Skip to content

Knowledge base

Senior Engineering Skills

Senior engineers are judged less on code than on the skills around it: running meetings, writing design docs, leading without authority, picking battles and taking feedback. Agentic work raises the stakes, because design docs and shared conventions become the context AI agents build from.

Written by

Updated:

Software Engineering 5

Our position. We hold that the minimum bar for an effective agentic developer is closer to today’s senior architect than to today’s mid-level engineer. Career ladders built for scarce code output point at the wrong skill now: level on decision quality, not on how much work someone supervises (skills outrank titles (opens in a new tab)).

The 23 skills come from An incomplete list of skills senior engineers need, beyond coding (opens in a new tab), first published on 22 September 2020. They apply from senior to staff and beyond, and each carries a note on what changes when agents write the code.

How do senior engineers communicate and influence?

1. Running a meeting

How to run a meeting, and no, being the person who talks the most in the meeting is not the same thing as running it.

Running a meeting means setting an agenda, keeping people on track, making sure decisions get made, and following up on action items. The loudest voice in the room is rarely the one actually driving outcomes.

In the agentic era: Meetings increasingly exist to make decisions that AI cannot make autonomously. If you cannot run a meeting efficiently, you are wasting the one resource AI cannot replace – human judgment time.

2. Writing design docs

How to write a design doc, take feedback, and drive it to resolution, in a reasonable period of time.

A design doc is not a novel. It states the problem, proposes a solution, lists alternatives considered, and identifies risks. Taking feedback means actually changing the doc, not defending every sentence. Driving to resolution means setting a deadline and making a decision, even with incomplete information.

In the agentic era: Design docs are context for AI agents. A well-written design doc becomes the system prompt for every agent that touches your codebase. Poorly written ones produce poorly generated code. This skill went from important to load-bearing.

3. Explaining technical concepts to senior people

How to explain a technical concept behind closed doors to a senior person too embarrassed to openly admit that they don’t understand it.

This requires empathy, not condescension. The senior person has context you lack (budget, politics, strategy). You have context they lack (how the system actually works). Your job is to bridge that gap without making them feel small.

In the agentic era: Non-technical leaders are now making decisions about AI tooling, agent architectures, and automation boundaries. If you cannot explain what “context engineering” means to a VP without jargon, bad decisions get made at scale.

4. Communicating project status

How to communicate project status to stakeholders.

Stakeholders do not care about your sprint velocity or story points. They care about: are we on track, what are the risks, and do you need anything from me. Learn to translate engineering reality into business language without lying about it.

In the agentic era: AI-accelerated delivery changes timelines unpredictably. Communicating status when velocity itself is volatile requires even more skill. “The AI wrote it fast but we are still validating” is a status your stakeholders need to understand.

5. Convincing management to invest

How to convince management that they need to invest in a non-trivial technical project.

Management speaks ROI, risk, and timeline. You need to translate “we need to refactor the auth system” into “we are losing X hours per week and risking Y in security exposure.” Numbers beat passion every time.

In the agentic era: Convincing leadership to invest in context engineering infrastructure, proper CLAUDE.md files, architectural decision records, and agent-friendly codebases requires exactly this skill. The ROI is real but non-obvious to anyone who has not seen it work.

6. Influencing other teams

How to influence another team to use your solution instead of writing their own.

This is sales, whether you like it or not. You need to understand what the other team cares about, show how your solution addresses their needs, and make adoption easier than building from scratch. Nobody adopts your library because it is technically superior. They adopt it because it solves their problem with less effort.

In the agentic era: AI agents operate across team boundaries. Shared solutions, shared context, shared conventions – the cost of fragmentation multiplies when every team’s AI agents produce incompatible output.

7. Repeating yourself

How to repeat yourself enough that people start to listen.

Saying something once is not communicating. Saying it in a meeting, writing it in a doc, mentioning it in a PR review, bringing it up in a 1:1 – that is communicating. People are busy. Your message competes with hundreds of others. Repetition is not nagging, it is how organizations absorb information.

In the agentic era: Context engineering is literally the discipline of ensuring the right information is available at the right time. If you cannot do this with humans, you will not do it with agents either.

How do senior engineers lead and mentor others?

8. Mentoring at every level

How to mentor an early-career teammate, a mid-career engineer, a new manager who needs technical advice.

Each requires a different approach. Early-career engineers need guardrails and encouragement. Mid-career engineers need to be challenged and given ownership. New managers need honest technical advice without you undermining their authority. One mentoring style does not fit all.

In the agentic era: Mentoring now includes teaching people how to work with AI effectively. An early-career engineer who learns to pair with AI agents from day one develops differently than one who learned to code solo first. Your mentoring approach needs to account for this.

9. Leading without authority

How to lead a project even though you don’t manage any of the people working on the project.

This is the core staff engineer skill. You cannot tell people what to do. You can only make them want to do the right thing by setting a clear vision, removing obstacles, and making it easy to contribute. Influence beats authority every time.

In the agentic era: You are now leading both humans and AI agents. AI agents need clear context, well-defined boundaries, and good architectural guardrails – exactly the same things human teams need from technical leaders.

10. Helping someone get promoted

How to help someone get promoted.

This means giving them visible projects, advocating for them in rooms they are not in, giving them honest feedback about their gaps, and coaching them on how to demonstrate their impact. It is one of the most generous things a senior engineer can do.

In the agentic era: The skills required for promotion are shifting. Helping someone get promoted now means helping them demonstrate impact in an AI-augmented environment, where raw output volume is no longer the differentiator.

11. Teaching others to care

How to teach another engineer to care about that thing you really care about (operations, correctness, testing, code quality, performance, simplicity, etc).

You cannot force people to care. You can show them the consequences of not caring. Let them feel the pain of a production incident caused by skipped tests. Let them see the beauty of a well-designed system. Care is caught, not taught.

In the agentic era: Teaching engineers to care about code quality matters more than ever, because AI-generated code that nobody cares about reviewing becomes technical debt at machine speed. Whoever guards code health on a team that ships agent-written code depends entirely on this skill.

How do senior engineers navigate organizational politics?

12. Indulging senior managers

How to indulge a senior manager who wants to talk about technical stuff that they don’t really understand, without rolling your eyes or making them feel stupid.

Senior managers who want to talk about technology are your allies. They are trying to understand your world so they can support it. Meet them where they are. Answer their questions honestly. Correct their misconceptions gently. The alternative is managers who stop asking and start guessing.

In the agentic era: Every executive now has opinions about AI. Some are informed, many are not. Your ability to engage with their enthusiasm while steering them toward reality directly impacts whether your organization adopts AI sensibly or chaotically.

13. Picking battles

How to pick your battles.

You cannot fight every fight. Some hills are not worth dying on. The art is knowing which battles matter for the long term and which are just ego. Save your political capital for the fights that actually affect the codebase, the team, or the product.

In the agentic era: The number of potential battles exploded. AI code style, agent boundaries, context management practices, which tools to adopt – you cannot fight all of them. Pick the ones that compound.

14. Networking and gossip

How to get information about what’s really happening (how to gossip, how to network).

The org chart tells you the formal structure. Gossip tells you the real one. Who is actually making decisions? Which project is about to get cancelled? Who is about to leave? This is not about being political. It is about having the information you need to make good decisions.

In the agentic era: AI adoption decisions happen fast and often informally. Knowing that another team already tried and failed with a particular approach saves you months. The informal network is your early warning system.

15. Telling someone they’re wrong

How to tell someone they’re wrong without making them feel ashamed.

“I think there might be a problem with this approach” works better than “this is wrong.” Focus on the idea, not the person. Offer alternatives. Make it a conversation, not a verdict. The goal is a better outcome, not a victory.

In the agentic era: AI generates plausible-looking code that is subtly wrong. You will need to tell colleagues that their AI-generated solution has problems without implying they are incompetent for trusting it. This requires finesse.

Which personal habits keep a senior engineer growing?

16. Asking for help

How to get another engineer to do something for you by asking for help in a way that makes them feel appreciated.

“Can you help me with X? You know more about this than I do” works. “Do this for me” does not. People want to feel competent and valued. Frame your request as recognition of their expertise, not as a task assignment.

In the agentic era: AI can answer many questions, but knowing when to ask a human instead of an AI – and how to do it well – is a skill that separates effective engineers from isolated ones.

17. Listening without feeling threatened

How to listen to other engineers’ ideas without feeling threatened.

Someone else having a good idea does not diminish yours. If their approach is better, adopting it makes you smarter, not weaker. The engineers who cannot listen without defending are the ones who stop growing.

In the agentic era: AI will suggest approaches you did not think of. Engineers who feel threatened by a machine’s suggestions will fight every AI-generated recommendation. Those who can evaluate ideas on merit – regardless of source – will thrive.

18. Getting other engineers to listen

How to get other engineers to listen to your ideas without making them feel threatened.

Present ideas as explorations, not declarations. “What if we tried X?” invites collaboration. “We should do X” invites resistance. Show your reasoning. Acknowledge tradeoffs. Make it easy for people to say yes.

In the agentic era: Proposing AI-driven changes to established workflows triggers defensiveness. The ability to introduce agentic practices without threatening people’s sense of competence is critical for any engineering leader driving adoption.

19. Giving up your baby

How to give up your baby, that project that you built into something great, so you can do something else.

Your identity is not the code you wrote. Handing off a project means trusting someone else to care for it, even if they care for it differently than you would. Holding on too long stunts your growth and theirs.

In the agentic era: AI will refactor, rewrite, and restructure code you spent months crafting. If your ego is attached to specific implementations rather than outcomes, you will spend your career fighting the inevitable.

20. Taking negative feedback

How to take negative feedback gracefully.

Negative feedback is data, not an attack. Thank the person for telling you. Ask clarifying questions. Sit with it before reacting. The engineers who improve fastest are the ones who make it safe for others to give them honest feedback.

In the agentic era: AI code review tools will give you feedback constantly and without diplomacy. If you cannot handle blunt critique from a human, automated critique at scale will be miserable.

21. Finding interesting work

How to find interesting work on your own, instead of waiting for someone to bring it to you.

Interesting work does not appear on your Jira board. It lives in the gap between what the system does and what it should do. Read the codebase. Talk to users. Look at metrics. The engineers who find their own work are the ones who get promoted.

In the agentic era: AI handles the routine tasks. What remains is the interesting, ambiguous, judgment-heavy work. Engineers who can identify and claim that work are invaluable. Those who wait to be assigned tasks will find themselves assigned the ones AI cannot do – not because they are interesting, but because they are left over.

How do senior engineers deliver value and win buy-in?

22. Delivering incremental value

How to build software while delivering incremental value in the process.

Big-bang releases are a gamble. Shipping incremental value means every week, something gets better for users. This requires the discipline to slice work into meaningful increments, not just technical layers. “We refactored the database” is not value. “Users can now search 10x faster” is.

In the agentic era: AI accelerates implementation but does not accelerate validation. Incremental delivery becomes even more critical because the gap between “built” and “verified” widens when code is generated faster than it can be tested and deployed.

23. Crafting proposals

How to craft a project proposal, socialize it, and get buy-in to execute it.

A proposal needs a clear problem statement, a proposed solution, cost/benefit analysis, risks, and a timeline. Socializing means sharing it with key stakeholders before the formal review so there are no surprises. Getting buy-in means making people feel like they are part of the decision, not being sold to.

In the agentic era: Proposing AI-driven initiatives requires an extra layer of persuasion. You need to address fear, uncertainty, and legitimate concerns about quality and control. A well-crafted proposal for agentic adoption is part technical document, part change management plan.

How do these skills show a team’s agentic readiness?

These 23 skills map directly to the gaps we see in engineering organizations adopting agentic practices. Technical skill with AI tools is necessary but not sufficient. The organizations that succeed with agentic engineering are the ones where senior engineers already have these skills – or are actively developing them.

When assessing engineering teams for agentic readiness, evaluate these skills alongside technical competency. A team of brilliant coders who cannot run meetings, write design docs, or give each other honest feedback will struggle with AI adoption just as much as a team that cannot code.

Key assessment questions:

  • Can your senior engineers write design docs that would serve as effective context for AI agents?
  • Do your tech leads know how to introduce new practices without threatening their teams?
  • Can your staff engineers influence across team boundaries?
  • Do your engineers find interesting problems, or do they wait to be assigned work?

The answers tell you more about agentic readiness than any technical assessment ever will.

Resources

Articles

Posts