, the shared context of a team. So if everyone has their own private AI workbench, how do we keep collaboration vibrant? And more importantly, how do we let our best users, the DeV users, help build the product without handing over our intellectual property?

SPEAKER_01

This is where we have to explicitly define the open commit model. It's an AI-mediated community contribution model.

SPEAKER_02

Aaron Powell How exactly does that work for CTOs worried about IP?

SPEAKER_01

The source code remains completely closed. It is proprietary behind strict firewalls, but you grant authorized AI agents access to it. Okay. So a community member, maybe a major enterprise partner, proposes a new feature in natural language. They say, I need an export tool that formats to our specific ledger.

SPEAKER_02

And they submit that to the AI agent.

SPEAKER_01

Right. The AI evaluates the proposal, implements the code safely behind the firewall, runs penetration tests, and then integrates it, pending a final human editorial review by your team.

SPEAKER_02

That is brilliant for business leaders. You can reward those external contributors with utility tokens or credits to offset their sauce costs. It's a new form of maker culture without any of the IP risk.

SPEAKER_01

It builds incredible loyalty. But as AI gets closer to the end user, we have to establish boundaries. Which brings us to the key moments principle.

SPEAKER_02

Let's define the key moments principle. Simply put, not everything should be automated. Humans must be kept in the loop for the moments that actually matter.

SPEAKER_01

Our VP of digital innovation, Rafa Jimenez, summed this up perfectly. He said, AI can handle the long tail of transactional engagement, freeing the same people to give their full attention to the key moments that actually matter.

SPEAKER_02

So what actually is a key moment?

SPEAKER_01

It's an inflection point where trust is formed, problems escalate, or relationships are made or broken.

SPEAKER_02

Like resetting a password. Purely transactional. Automate it completely.

SPEAKER_01

Right. But a complex onboarding for a new seven-figure enterprise client, or a critical support failure where a client thinks they locked their data.

SPEAKER_02

If a user is panicking about data loss and you route them to a hyper cheerful AI chatbot, you will destroy that commercial relationship instantly.

SPEAKER_01

Exactly. Frictionless AI perfection causes more psychological damage and high anxiety moments than a slightly slower, empathetic human response. Trust is a deliberate design outcome.

SPEAKER_02

How do you design for that trust?

SPEAKER_01

Two ways. First is legibility. Clearly signal when AI is being used. Don't hide it. Opacity destroys trust the second it's discovered.

SPEAKER_02

And the second.

SPEAKER_01

Graceful escalation. Systems must know their own limits and hand off context to a human smoothly before the user gets frustrated.

SPEAKER_02

But that only works if the human team is ready. Which brings us to the org of the future and cultural readiness. How do CTOs and HR leaders need to hire differently now?

SPEAKER_01

Coding is still necessary, but no longer sufficient. You ought to hire for critical evaluation, customer empathy, and the ability to facilitate cross-functional conversations.

SPEAKER_02

We need blended teams, AI native talent for speed, and seasoned engineers for deep system integration knowledge.

SPEAKER_01

And if you're a tech leader, your role has shifted from a factory resource manager focusing on headcount to an editor-in-chief focusing on vision and curating output.

SPEAKER_02

But the hardest part isn't the tech, right? It's the mindset, psychological safety. Because it costs nothing to build 10 prototypes now, the fear of failure should drop.

SPEAKER_01

But the muscle memory of caution is deep. Leaders must explicitly protect this culture. You have to reward learning and normalize discarding prototypes without any stigma.

SPEAKER_02

You suggest starting an internal ritual called the digital pulse, right?

SPEAKER_01

Yeah, a recurring conversation where devs, product, and business talk honestly about AI's impact, surface their fears, share experiments, and align on where humans stay in the loop.

SPEAKER_02

All right, we have covered so much ground today. To synthesize this whole conversation, I want to leave you, the listener, with five crisp prompts to take back to your executive meetings. Number one, talent and roles. Are your developers treated as ticket takers or as editors and architects of meaning?

SPEAKER_01

Number two, org design. Do your product teams look more like software factories or more like editorial boards with high judgment density?

SPEAKER_02

Number three, experimentation muscle. If you can build five to ten prototypes in the time it used to take to build one, do you have enough good questions to decide which ones matter?

SPEAKER_01

Number four, defuser readiness. Are you ready for your users, both inside and outside the company, to become active co-creators of your software?

SPEAKER_02

And number five, culture and psychological safety. Have you talked honestly with your engineers about how AI changes their roles and how you plan to support them in that shift?

SPEAKER_01

I want to leave you with one final provocative thought. The gap between an idea and a product is rapidly approaching zero. But the gap between a good idea and a meaningful one has never been wider. The organizations that realize they are now in the business of meaning rather than the business of building will be the ones that define the next decade. Because eventually, autopoietic code bases will optimize themselves beyond human readability. They will become alien to us. The transition to editor isn't just strategic, it will be a technical necessity.

SPEAKER_00

And that's it for this episode of Digital Pulse. Now bringing it back to a human voice to say, thank you for listening. Remember to subscribe and follow Digital Pulse. Real Insight, Zero Noise.