Example · Senior Software Engineer / Tech Lead
Coding agents now write the routine code, and companies are asking senior engineers to sign for it before it ships.
30 of your 40 hours still need you
- 22 stay yours
- 8 faster with AI
- 10 tools could take
A typical week at the middle pace. Your plan uses your own tasks.
Are senior software engineers safe from AI?
By our task-by-task count, 30 of a typical 40-hour week as a senior software engineer still need a person. Tools could already take about 10 hours. Routine code now comes from agents. Google's CEO said in April 2026 that 75% of the company's new code is AI-generated, and Sonar's January 2026 survey of 1,100 developers put the share of committed code at 42%.
Read more
That is the part of your week a model already does well: well-specified feature work, tests, refactors and PR descriptions. It adds up to about 25% of your week.
The rest holds, and some of it is getting heavier. After a run of outages in early 2026, Amazon told junior and mid-level engineers they would need a senior engineer to sign off on AI-assisted changes (Financial Times, March 10, 2026; Amazon disputes that AI caused the outages). New Relic's 2026 survey found 86% of tech leaders saying senior staff spend more time fixing code. Your bigger risk is a layoff round, or two years as the unpaid bottleneck for other people's agents. The plan turns review from a pile into a policy with your name on it.
The hard truth
More of your week will go to checking code.
AI writes a lot of code that is almost right. Someone has to catch the almost, and often that is you.
Sources (2)
- 66% of developers named 'AI solutions that are almost right, but not quite' as a frustration, the most cited. Stack Overflow 2025 Developer Survey, 2025-07
- 45% say debugging AI-generated code is more time-consuming. Stack Overflow 2025 Developer Survey, 2025-07
What stays yours
Deciding the architecture: what to build, what to refuse, what has to stay reversible
This takes context and accountability. The tradeoffs depend on history, team skill and money nobody wrote down, and a wrong call shows up a year later, which makes it expensive to check.
Signing off on risky changes: live-data migrations, auth, money movement
Someone has to answer for these changes. Companies are formalizing that a senior approves AI-assisted changes. The protection is delegated by policy, so it holds inside the company but can be moved by a reorg.
Leading the incident and owning the postmortem
This takes context, and you hold the accountability informally. You know which graph lies at 2 a.m., and the organization needs one name on the timeline.
Finding the cause when it is unclear
This takes context. Agents read logs fast, but they don't know that a job double-runs after a failover, or which alert has been noisy since the last migration.
Growing the people behind you
This rests on relationships. Someone has to train the next reviewer, and an agent can't sit with a nervous engineer through their first incident.
What's next for a tech lead?
Plan A is your best move inside the work you do now. Most sessions in the 13-week plan build toward it.
- A
Own change safety for agent-written code
Why it fits
Why you. You already review the risky changes and carry the pager, so you know where agent output fails in your system.
The gap. You need a failure rate split by agent-assisted and other changes, a written review contract, and evidence that it changed something.
The trade-off. It makes you the official reviewer, which can mean more review at first. If the policy later moves into automated gates, the title can go with it.
Other moves that fit
These are also good bets in this field. Read them if Plan A doesn't suit you.
Forward deployed or applied AI engineer
Why it fits
Why you. Tech leads already scope vague asks and ship under pressure, and this is that job with a customer in the room. Anthropic's Staff Forward Deployed Engineer posting (updated Oct 14, 2026) lists $280,000 to $320,000; OpenAI's Forward Deployed Software Engineer roles in New York and Seattle list $153,000 to $325,000.
The gap. You need shipped work with large language models that you can show, and evidence of working directly with customers.
The trade-off. Expect travel, customer pressure and a loosely used title. Some roles called forward deployed are presales. Most openings are at AI companies in large cities.
Engineering manager
Why it fits
Why you. You already interview, mentor and set direction.
The gap. You need formal people management experience, and a team that needs a manager.
The trade-off. Manager seats are what flattening cuts first, and spans are widening. You trade the most protected work of an individual contributor for a seat that is easier to remove.
What's a plan B for senior engineers?
Plan B is a job AI is creating that these skills already fit. One session a week in the plan trains you for it, so it is ready if your job shrinks or you want a change.
BPlan B, picked for this role
Agent reliability and quality engineer
You already set reliability targets, review risky changes and lead incidents, and this role applies the same work to agents.
Other options
AI evaluation specialist
Your code review judgment is exactly what turns into test sets and rubrics that say whether an AI system is ready to ship.
Developer AI enablement lead
You already write your team's review rules for agent code and coach engineers through them, which is this job across more teams.
In your own plan you can switch to any of the 47 roles.
The first two weeks
The first 2 of 13 weeks. Three sessions a week of 30 to 90 minutes, about 2 to 3 hours in all, plus one 40-minute Plan B session. In your on-call week, do only the first session. By week 13 you'll have a review contract and PR template your team uses, failure rates split by agent-assisted and other changes, and a memo asking for ownership of change safety.
- Week 1: Measure your week
What you do
- ALog your time for one week. Match each block to a task in this report's task list.
- ACount the pull requests (PRs) you reviewed in the last 30 days. Split them into agent-assisted and not, and note the median time per review.
- AList the last five times you were paged and the change that caused each one.
- BTake AI Capabilities and Limitations, a free Anthropic Academy course. Name three ways you have seen agents fail in your own work.
- Week 2: Draft a review contract
What you do
- ADraft a one-page PR contract for your team. Include a size cap, required tests, proof the code ran and a risk tier for each change.
- ATag the last 40 merged PRs with a risk tier.
- ABring the contract to your team's retro meeting and collect their edits.
- BWork through the first half of AI Fluency: Framework & Foundations, Anthropic's free course on working with AI. Write your own rules for what to hand an agent and what to keep.
Three courses from the plan
Checked 2026-10-15.
- Site Reliability Engineering, chapters 4 (Service Level Objectives), 11 (Being On-Call) and 15 (Postmortem Culture)Google · Free online
Yes, for Weeks 6 and 10.
- StaffEng guides: "Manage technical quality" and "Writing engineering strategy"staffeng.com · Free
Yes. Read those two guides only.
- State of AI-assisted Software Development 2025DORA (Google Cloud) · Free download (form required)
Yes, for framing the Week 12 memo.
Risks beyond AI: highLayoff and offshoring risk
Layoff rounds and offshore engineering centers have moved more senior seats in 2026 than coding agents have. Software postings are well below their 2020 level, tech cuts are running ahead of last year, and US companies keep opening engineering centers in India. Being senior helps, since most of the postings that remain ask for experience, but it doesn't stop a round from reaching your team.
Where this could be wrong
- Approving changes could become automated. If automated review gets good enough that a CI gate, the automated check before merge, can sign off, the sign-off role shrinks faster than we scored it.
- Shares assume a tech lead at a product company. At an agency or consultancy, feature work is a bigger share and so is exposure.
- The 75% figure is a CEO's claim, and Semafor noted the timeframes are hard to verify. Amazon disputes the reporting on what caused its outages.
- We rate layoffs above the AI risk based on 2026 cut counts and posting data. If funding loosens quickly, that risk falls faster than the AI one.
Also matched: senior software engineer, tech lead, staff software engineer. Research checked 2026-10-15. Page updated .