The Cost of Manual Checking

Start with the most direct cost: the time qualified professionals spend checking their own work against rules. Before a drawing set goes anywhere, the architect and consultants must satisfy themselves that it complies - that corridors meet minimum clear widths, that every habitable room has the required light and ventilation area, that travel distance to a fire exit is within limits, that setbacks, ground coverage and floor-space index are respected, that stair and ramp dimensions are legal. None of this is design. It is verification: reading a clause, finding the relevant elements in the drawing, and comparing a number to a threshold, over and over.

The labour is heavy because it is repetitive and interruptive. A single change - moving a wall, adding a room, revising a floor plate - can invalidate a dozen earlier checks, and each must be redone by hand. Because the rulebook is vast and fragmented across the National Building Code, local byelaws and several IS standards, much of the effort goes into simply looking things up and holding them in mind. And because it competes for the same hours as designing, self-checking is often deferred, rushed near a deadline, or done once and not repeated after the design moves on - which is exactly how errors survive.

Think of it as a tax rate on professional time. On a complex project a meaningful fraction of the design team's hours goes not into shaping space but into the mechanical bookkeeping of compliance. Those are among the most expensive hours in the practice, spent on work that is genuinely necessary but largely rote for the checkable rules. That is the target. The judgement-laden part of checking - deciding whether an arrangement is actually safe or usable - is real professional work and stays human. But the part that is 'measure this, compare it to that number, flag if it fails' is a mechanical burden a structured model and an encoded rule can shoulder, freeing the professional's scarce, costly attention for the design decisions and the judgement that only a person can make. The first cost of manual checking, then, is simply the price of skilled people doing arithmetic a machine could do.

The second bill is time, but on the authority's side. Once a design is submitted for building permission, a plans-examiner or scrutiny cell must read the drawings and re-run the same checks from the outside - comparing the proposal against setbacks, coverage, floor-space index, height limits, fire provisions and the rest of the local development-control regulations. This is careful, manual work, and there is a queue of other projects ahead of yours. In many Indian cities and elsewhere, that queue and the back-and-forth of objections and resubmissions can stretch approval over weeks or months.

Waiting is not neutral; it costs money continuously. Land and finance carry interest whether or not work has started. A stalled project delays revenue for a developer and occupancy for a client. Design teams cannot fully close out a job and must keep revisiting a file as queries trickle back. And each cycle of objection and resubmission restarts the wait: a clarification on a setback or a missed fire clause sends the file back into the queue behind everyone else. The mechanical, checkable part of scrutiny - does this number meet that limit - is a large share of what the queue is actually processing, and it is precisely the part that could be pre-checked instantly.

This is why approval delay features so heavily in ease-of-doing-business conversations: slow, manual, sometimes discretionary permitting is a well-known drag on construction. Automation does not remove the authority's judgement or its accountability - the authority still approves, and much of scrutiny is genuinely interpretive - but it can shrink the mechanical backlog. If a designer can self-check the quantitative development-control rules before submitting, fewer files arrive with obvious, fixable violations, fewer objection cycles are triggered, and the examiner's scarce time goes to the parts that need human judgement. India's move toward online single-window and common-application building-permission systems, some with automated scrutiny of drawings against byelaw parameters, is aimed squarely at this cost. The waiting itself is the point: every week in the queue is carried cost, and much of what fills the queue is exactly the checkable arithmetic that need not have waited at all.

The third and often largest cost is not the checking - it is the checking that fails. When a compliance problem is missed during design and surfaces later, the price of fixing it climbs steeply with every stage it survives. A violation caught at the sketch stage is a few minutes with an eraser. The same violation caught at detailed design means reworking coordinated drawings across disciplines. Caught at submission, it means an objection, a redesign and another trip through the approval queue. Caught on site, it means tearing out and rebuilding work already paid for. And caught after occupancy - an inaccessible route, an inadequate escape path - it can mean legal liability, forced alteration, or genuine danger to people.

This rising-cost curve is one of the most important ideas in the whole subject, and it is why *when* you check matters as much as *whether* you check. The violation is the same size at every stage; what grows is the amount of committed, coordinated, built work that must be undone to correct it. A half-metre setback encroachment is trivial to fix on a plan and ruinous to fix in concrete. An under-width corridor is a line to move early and a demolition later. Manual checking, because it is slow and interruptive, tends to happen in bursts near milestones - which means violations are often discovered late, on the expensive part of the curve, rather than early where they are cheap.

That is the real argument for checking early and often, and it is the cost automation most directly attacks. Not because a computer's judgement is better - for the checkable rules it is simply faster, so it can run continuously as the design changes and flag a new violation the moment it appears, while it is still cheap to fix. The saving is not mainly the checking time; it is the rework avoided by catching the problem on the left of the curve instead of the right. Every guide in this course insists an automated pass is never an approval and never a guarantee - but as an early-warning system that surfaces likely quantitative violations while they are still cheap, its value is measured precisely in the expensive late-stage rework it helps you never pay for.

The last cost is the hardest to see because nothing visibly goes wrong: it is what does not happen. Every hour of scarce professional attention spent on mechanical checking is an hour not spent on the things only a skilled designer can do - resolving a difficult plan, improving daylight, refining a section, thinking harder about how people will actually move through a space or whether an accessible route is genuinely usable. Compliance arithmetic crowds out design judgement, and because the arithmetic is urgent and concrete while good design is open-ended, the arithmetic usually wins the calendar. That is the opportunity cost, and over a career it is enormous.

There is a subtler version too. Because manual checking is tedious, it is often the first thing compressed under deadline pressure - skimmed, deferred, or trusted to memory. So the same cost that drains attention also degrades the checking itself: rushed verification misses clauses, which feeds straight back into the late-discovery cost. The mechanical burden and the error rate are two faces of one problem - a person doing rote work under time pressure both loses the time and makes the mistakes.

This is where the honest case for automation rests, and it is a case about *reallocation*, not replacement. If a structured model and a set of encoded rules can absorb the checkable arithmetic - run it instantly, run it again on every change, and surface the flaggable quantitative issues - then the professional's costly, finite attention is freed for the judgement-laden work that no rule engine can do: interpreting an ambiguous requirement, weighing a performance trade-off, deciding whether a design is genuinely safe and humane. The point of counting these four costs - designer hours, approval queues, late-discovered violations, and drained attention - is not to argue that checking is bad or that machines should take it over. It is to locate precisely the mechanical, checkable slice that is a heavy tax, and to make the case that automating exactly that slice, and nothing beyond it, buys back time, money and - most valuable of all - the attention that good design and sound professional judgement actually require.