Requisition volume spikes at budget close, site openings and year end. See why extra reviewers rarely help, what breaks first, and how AI agents absorb a peak without losing a check.
Why does requisition volume spike at all?
It’s the second week of December. Three departments have just realized they’ll lose next year’s budget if they don’t spend it, a new site is being fitted out, and the plant has scheduled its shutdown maintenance for January. Everything lands in your queue at once.
Volume isn’t flat, because the business isn’t flat. It follows the calendar. Budget closes, sites open, factories shut down, and year end arrives whether anyone’s ready or not.
So do the requests get harder during a peak? No. There are just three times as many of them, arriving in the same fortnight, into a team that’s sized for an average month.
Why adding people doesn’t fix it
The first instinct is temporary help, and it almost never works. Why not?
- The ramp is longer than the peak.
Reviewing a requisition properly means knowing your category rules, your contracted suppliers, and which exceptions are real. That takes months to learn. A contractor who starts when your peak starts is still learning when it's over.
- Quality drops exactly when you need it most.
Give someone a throughput target and reviews get shallower. That missed contract link or wrong commodity code during December? It's the audit finding you're explaining in June.
- You pay for it all year.
Headcount you need for three weeks gets paid for across fifty-two. Which is why you probably didn't get approval for it last time you asked.
If the queue is slow in a quiet month too, the peak is not your real problem. What to automate first in a manual intake process deals with the baseline.
What actually breaks when volume spikes
Requests don’t stop moving, which is what makes this so hard to spot. What stops is the checking, and you find that out months later.
Reviews get skipped or rubber-stamped
When your queue is three times its normal length, you can clear it or you can check it. Be honest about which one your team picks.
Requesters stop waiting for you
Someone needs it this week, sees a two-week queue, and finds another way to buy it. Your peak season is where your maverick spend comes from.
Negotiation goes first
Checking is mandatory. Negotiating isn't. So the part that was going to save you money is the first thing to go.
Late rejections pile up behind you
Problems nobody caught at the start bounce back after approval, so you handle the same request twice, and it lands in a queue that's already full.
A peak doesn't create new work. It shows you how much of your existing work was waiting on someone being free.
How AI agents absorb a peak
Here’s what changes. AI agents don’t have a throughput ceiling that moves with headcount, so they handle ten requests a day or a thousand at the same depth. Checking a request is the same work every single time.
With Pactum, three things happen to every request, no matter how many turn up at once.
The request arrives complete
Your colleague describes what they need in plain language, in Microsoft Teams, Coupa, or SAP Ariba, and Pactum's Intake Agent turns it into a correctly coded requisition. Nobody's guessing at a commodity code at 6pm on the last day of the quarter.
Every request gets checked, not a sample
The Alignment Agent checks each requisition against your policies, your category rules, your approved suppliers, and your contracts before money is committed. It fixes what it can and sorts out the rest with your requester in the same conversation.
The buys you should negotiate still get negotiated
Anything that qualifies goes to Tactical Sourcing or Single-Source Negotiation, all in parallel. Your savings don't disappear just because it's a busy month.
At SUEZ, automating the review step cut requisition cycle time by 90% and took 80% of the manual review work away.
What stays with your team during a peak
Absorbing the volume isn’t the same as handing over control. You keep every decision that needs judgment.
- Your thresholds. You decide what the agents handle alone and what always comes to a person, by value, category, or supplier.
- Your exceptions. Anything outside your rulebook goes to a buyer, with the reason attached.
- The strategic buys. The agents take the volume, so you’ve got room for the handful of deals in a peak that genuinely need you.
- The audit trail. Every decision an agent makes gets written back to your P2P system, so a busy month is still a defensible one.
What you need in place before your next peak
Do you need a perfect policy framework first? No. You need three things.
- Your rules written down somewhere. They don’t have to be tidy. The Alignment Agent turns what you’ve got into one rulebook, and shows you where it contradicts itself, which is usually the more interesting part.
- Supplier and contract data that’s roughly current. The cleaner it is, the more your agents resolve without asking anyone.
- A decision on thresholds. What runs on its own, and what a buyer always sees.
Most teams start in observation mode. The agents check every request against your rulebook while your people still decide, and you compare the two. Once the agents match what your reviewers would have done, you hand over more.
One more thing worth settling before the peak: whether people will actually use the new route when they are under pressure. Why intake adoption stalls after a system change covers that, and the guide to intake agents in procurement sets out what each agent handles.