Excellence on Paper, Workarounds on the Floor
Why formally mature engineering firms still drift — and what quiet friction reveals about where capability actually sits
An engineering team delivers a project on time and on budget. The order is fulfilled, and the performance indicators stay green.
Six months later, the customer asks for a variation. Answering that request takes three weeks.
Nothing in the equipment failed. The person who made the original design decision had moved to another project, and the reasoning behind that choice had never been written down. In my experience, standard delivery procedures verify what was approved and shipped, rarely which design alternatives were rejected or why — so every formal gate can pass while the capability stays in a person rather than in the structure.
Leaders in industrial businesses recognise that three-week pause immediately. Fewer treat it as an early structural warning, because every formal report around it still describes a healthy system.
Excellence on paper
Earlier in my career, I spent a decade (1997–2007) assessing organisations against formal excellence models as a National Quality Award assessor, and in 2002 contributed to international assessments as a European Quality Award assessor. That work meant sitting inside dozens of formally mature institutions: documented workflows, structured review cycles, clean scorecards, and awards on the wall.
Across the organisations I assessed, one pattern surfaced repeatedly. An organisation could score well, document everything, win the award, and still be drifting in direction. Excellence on paper was not alignment in reality.
None of that makes excellence models or quality systems pointless; operational rigour matters, and many assessed firms used those frameworks to build genuine discipline. The blind spot lies in what leadership reads into a favourable result.
Excellence models evaluate how systematically processes are deployed and reviewed against stated goals, while compliance audits verify whether documented procedures are followed. Neither can guarantee that the market assumptions underneath those goals still hold, or that critical design judgment made it into the files. When metrics stay green, a high score or clean audit can delay those harder questions, because formal recognition itself becomes part of management’s reassurance.
What silence hides on an engineering floor
Across the industrial and engineering-intensive businesses I have worked in and served on the boards of — from automotive supplier operations to mobility, defence, and manufacturing technology — that reassurance is reinforced by how technical teams handle friction. Technical capability in these companies is often strong, even when the structure around it was built for a different scale.
When a process or handoff no longer fits the work, engineers rarely complain. They do not argue in steering committees or escalate every gap in a workflow. They quietly work around it, make a mental note, and keep the delivery moving.
In my experience, reading that silence as approval is one of the costliest mistakes a management team can make. Zero escalations does not prove a structure is sound; it can just as easily mean the damage is not visible yet. Behind a clean dashboard, engineers normalise workarounds — repeated revisions, constant *”could you just take a quick look?”* requests, and critical technical calls that keep landing on the same person.
Once relying on one specialist to clear friction becomes routine, dependency becomes the operating standard. The schedule holds and the escalation log stays empty, so nobody examines what kept delivery from slipping until that specialist is no longer in the seat.
The same structural gap shows up at departmental boundaries long before it reaches the income statement. Engineering sees a job as a technical difficulty; finance sees it as a cost centre; sales sees it as a delivery date. The customer experiences none of those functions separately — only one commitment.
When a customer has to ask separately after every delivery for the maintenance plan, the spare-parts list, or the reasoning behind a revision, the order was technically fulfilled. Each department met its own definition of the job, yet a predictable customer need fell between them because nobody owned the commitment end to end.
When the reflex answer is headcount
Eventually that accumulated friction slows visible work: variations take weeks to turn around, senior specialists turn into bottlenecks, and commercial commitments wait in a queue. In the boardrooms and leadership teams where I have seen this play out, the first reflex is often to diagnose a capacity shortage — hire more engineers, or add another layer of reporting.
I had to test that diagnosis directly when I stepped in as acting general manager of an engineering services company, carrying full budget and profitability accountability to the board. The organisation had strong technical depth, all of it historically directed inward to serve internal group programmes. When that funding model stopped working and internal services had to pay for themselves through external revenue, the rigid functional layout made the team look perpetually short-staffed — and hiring was the standard answer on the table.
Addressing that constraint under financial pressure required two distinct decisions, not one. First came a commercial and strategic question nobody had needed to ask under the old mandate: who is this capacity actually for Turning inward-facing engineering capability into an external revenue-generating service — including setting up an embedded R&D operation at a key customer’s site, which increased repeat project flow — changed the commercial equation while exposing how rigidly people were locked inside single-project functional boxes.
Get Samih Yedievli’s stories in your inbox
Join Medium for free to get updates from this writer.
Second came redesigning how work and roles were assigned so we could carry those commitments without hiring into the old structure. Instead of fixed functional departments, we moved to a project-based organisation where engineers took different roles across programmes according to the capability each job demanded. We met our committed customer programmes with a much smaller team, and continued delivering when key individuals later left.
Running customer programmes with a leaner team under commercial pressure carries real workload strain, and one turnaround does not mean every staffing shortage is structural — sometimes an engineering unit simply needs more people. Simply spreading the same scarce specialist across three projects would only multiply the bottleneck; what reduced dependency on any single individual here, in my reading, was breaking single-project silos so engineers built shared experience across active programmes before departures happened.
At leadership level, we closed the mandate on a separate application of the same principle: when my acting general manager term ended, decision rights were transferred in planned stages rather than on a single handover date.
Empty logs, scattered traces — and where AI actually belongs
Catching these blind spots before a funding model breaks or a specialist leaves does not require another compliance audit. It starts by distinguishing between what formal reports leave blank and where operational traces actually sit.
One of my working rules came out of an industry analysis that needed a sector figure published nowhere — not by the ministry, not by the industry association, not by the sector body. Instead of estimating it, the analysis reported that no published figure existed: absence of data is a finding, not a gap.
Inside an engineering firm, the mechanism differs in one important way: an escalation register or a post-mortem form already exists, and reads zero or routine. A clean log cannot separate genuine health from quiet workarounds on its own — which is why an unwritten design rationale discovered months later should be read as a structural finding, not a filing oversight.
That distinction also sets the practical boundary for how management uses artificial intelligence. AI cannot read a mental note an engineer never wrote down, or hear an unlogged ”could you just take a quick look?” conversation in a corridor. And if sign-offs or revisions already sit in one clean table, management does not need AI to count them.
Where AI earns its keep is across unstructured, fragmented records — linking patterns across revision notes, post-delivery customer correspondence, and approval trails that different teams record in different words across separate projects, while formal escalations stay at zero. Brought together in a privacy-preserving, anonymised way, those scattered records show leadership where friction and sign-off load are concentrating.
Concentration alone does not prove a process was bypassed; approvals piling up on one desk may reflect an informal workaround, or they may be the unintended queue of an authority matrix that deliberately centralises sign-off in one role. The tool surfaces where the load sits, and decisions remain human — management still has to examine why the work keeps converging there.
Underneath *The Depth Mirror* sits one alignment question: what will still be working when the decision-maker is gone? A leadership team can test the three blind spots traced in this essay — siloed handoffs, person dependency, and aged assumptions — through two simple counts and one review over the next month:
- Siloed handoffs (post-delivery requests): How many times after a completed delivery does a customer have to ask separately for something — a maintenance plan, a spare-parts list, a revision rationale — that should already have been built into the design?
- Person dependency (single-desk clearance): Across active projects, how many technical or commercial decisions could only be cleared by one specific person?
- Aged assumptions (post-mortem challenge): Next to every finding in your last project post-mortem, can your team write which management assumption that finding challenges — or did it stay on paper as an operational note?
Seeing where capability sits
As I write in The Depth Mirror: ”The issue is rarely visible failure. The issue is that drift becomes normalized before it becomes measurable.”
In the industrial and engineering firms I have assessed and led, drift rarely announces itself through broken machinery or failed audits. Far more quietly, formal procedures verify outputs and resourceful engineers route around friction while critical judgment concentrates in a few heads and strong technical capacity stays pointed at an older mandate.
When an organisation is caught in that pattern, what it needs first is not more capacity, and not another review confirming that the documented process ran. It needs the clarity to see where the capability it already has is sitting — before a routine variation request takes three weeks to answer.
Samih Yedievli, November 2026
Notes:
This essay extends The Depth Mirror: Why Organizations Continue to Work While Moving in the Wrong Direction ( thedepthmirror.com ), available on Amazon: ( amazon.com/dp/B0H6Q6KZ14 )
Visuals generated with Gemini; art direction by the author.
I prepared this essay with the support of Synapta OS, my AI-assisted thinking system. The argument, field observations, and final text are mine.
