Why One Hiring Process Quitly Becomes Seven

Ask three people on your team how a hiring requisition really gets approved. The number of different answers you get is quietly costing you more than any line item shows. Here is the mechanism, the true cost, and why the fix is almost never the one teams reach for first.

Here is a diagnostic you can run before lunch, and it will tell you more about the health of your talent function than most of the dashboards you paid for.

Ask three people on your team to describe how a hiring requisition gets approved. Not the version written on the wiki. The version they actually follow on a Tuesday, when something is urgent and the clock is running.

If the three answers line up, stop reading. Your operation is in rarer shape than you think.

If they diverge, and they almost always diverge, you have just found something worth understanding. Hold that result for a moment. We will come back to what it means. The more interesting question is why they diverge at all.

One process becomes seven

It helps to name the gap the way Anju did in opening the roundtable.

One process becomes seven

The distance between the diagram and the Tuesday is the entire subject. So start with the part most diagnoses get wrong from the first step. Nobody on your team is trying to break the process.

Pratik framed it in a way that reframes everything that follows.

Inconsistency isn’t a failure

Hold onto that, because it points the fix somewhere most teams never think to look. People want to do the right thing. The trouble is that the right thing is rarely the most available thing when there is no time. A manager who cannot find the current approval policy reaches for last year’s email, because last year’s email is right there. A recruiter under pressure skips a step that felt optional the last time it was skipped without consequence. None of this is carelessness. It is what capable people do when the system makes the correct path harder to find than the workaround.

Then the workaround gets taught.

This is the part most operators underestimate. A new hire does not learn the documented process. They learn the process from whoever trains them, and that person is often passing on their own adaptation, in good faith, as if it were the standard.

Keith has watched this happen across enough programs to stop being surprised by it.

One process quitly becomes seven

Tribal knowledge has a flattering name. It is not knowledge. It is interpretation, inherited and slightly mutated at every handoff. And the moment interpretation replaces a defined path, you lose the one quality that makes an operation an operation. Predictability.

One process. Seven realities.

The cost no one budgets for

Here is where it gets expensive, and not in the way finance would catch. Pratik, who thinks about this in terms of flow and waste, makes a point worth sitting with. The real cost of inconsistency almost never shows up as a line item. It shows up as everything around the line item.

• Trust erosion. A manager checks in at every stage, because the output feels random and trust has quietly drained out of the process.

• Escalations. The documentation says one thing and the work went another way, so small gaps turn into avoidable reviews.

• Cycle-time creep. Exceptions stopped being exceptions and became the actual process. Every step quietly takes longer.

• Rework. The cost that compounds. Every inconsistency you tolerate this quarter becomes a correction project the next, and the cleanup never ends because the cause was never addressed.

There is one cost that outranks all of these, and it is the one most likely to end quietly rather than loudly.

It is credibility. When a business owner asks where their requisition is, and the honest answer changes depending on who they happen to ask, that is the moment a talent function stops being trusted to run itself. Inconsistent processes produce inconsistent expectations. And once expectations are inconsistent, nobody in the building agrees on what good even looks like anymore. Credibility lost inside an operation is slow and expensive to rebuild. Most teams never fully do.

What inconsistency actually costs

The fix everyone reaches for first

So the instinct takes over. Inconsistency must be a knowledge problem. People have forgotten the process. The answer must be more training.

This is the most expensive wrong turn in the entire sequence.

Training treats the symptom as the disease. It assumes that if people simply remembered the process correctly, they would follow it correctly. So you run the session. Awareness spikes. For a few weeks, behavior improves. Then pressure returns, priorities shift, memory does what memory does, and the old workarounds resurface. So you schedule another session. The curve repeats, indefinitely, and somewhere along the way you start mistaking the treadmill for a strategy.

Pratik’s framing is the one that lands. You do not fix inconsistency with more training. You fix it in the design. A process that depends on human recall to stay consistent will always drift, no matter how good the training was. The fix is not better memory. It is a process built so that memory is never required in the first place.

Training fades. Design holds.

Guided beats remembered

When a process is guided rather than recalled, the entire experience changes. The shift is small to describe and large in effect: remembered becomes guided.

The manager does not retrieve the next step from memory. The step is presented to them, in sequence, at the moment it is needed. Intake, assessment, evaluation criteria, decision points, each one defined and delivered rather than recollected. The process no longer varies based on who is running it, or how long they have been around. Every person moves through the same path and produces outcomes you can actually compare.

That last part matters more than it sounds. Puneet’s point on this is sharp. When everyone evaluates talent in their own way, you get different outcomes by definition. Consistency in how candidates are assessed is not bureaucracy. It is the only thing that makes the results mean anything, and it is also how bias quietly leaves the room. Same path, same criteria, comparable outcomes, regardless of who is in the chair.

Which brings us to the word everyone wants to lead with, and almost everyone misuses.

AI is not the magic

There is a version of this conversation where AI is the hero. That version is worth resisting, because it is the version that produces the most expensive disappointments.

AI does not repair a broken workflow. It cannot. Point it at a process that exists in seven conflicting versions and you do not get consistency. You get a faster, more sophisticated version of the same disorder, now harder to untangle because it runs at scale.

Workflow is the magic. AI is not.

Puneet, who leads the AI side of this work, is precise about what that leaves AI to do.

Workflow is the magic. AI is not.

Removing ambiguity, not supplying brilliance, is the job. What AI does, when the structure beneath it is sound, is enforce.

• It holds the designed process in place, so the version you intended is the version that runs.

• It standardizes how every candidate is assessed, so the result does not depend on who happens to be evaluating.

• It keeps the process honest at scale, regardless of who is managing it on a given day, or how busy that day is.

The organizations that get real value from AI in talent operations almost always solved the structural problem first. They had a process worth scaling. AI became the layer that keeps it honest. The ones who adopt AI as a substitute for the design work end up roughly where they started, paying more for the privilege.

So the useful question was never how to apply AI to a talent process. It was the quieter one underneath. Is the process coherent enough to be worth applying anything to at all?

Workflow is the magic. AI is not.

The list that is quietly six years old

The same logic explains a failure most enterprises live with and rarely name. Talent pools decay. The intention to keep them current is always there. The bandwidth is not. Months pass. People change roles, change skills, change their minds about moving. And when the moment of need finally arrives, the pipeline reflects an old reality instead of the current one.

Keith said something about direct sourcing that stays with me. Without ongoing engagement, a talent pool is just a big list of people. And that list could be six years old, not six months.

Nobody decided to let it rot. It rotted because keeping it alive depended on someone remembering to, and remembering does not scale. The fix is the same fix. Not a heroic quarterly cleanup, but a structure that keeps the pool current and the people in it engaged by design, so that the question of whether they are still interested is already answered before anyone has to ask.

Back to the three answers

Which returns us to the diagnostic from the top. When you asked three people to describe your approval process and the answers diverged, the divergence was never the real finding. The real finding was the reason. Not that your people are undisciplined, but that your process was built to be remembered instead of built to be followed. The gap between those two designs is where your cycle time, your escalations, your rework, and your credibility are leaking, quietly, every day, in amounts no dashboard is itemizing for you.

Inconsistency feels like a collection of one-offs. It is not agility. It is disorder. And disorder, at enterprise scale, is one of the most expensive things a talent function can carry.

The honest starting point is not a new tool. It is the question. Ask the three people. Listen for whether the answers converge or diverge. That answer will tell you more about the state of your operation than most of what you are currently measuring.