Techbanq

Cloud · continuity

The engineer who knows everything is a single point of failure

You would never run production on one server with no backup, yet the knowledge of how production works lives in one head with no copy. A resignation, an illness or a long holiday turns every incident into archaeology. The fix is not a wiki mandate; it is engineering the knowledge and the access into the organisation.

Enterprise delivery credentials

  • ISO

    ISO/IEC 27001

    ISO/IEC 27001 certified

Certifications & registrations

The conversation

The bus-factor conversation everyone keeps postponing

We do not have an office in Saudi Arabia

We serve organisations there from our Riyadh office and our engineering hubs in Lahore and Karachi, travelling for the parts of an engagement that need to happen in person.

Every organisation with this problem knows it has it. The engineer is named in half the incident channels; holidays are scheduled around them; the honest answer to "what happens if they leave" is a nervous laugh. The conversation gets postponed because every framing of it sounds like an accusation — as if planning for their absence means doubting their loyalty, or preparing to replace them. So the risk compounds silently: every quarter they stay, the system grows, and the share of it that exists only in their memory grows with it.

The framing that works is the one that is actually true: this is a reliability problem, and the engineer is its biggest victim. They cannot take a real holiday, cannot be promoted out of operations, and carry every incident alone at any hour. Redundancy in the knowledge is what frees them — the same argument they would make about any other single point of failure. Put that way, the strongest engineers usually become the effort’s sponsor rather than its obstacle.

The method

Moving a system out of one head without a single interview

Sitting the engineer down to "document everything" fails twice: it produces pages nobody reads, and it tells them their memory is the deliverable. Knowledge moves when it is a by-product of real work.

STAGE 01

Pair on the living work

A second engineer works beside them on real changes and real incidents — not shadowing, doing, with the veteran advising. Context that would never occur to anyone to write down transfers because it comes up. A second person who has done it, not read about it.

STAGE 02

Write runbooks from incidents, not from memory

After each incident or maintenance task, the pair writes down what was actually done, while it is fresh. A runbook born from a real event is specific and testable; one written from recollection is an essay. Runbooks that were true at least once.

STAGE 03

Make documentation a side effect of change

Every change to the system carries its explanation with it — in the change description, the decision record, the diagram updated in the same commit. Nobody sits down to document; the trail accumulates because work happened. A record that stays current by construction.

STAGE 04

Rotate the duties that force learning

On-call, releases and vendor calls rotate through the team, with the veteran as escalation rather than first responder. The gaps this exposes are the syllabus: each one becomes the next pairing session. A team that finds its own gaps before an emergency does.

The security half

Concentrated credentials are the part a resignation letter cannot wait for

Knowledge concentration hurts you slowly; access concentration can hurt you in an afternoon. Wherever the answer to "who can get into this" is one name, continuity and security are the same problem.

01

Root accounts only one person can open

Cloud root credentials, domain registrars, the payment provider’s console — if the recovery email and the second factor both point at one person, the organisation is locked out the day they are unreachable.

02

Secrets that live on one laptop

SSH keys, API tokens and deployment certificates accumulated over years in one home directory. They belong in a vault with role-based access and an audit trail — not because the engineer is distrusted, but because laptops fail and people travel.

03

Vendor relationships routed through one inbox

The hosting account manager, the licence renewals, the support contracts — when every vendor knows exactly one contact, a departure severs relationships the business did not know it had.

04

Approvals that assume one particular approver

Deployment gates, firewall change approvals and access grants that name a person rather than a role. Each is a queue that stops moving the week they are away — and a record of how much authority was never formally delegated.

Our role

What we restructure around the veteran and what stays theirs

  1. 01

    The map before the plan

    We audit what actually routes through one person — systems, credentials, approvals, vendor contacts — and rank the dependencies by how badly each would hurt, so the effort starts where the exposure is worst.

  2. 02

    The transfer, run as engineering

    We set up the pairing, the incident-born runbooks and the rotation, and we do part of the pairing ourselves where the team is too small to spare a second engineer.

  3. 03

    The access, moved onto roles

    Credentials move into a vault, root accounts gain recoverable ownership, approvals name roles instead of people — done with the engineer, whose knowledge of where everything is makes the work weeks instead of months.

The veteran keeps what makes them valuable — the judgement, the seniority, the hardest problems; what the organisation gains is the ability to survive their holiday, and eventually their next role, without either being a crisis.

Common questions

What buyers ask

How do we reduce the dependency on our key engineer without pushing them out?+

By being honest about whose problem it is. The dependency is a reliability defect the engineer suffers from more than anyone: no real holidays, no promotion path out of operations, every incident theirs at any hour. Framed that way — and staffed that way, with pairing and rotation that visibly lighten their load rather than an interrogation of their memory — most senior engineers champion the work. What pushes people out is the other framing: treating them as a risk, hiring their "backup" without telling them, or demanding documentation as if their memory were company property to be extracted. The order matters too: start with the access and credential half, which is mechanical and obviously sensible, and let the trust that builds carry into the knowledge half.

Our senior engineer says writing documentation is a waste of their time. Are they wrong?+

About the kind of documentation they are picturing, they are largely right. A standalone documentation project — weeks spent writing descriptions of a system from memory — produces text that is stale on arrival and read by nobody, and it consumes exactly the person whose time is scarcest. The version that works costs them almost nothing: runbooks written by the pair immediately after a real incident, decisions recorded in the change that implemented them, a diagram updated in the same commit that changed the architecture. The veteran’s role in that model is reviewing and correcting, not authoring from scratch. If the argument continues, agree on a test: pick one recurring task, have someone else attempt it from the existing material, and let the result settle whether the current state is sufficient.

What should happen to production access on the day a critical engineer resigns?+

If the preparation happened, the day is administrative: their accounts are disabled, the vault rotates the secrets their roles touched, vendor contacts were already shared, and the runbooks and pairs carry the operational load through the notice period. If it did not, resist the tempting shortcuts. Do not copy their credentials into a spreadsheet, and do not ask them to hand over personal accounts — negotiate a structured handover instead: root-account ownership transferred properly, secrets moved into a vault and rotated, one pairing pass over the systems only they have touched, priced by how much of the estate routes through them. It is an uncomfortable position to negotiate from, which is the strongest argument for doing the work while nobody is resigning.

Plan the succession with us

A dependency review that names what only one person can currently do — systems, credentials, vendor contacts — and the order in which each dependency should be shared.

Start the conversation