1. LandOffer
  2. Blog
  3. All articles
  4. How Hands-On Should an Engineering Manager Be?

People Management

How Hands-On Should an Engineering Manager Be?

Choose technical activities that improve understanding without putting the manager on the delivery critical path. Compare two complete weekly ledgers and explicit ownership boundaries.

10 min readLandOffer team

An engineering manager should stay technically engaged enough to understand the team’s work, challenge assumptions, and support sound decisions without becoming the person every delivery depends on. Choose the activity and its ownership boundary before choosing a coding percentage. Protect people management, planning, and coordination, then select technical work that can stop or transfer when those responsibilities require attention.

This guide from LandOffer.ai uses GitLab’s public practices as one employer example and two entirely fictional weekly cases. It does not establish a universal management role or measure any team’s results. The useful outcome is a specific technical-engagement agreement and a reconciled work ledger, including the work that remains unfinished.

An instrument-making leader studies a separate practice mechanism while a craftsperson retains the main assembly.

Define the contribution before counting the hours

Hands-on work can mean reading a design, inspecting a trace with its owner, reviewing code, pairing on a learning exercise, or implementing a change. These activities have different interruption costs and ownership consequences. Combining them into one coding target can hide whether the manager is improving understanding or taking over delivery.

Identify the decision, learning need, or support the team needs from your participation. Perhaps you need to understand an unfamiliar subsystem to ask better questions. Perhaps a decision needs cross-team coordination that only you can arrange. Perhaps an engineer needs a second opinion rather than a substitute implementer. Describe that need in terms of a result, not time spent looking busy in the repository.

GitLab’s engineering-management guidance expects technical credibility while advising managers to avoid critical-path coding and review more than they implement. Its engineering-manager role also includes people development, feedback, hiring, planning, and coordination. Those are that employer’s expectations; their transferable lesson is to evaluate technical activity alongside the obligations it can displace.

If your role explicitly combines delivery and management, make the trade-off visible with your manager. Fit the combined responsibilities within the agreed work capacity. Agree which commitments change, who owns the main delivery, and what happens when an urgent people issue interrupts the implementation block.

Set a boundary that survives your absence

Before taking technical work, identify its delivery owner, required reviewers, dependency, stop condition, and handoff material. “It is only a small task” is not enough. A small change can still become critical if nobody else understands it or if others cannot proceed until you finish.

Code review has a similar risk. Being available to review does not mean every change must wait for the manager. Name other qualified reviewers and the decisions that genuinely require managerial involvement. Otherwise, a technically engaged manager can become a review queue rather than an implementation bottleneck.

GitLab’s tech-lead guidance describes technical leadership as partnering with engineering and product managers, with ownership agreed between them. You do not need to import its role structure to use that distinction. Specify whether you are contributing context, making a management decision, or accepting technical ownership. Keep the designated technical owner visible so contributors know whose implementation decision and delivery plan they are following.

The test is practical: if you leave for an unexpected conversation, can the team continue using the agreed plan? If the answer is no, either change the task, arrange genuine alternate ownership, or acknowledge that you are taking a delivery commitment that needs capacity and management coverage. Do not hide the dependency behind the word “support.”

Worked case: technical learning with a pauseable implementation

The following people, work, time allocations, and results are fictional. Manager Beatrice leads six engineers maintaining package-metadata validation tooling. Technical lead Omar owns the current release’s rule changes. Beatrice wants enough domain context to discuss validation failures without pretending she has reviewed every implementation detail.

She considers four activities and records a completed decision for each.

Proposed participation Team need and owner Decision and boundary
Implement the release’s required rule migration Omar owns delivery; release depends on completion Decline personal implementation ownership; protect a focused engineer’s delivery time
Review two selected design/code changes Authors need feedback on assumptions and maintainability Join as a reviewer; qualified alternate Renata can proceed if Beatrice is unavailable
Reproduce one confusing validation message in a learning fixture Beatrice needs to understand the workflow Use a bounded pairing session with Omar; no production data or release dependency in the case
Draft a noncritical explanatory example for the contributor guide A small learning artifact may help future discussion Allow up to two hours; keep notes transferable; stop if management work requires the time

Renata explicitly accepts the alternate-review role within her existing review allocation; Beatrice is not assigning unlimited additional work. Omar remains accountable for the release’s technical decisions. The contributor-guide example is useful only if it is accurate and reviewed; the release does not depend on it.

Here is Beatrice’s filled weekly ledger. Each hour is assigned to one category; pairing time is not counted again as coaching or code review.

Work category Planned hours Illustrative actual hours
One-to-ones, feedback, and people follow-up 8 9
Delivery planning and dependency work 6 6
Stakeholder coordination 4 5
Hiring responsibilities 3 3
Administrative work 3 3
Technical review and bounded learning 6 5
Noncritical guide-example implementation 2 1
Reserved contingency → actual unexpected work 8 8
Total 40 40

The actual column is an authored result, not an observed employee time study. Beatrice uses two more hours on people and stakeholder work than planned. She reduces technical review/learning by one hour and the optional implementation by one hour. The eight-hour contingency allocation is fully used by other unexpected work; the two extra hours in people and stakeholder categories are funded by the two technical reductions, not counted again as contingency.

At the Friday review, the guide example is unfinished and explicitly parked with its notes. No engineer is waiting for it. Renata completed one review when Beatrice was unavailable, and Omar retained release ownership in the story. Beatrice can explain the validation workflow more precisely, but the case does not claim a measured productivity gain or prove complete technical fluency.

The next decision is to keep the selected-review arrangement and remove any implied promise to finish the example next week. If the team later makes that artifact a delivery requirement, it needs a new owner and capacity decision rather than inheriting Beatrice’s optional learning allocation.

Worked case: a week that calls for less implementation

A workshop leader reviews the whole assembly with its owner while a detachable practice piece remains safely set aside.

Manager Ellis leads an eight-engineer team taking responsibility for a new subsystem. Two engineers need onboarding support, and the team is negotiating interface ownership with another group. Senior engineer Cora owns the technical transition. Ellis knows the old subsystem well and is tempted to take the transition patch personally to “help it move.”

The completed choice is to decline that implementation, join selected technical reviews, and use managerial time to resolve the ownership disagreement. Cora accepts the transition role with four hours removed from an unrelated investigation by Ellis and the product counterpart. Cora’s capacity is a real constraint in the fictional plan, not an assumed free resource.

Ellis’s technical agreement has three parts: Cora owns the patch and design recommendation; Ellis attends one decision review and can ask for evidence; alternate reviewer Tomas can continue routine reviews without waiting for Ellis. No one is required to wait for the manager’s optional code comments. A disputed priority or staffing decision still comes back to Ellis.

Work category Planned hours Illustrative actual hours
People work and onboarding support 10 11
Planning and dependency resolution 7 8
Stakeholder coordination 5 5
Hiring responsibilities 3 2
Administrative work 3 3
Selected technical review 4 3
Manager-owned implementation 0 0
Reserved contingency → actual unexpected work 8 8
Total 40 40

The extra two hours of people and planning work are funded by one fewer hiring hour and one fewer technical-review hour. The recruiting coordinator, Soren, confirms the displaced interview can move to the following Tuesday with interviewer Nell, who accepts the one-hour slot within her existing Tuesday interview allocation. That is a changed commitment recorded in the case, not a claim that all original duties were completed as planned.

The technical review produces a useful ownership decision, but one failure-handling assumption remains unverified. Cora records the missing evidence and the next check; Ellis does not declare the transition technically complete. The week’s result is a narrower management contribution and a visible remaining risk, not an assertion that less coding always improves delivery.

At the Friday review, Ellis keeps zero personally owned implementation for the next planning cycle. If Cora cannot sustain the transition workload after the four-hour reduction elsewhere, Ellis must renegotiate scope or staffing. Taking the patch back without changing management commitments would simply conceal the capacity problem again.

Make the technical contribution inspectable

For a selected review, agree on the question you are helping answer. In Beatrice’s case, that question is whether the validation message accurately explains the rule and gives contributors a usable next step. Omar and the author still own the implementation. Beatrice can identify an unclear assumption without rewriting the change to match her preferred style.

For Ellis, the important question is which team owns the subsystem interface and who can supply the missing failure-handling evidence. Writing the patch would not itself settle that agreement. The manager’s technical understanding supports a coordination decision; it does not replace Cora’s assessment of the transition.

Keep a short result record for each contribution: question, evidence considered, decision owner, unresolved point, and next check. Avoid counting comments or commits as proof of useful leadership. A large review can consume engineer time without changing a decision, while a short, well-grounded question can expose a missing dependency. Ask the owner whether your involvement clarified the work and whether it created avoidable waiting.

Review technical engagement by its effects on ownership

Both planned and illustrative actual columns reconcile to 40 hours in each case; they are examples, not recommended calendars or observed time studies. Their technical allocations differ because the team needs and management obligations differ. Neither yields a universal percentage, and the hours alone do not establish whether the manager was helpful.

Review the actual consequences with the team and your manager. Did an engineer wait unnecessarily for your review? Did a one-to-one or promised feedback move because you were implementing? Did your contribution clarify a decision while leaving the technical owner in control? Is unfinished work documented well enough for someone else to assess it without reconstructing your thinking?

Watch for technical activity that becomes a refuge from difficult management work. A satisfying patch does not complete a delayed feedback conversation or resolve an unclear ownership boundary. Conversely, avoiding the codebase entirely can leave you unable to understand the risks you are coordinating. Choose an explicit way to learn rather than treating detachment as the only alternative to implementing.

A bounded experiment can be useful: agree on a small set of technical activities, protect management commitments, name alternate owners, and review the result after an agreed period. Stop or revise the experiment when someone starts depending on work you cannot reliably finish; assign the dependency explicitly before accepting another implementation task. Do not turn the initial allocation into an entitlement to keep coding regardless of what the team needs.

For your next planning conversation, bring a technical-participation choice and a complete work ledger, not a preferred coding ratio. Make the delivery owner, pause condition, and displaced work visible. Further workplace guides are available through LandOffer.ai; this guide does not assess actual employees or prescribe a staffing model.

Sources and Further Reading

  • GitLab: Engineering Management — Employer-specific technical credibility, avoiding critical-path work and reviewing more than implementing. Last modified March 19, 2026; checked October 25, 2026.
  • GitLab: Engineering Manager role — People, delivery, feedback and technical-awareness responsibilities. Last modified March 12, 2026; checked October 25, 2026.
  • GitLab: Tech Lead — Technical-lead partnership and explicit ownership division. Last modified March 19, 2026; checked October 25, 2026.