9 underrated engineering skills in this ai era

Beyond coding: Underrated skills that'll help you stand out as an engineer while AI takes more of our day-to-day code writing. Not many of us are thinking about #6.

Generative AI has effectively rewritten the rules of software engineering in the last year. This has caused many in our industry (including me) to wonder if our jobs are in jeopardy.

As Senior+ engineers, coding is only about 30% (if we are lucky) of how we spend our time.

With ai handling more and more of the raw coding part of our job – what’s left for us?

The path to standing out isn't just cranking out more code / features. It’s doubling-down on nine key areas that will help you increase your impact and value as an engineer in this new age of ai.

Let's dive in... 👇🏼

1/ Communication

You’d almost think this would become less important, right?!

Ai can spew out 1000s of lines of a blog post in seconds.

However, verbal communication is where so many engineers struggle.

These days you can sorta get around that as each member on the team has different skills and strengths and can help balance each other out.

In the future I can imagine daily standup consisting of reviewing the 3 or 4 projects that a couple seniors are managing and building with the help of ai agents, juniors, and product docs.

In this future, clear thinking and communication is key.

• Presenting demos

• Talking with cross-functional stakeholders

• Giving ai detailed context and overviews to improve its output

• Translating engineering terms and concepts into things others can understand

• Bridging the gap between product, business, engineering, support to help teams build out the right product for its users

Communication is one of those skills that has infinite value. Working on it levels you up as a human in so many different areas of life.

2/ Code review

As ai and engineers using ai generate more code, performing good code reviews will be key to our success as engineering orgs.

The more advanced or broad the code, it’s harder to know if it’s correct.

Sure ai can refactor our entire codebase to add missing types and tests, but how do we know if it was hallucinating or not?

This is where really through code review comes into play.

Photo credit: https://www.morling.dev/blog/the-code-review-pyramid/

As an engineer in the age of ai, it becomes more and more important to focus on Implementation Semantics and API Semantics.

Here are some questions you should be asking:

• Does it satisfy the original requirements and intent?

• Is there unnecessary complexity to the generated code?

• Are errors handled properly and appropriate logging / monitoring added?

• Is it secure? (big one with ai generated code)

• Are the dependencies used up to date and supported? (ai often pulls in outdated and insecure dependencies)

I know there are ai code review tool apps being worked on that will likely solve the top of the pyramid – code style and tests – in my mind that’s great!

Let an ai show where missing types and spelling mistakes and missing tests are and fix them, and let engineers focus on the big picture and how that code integrates into our system and product.

3/ Observability

As more ai generated code is deployed, we’ll need better e2e observability of our systems.

The first thing is just how is the new code working in production?

Do we have all the appropriate overview dashboards, anomaly detection monitoring and immediate rollback options configured and setup?

If not, that’s the first step. It’s so easy to get screwed without these.

The second step of observability is completely new – how do we monitor the percentage of ai generated code and its quality in our systems?

For this we’ll need new tools and new ways of operating.

I can imagine something like a Synk, Trivy, or Dependabot that expands to detect ai generated code and ensure its well-written, performant, readable, in line with company standards, secure, and using up to date dependencies.

4/ Legacy codebase understanding

AI is really good at green field projects, not so good at understanding large legacy codebases. It will get there eventually, but for a while engineers will have the lead here.

One of the most underrated skills of a strong engineer is being able to jump into a legacy codebase with anti-patterns, typos, security risks, and tech debt abounding – understand it, and then help improve it 1% every day.

This becomes even more critical in the AI era for several reasons:

Context is everything

AI struggles with the "why" behind architectural decisions made years ago. That seemingly odd pattern might exist because of a critical customer requirement or a performance bottleneck discovered through numerous incidents and custom configurations.

This historical context might be invisible to AI but invaluable to rearchitecting a robust solution.

Engineers who can reverse-engineer intent from implementation, build test coverage around critical paths, and gradually improve documentation create the foundation that will allow ai to assist in refactoring these systems.

Refactoring requires nuanced strategy

When faced with a legacy codebase, it's so easy to miss the whole picture or just get overwhelmed and want to rewrite the entire thing. Sometimes that's the best path, but other times the slow and steady improvement while painful, is the best approach forward.

AI often suggests the complete rewrite. It will be key to make a strategy call as an engineer with years or context / experience rather than just blinding accepting massive rewrites.

5/ Mentorship

One thing that’s blown my mind over the last few months is how much ai can help unlock quicker learning for me.

I can have a conversation with Claude or other about a topic I’m interested in or a difficult problem I need to solve. Within 30 mins I feel like I’ve just leveled up massively in my understanding of the problem or new concept.

Here’s the interesting part though – most of the things I’m learning or asking ai about are because of mentors or things I’m thinking about or listening to. It’s not just random things that popped into my brain.

If we think back to pre ai – circa 2022 – we all relied on mentors to help us learn and grow. Here’s the thing about ai that’s different from mentors – it’s only as good as the questions you ask it and the context it has.

One of the amazing things about Seniored ICs is how they can take in your current skills, experience, and interests, the companies projects, goals, technologies, processes, weaknesses, their experience, industry trends, etc. and use all that context to help you level up.

You don’t know what you don’t know. vs. a senior engineer knows what they know and what you don’t know.

Mentorship + using ai for learning + practicing on real world challenging problems =

a massive growth unlock 🚀🚀🚀

6/ Influence, Strategy and Storytelling

One of the most underrated skills of Senior+ engineers is the ability to influence others. To craft a strategy, a technical plan, and a narrative about the “why”, the “what”, and the “how” to guide a team in a new direction.

Without this you are left with sharing ideas and trying to get work on the roadmap and constantly getting rejected or hearing “ok we hear you, maybe next quarter…”

Let’s look at an example:

System Design / Architecture

Let’s say you want to propose adding a new tool to your stack. You are hitting up against the limits of what Postgres can support for your product and want to propose adding in elasticsearch for advanced querying and searching.

Currently you are using a simple `ILIKE %SearchTerm%` fuzzy search which works ok for simple searches on indexed columns in your crm, but is falling apart for more advanced searches.

Not to mention in the future your product team wants to add in natural language query ability 😅

Here’s the thing, building out the ETL pipeline to transform and get your data into postgres, and building out the querying capabilities is going to take 6+ months.

How are you going to get buy-in for that massive investment when the value today is “better search results.”

As a Senior+ engineer, your strength comes not just from designing performant systems, but your ability to tell compelling stories about architectural decisions, tradeoffs, and what’s needed to evolve for your product and users.

Getting buy-in for the right things is key to your success and your company’s success.

As ai starts to handle more of the implementation details of code writing, the "why" behind design choices almost becomes more important than the "how".

Engineers who can build trust and influence and a proven track record will stand out and be assigned to all the interesting and important projects.

7/ Security and Risk Assessment

This is related to code review, but I felt like it was worth it’s own section.

As more and more ai code gets deployed into production, monitoring it for security risks and vulnerabilities will become critical.

Shipping api keys into production becomes so much easier if a model just slips it into one of the 5 files it edited and you don’t have time to review them all, or a security vulnerability tool setup to catch that.

Another example to look out for is ai using old libraries and techniques when it generates new features and components.

Ai has been known to pull in old versions of something like create-react-app that has 100+ vulnerabilities right out of the gate of starting a new project.

As ai agents come into the picture and not only handle the code generation, but also code review and deployment, we are going to need new processes and tools for ensuring our products and codebases remain secure.

I think security-minded engineers will only stand out more and become more in demand as generative ai usage grows.

8/ Product-minded engineering

As ai replaces more of the “raw coding”, the product thinking side of our jobs becomes more important.

Anyone can command ai to crank out a “compelling sales site”, or a “clean code formatted REST api backend for these database tables”…

But what is behind that sales site or those database tables? What is your compelling product that solves users pain or needs in a way others aren’t?

Connecting the dots for your users is a key part of our job as “product-minded engineers”. That’s not just the job of PMs.

Just like it’s not the job of designers to drop pixel perfect mocks to us and then we build them. Just like it’s not our job to code up a new feature and then throw it over the fence for QAs to do all the testing.

The cross-boundary thinking and perspective we have as engineers is one of the ways we can have massive impact.

That creative idea no product person thought of because they couldn’t see the code and systems and where they were all pointing towards.

That inefficiency you spotted and optimized that lowered the onboarding time by weeks because you had deep into into customer tickets, what was possible in your systems, and the skills you had on the team.

Those times are where you show super strength as an engineer. That’s where you are shinning as a “product-minded engineer”.

These skills are only going to become more and more valuable as ai abstracts away more of the “raw doing”. Guide AI tools toward solutions that actually solve user problems, not just technical challenges.

9/ Ai tech stack engineering skills

These days becoming an “ai engineer” doesn’t requiring writing and tuning custom models.

The bar has lowered for learning how to build ai into valuable products and experiences for users.

And no I’m not talking about just adding the ol’ ChatGPT wrapper chatbot to your app.

Less than 10% of engineers have learned how to do prompt-chaining, prompt workflows, how to build / tune ai agents, RAG workflows, etc.

If you can start learning these skills and build up your understanding, you’ll stand out in your organization and for future job opportunities as an “Ai Engineer.”

Here are some resources to help you get started:

• RAG: What is it? How to use it? (intro)

• Applied AI Software Engineering: RAG (paid deep dive)

• How do AI software engineering agents work? (paid deep dive)

• Ai Engineering by Chip Huyen (deep-dive book)

• A deep-dive into ai engineering from an ai researcher / author (youtube)

• How my friend Jordan is building an AI side project using AI in 2025

• Anthropic: Building Effective Agents

• Building your own 2nd brain ai assistant (awesome free tutorial)

• AI is eating saas (interesting read)

• Prompt Engineering Interactive Tutorial

• Prompt Engineering Guide / Mini Course by DAIR.AI

• Prompt Chaining Tutorial with code examples

• Building Gen AI Prompt Workflows with Human Signoff

---

Well, that’s it for this week!

I hope you found today’s newsletter practical, thought-provoking, and helpful to your career growth.

If you have any helpful resources / thoughts on ai… I’d love to hear from you in the comments!

Until next week 👇🏼

Catch me daily on LinkedIn where I talk about everything software engineering, startups, and growing in your engineering soft skills.

– Caleb

P.S. Don’t forget to like, comment, and share with others if you found this helpful!