We hit 100% automation on purchase order confirmations months ago. It wasn't the victory we thought it would be.
When we started, the goal seemed clear: automate the work. Extract the data from the mess — emails, EDI, images, Excel sheets — and process confirmations without human touch. It was hard. Parsing messy supplier data at scale is genuinely difficult. But we solved it. Then we solved automation on top of it. By every metric, we won.
And then we realized we'd been optimizing the wrong thing.
The number stopped mattering the moment we understood what automation was actually for. It wasn't about eliminating work. It was about moving work upstream — away from triage and toward resolution.
The Old Model: Triage Everything
Most wholesalers live here. A confirmation arrives. A human opens it. They eyeball it against the PO. They spot a deviation — price jumped three percent, delivery slipped two weeks, quantity doesn't match. Then they have to decide: does this matter? They check stock coverage. They check margin contribution. They check whether production depends on this item. Then they decide what to do.
If they're thorough, it takes time. If they're drowning — and most are — they skip it. They accept the deviation. Downstream, the problem compounds: goods arrive late, fill rates drop, and they compensate by holding more safety stock. Higher working capital. Lower fill rates. People exhausted.
Some wholesalers don't process confirmations with any real diligence at all. They just accept what arrives and deal with the consequences later.
What Lisa Does Instead
Lisa — our purchasing agent — doesn't just read the confirmation. She resolves it.
A deviation lands. She evaluates it against context: coverage, criticality, margin contribution, production impact. If the deviation is acceptable — if a partial early delivery maintains coverage, if a price variance is within tolerance — she decides. She accepts it and updates the plan. No human involved.
But if the deviation breaks coverage, she doesn't escalate it to you yet. She negotiates. She sends an email to the supplier asking for an earlier or partial delivery. She tries to fix the problem before you ever see it.
When the supplier responds, she evaluates again. If the new terms work — if she can find substitute materials with coverage, if warehouse transfers solve it, if any mitigation path keeps the supply chain whole — she executes the mitigation and closes the issue.
Only when she's exhausted her resolution paths does it come to you. And when it does, it lands as a risk — not a raw deviation. It includes what she tried, why it didn't work, and what it means downstream. It includes how critical the item is, what the impact is, and what your options are. Crucially, it includes a probable cause — not just "this confirmation is off" but "this is likely off because the supplier shipped from a different origin." The value was never in catching the mismatch. It's in saving the buyer the twenty minutes of working out why it happened.
The human sees the exceptions that genuinely need judgment. Everything else is already resolved.
The Three Layers
This is the model:
Layer one: Auto-resolve if possible. If the deviation fits within acceptable parameters and maintains coverage, accept it and move on. No human, no delay.
Layer two: Auto-mitigate if resolution fails. Try to fix the underlying problem — negotiate with the supplier, find substitutes, orchestrate warehouse transfers — before the human ever sees it.
Layer three: Escalate intelligently if both fail. Hand it to a human as a curated risk, ranked by criticality and impact, with the work already done. The human makes the call, not the triage.
What This Actually Means
The outcome isn't "we automated stuff." It's three things your finance and operations teams actually care about:
Higher service levels. When exceptions are caught and resolved with diligence around the clock, supply reliability rises across the board. For wholesalers, this shows up as higher fill rates — goods arrive when promised, customer orders get fulfilled, and the order-to-cash cycle stays intact. For large-scale manufacturers, it shows up as dependable material and component availability: fewer line stoppages, less expediting, smoother production flow, and ultimately a higher service level on the finished product that reaches the end customer. In both cases the mechanism is the same — deviations are surfaced and resolved before they propagate downstream, so the disruption never reaches the part of the business where it becomes expensive.
Same working capital. You're not holding extra safety stock to cover the gaps that should have been caught. You maintain coverage with the inventory you already have.
People freed up. Your procurement team stops firefighting confirmations and starts resolving supply chain problems. They move from triage to strategy.
And underneath it all: consistent, intelligent exception handling that scales. Most wholesalers either skip this work entirely or do it halfway. You get diligence around the clock that humans can't sustain.
Why This Matters Now
We didn't invent this model in a vacuum. We built it because we work with large wholesalers every day and we see the reality: the ones that actually care about confirmations get better outcomes. Higher service levels. Lower working capital. Smarter people doing smarter work.
The ones that don't? They compensate with inventory. They accept lower fill rates. They accept that this work just won't get done with the necessary quality.
We've already proven this works. Our current customers are living it.
The Real Shift
When we started, we chased a percentage. We wanted to say "we automated 100%." And we got there.
But the percentage was never the point. The point was: can we make decisions with enough intelligence that humans never inherit a mess? Can we resolve what's solvable, mitigate what's mitigatable, and escalate only what genuinely needs judgment?
That's the model that scales. That's the model that moves the needle.
The 100% automation? That's just the byproduct.
