How to manage 10x the invoices in half the time

The short version

Your AP process worked fine at 50 invoices a month, but at 200 it starts to break. Processing 1,000 invoices manually averages 141 hours of work, and roughly 70% of AP time still goes to manual data entry. A single 15-second action repeated across 100 invoices a week costs nearly 20 hours a year. The math compounds before headcount can catch up.

Pair Eyewear cut invoice processing time substantially after overhauling their AP workflow with Ramp Bill Pay. Staci Robinson, Pair's AP manager, shared how she did it and how she now handles office management, facilities, and real estate on top of AP.

Top 5 takeaways from this webinar

1. Traditional OCR is why your "automated" AP still feels manual

Your AP platform probably runs on optical character recognition that works like a blind copy-and-paste, pulling from fixed positions on the page. That's fine for a clean two-line invoice and useless for a 50-line item invoice with scattered descriptions and dollar amounts.

Matt Mayer demonstrated this with a 58-line item invoice that had no due date listed, only "Net 45" payment terms. Ramp's OCR reads natural language, so it calculates the due date itself. The gap between OCR that copies and OCR that comprehends is where 70% of your time goes.

2. Auto-coding and suggested coding are different levers, and you need both

Auto-coding hard-assigns a GL category to every invoice from a given vendor. Suggested coding learns from your past decisions and proposes a category, which gets sharper the more bills you process.

Staci described the impact on her team:

"Because we're a manufacturing organization, we have so many lines from purchase orders and invoices. So the time savings cuts it by 80%. I've been able to move into other areas due to the time savings that I have, other areas of work like office facilities and real estate."

3. Controls don't have to be all or nothing

Your AP process probably forces a tradeoff of requiring CFO approval on every change or skip controls entirely. With Ramp Bill Pay, you can toggle preapproval on or off for four specific change types: vendor, payment amount, payment details, and payment scheduling. Everything else moves.

In Matt's demo, an inbound vendor email asked to update bank details before a payment went out. He made the change inside the bill, the system flagged it for approval based on policy, and the rest of the bill kept progressing.

4. The cumulative cost of 15-second actions is where AP actually bleeds

Ramp's team watched users click into a bill, scroll to the GL field, and update coding on competing platforms, and the average was around 15 seconds. At 100 invoices a week, that single action costs nearly 20 hours a year, and AP has dozens of them: toggling between email and the AP tool, opening contracts in a separate system, repeating coding on identical vendor bills.

Most AP teams optimize each task on its own and miss how those small time costs stack up across the whole workflow. Bulk actions only earn their keep when the rest of the workflow is already clean.

5. Stakeholder interviews come before software selection

When Staci joined Pair Eyewear 2 years ago, bills were arriving by mail and email across 3 offices, and her first move wasn't a vendor evaluation. She ran 2 or 3 in-person meetings and 2 or 3 virtual calls with every stakeholder who touched AP and asked one question: what is slowing you down?

"When you ask the question of a person, like, what can we do to help you and make this better for you? That's almost a bit of buy-in right then."

By the time she presented a deck on the new process, the team had already told her what to fix.

Why do most AP teams overhaul the wrong layer first?

Most AP teams overhaul the wrong layer first because they jump straight to software shortlists instead of mapping where the process is actually breaking. Staci started with conversations. Before Pair Eyewear evaluated a single platform, she mapped every channel invoices were arriving through, every stakeholder who touched a bill, and every bottleneck those people described in their own words. That groundwork is why the Ramp implementation moved fast. The team already knew exactly what the platform had to do.

Software can't fix an approval chain that no one agrees on. It can fix data entry, GL coding, and payment execution, but only after your team agrees on what the workflow should look like. This is also why Pair Eyewear’s expansion into procurement is happening now, not at implementation.

"We can request the purchase order right through Slack or set up the approval policy for the purchase order, and it works out.”

Procurement requests can be submitted through Slack, and approval policies are configured similarly to bill approvals. Sequencing matters.

The technique behind all 3 fixes

The thread connecting every part of the Bill Pay demo was the keystroke count. Matt processed a 58-line item invoice end to end and used zero keystrokes for data entry. The technique has 5 parts.

1. Let the OCR read the whole invoice, not just the fields it recognizes. Ramp's OCR processes natural language, so "Net 45" populates the due date automatically. No mental math, no manual entry on payment terms.
2. Decide where coding should be hard versus suggested. For vendors you use constantly, hard-code the GL category once. For everything else, let suggested coding learn from your decisions over time.
3. Configure preapproval triggers, not blanket controls. Inside Bill Pay settings, toggle preapproval on for vendor changes, payment amount, payment details, and payment scheduling. Leave everything else routable.
4. Batch the execution. Once coding and controls are clean, you can code, approve, schedule, and pay in bulk. Matt processed 20 bills in roughly 10 clicks.
5. Sync back to the ERP. Your coded bills sync directly to NetSuite without manual re-entry.

A good starting point is to open one of your messiest recent invoices, the kind with no clear due date and a long line-item table, and time how long it takes your current platform to extract it cleanly. That number is your baseline.

Staci Robinson's 3 fixes for scaling AP

Fix 1: Replace OCR-that-copies with OCR-that-comprehends

The problem. 70% of AP team time goes to manual data entry because most AP software still relies on traditional OCR that pulls from fixed positions on the page. The moment an invoice deviates (a long line-item table, scattered dollar amounts, no explicit due date, non-standard vendor formatting), it falls apart. You fill the gaps manually, which erases the automation savings the platform promised.

How it works. Ramp's OCR reads the entire invoice, not just fixed fields on the page. Using the same 58-line item invoice from the demo, Matt showed how Ramp parsed the natural language, calculated the due date from "Net 45" payment terms, and populated every line item (descriptions, quantities, and dollar amounts included).

Two configurable layers handle coding: hard auto-coding rules tied to specific vendors, and suggested coding that learns from your past decisions and refines over time. The suggested coding loop is the key safeguard. The system gets smarter as you correct it, so even invoices from new vendors get cleaner suggestions month over month.

What this looks like in practice. Zero keystrokes on data entry for a 58-line item invoice. Staci reported an 80% time savings at Pair Eyewear from the combination of auto-coding and suggested coding.

Fix 2: Build conditional approvals into the bill itself

The problem. You're stuck between two bad options: require CFO approval on every vendor change and slow everything down, or skip controls entirely and accept the fraud and error risk. Most platforms force one or the other.

How it works. Inside Bill Pay settings, you can toggle four preapproval triggers: vendor, payment amount, payment details, and payment scheduling. Turn on only the ones that matter for your risk posture.

In the live scenario, a vendor emailed asking to update their bank account before payment was processed, and Matt added the new routing and account number directly inside the bill. When he saved the change, the system prompted: "Your changes will be sent for approval based on your organization's policy."

The policy-based hold is the safeguard here. The change is captured, the bill isn't blocked from progressing, and the payment is gated until approval lands.

What this looks like in practice. The fire-drill scenario (vendor asking for a bank change while the CFO is on vacation) resolves without skipping controls and without blocking everything else on the bill.

Fix 3: Treat bulk actions as the payoff, not the feature

The problem. Clicking into one bill, scrolling to the GL field, updating it, and saving takes about 15 seconds. Multiplied across 100 invoices a week, you lose nearly 20 hours a year on one repetitive motion. AP has dozens of these.

Most platforms offer bulk actions but don't fix the data quality problems that make bulk processing risky in the first place.

How it works. Once invoices are auto-coded (or coded in bulk inside the dashboard view), the rest of the workflow runs in batches. Matt selected 20 bills, bulk-edited accounting fields for the uncoded ones, bulk-created the bills, bulk-approved them, bulk-updated the payment date to the last Friday in September, and reviewed the batched payment screen before executing.

The batch review screen is your last check. Before money moves, you see the total, the source bank account, the current balance, and every individual bill, and you can toggle any one off.

What this looks like in practice. 20 bills processed end to end in roughly 10 clicks.

"Before, I used to spend an hour or two just creating a batch for payments, and now it's like a couple of ticks and payments are done."

What should AP teams do first?

1. Run the stakeholder conversation Staci ran at Pair Eyewear. Get every person who touches AP (finance, ops, office managers, budget owners) on 2 or 3 calls and ask one question: what is slowing you down? Capture the answers verbatim. That document becomes your requirements list before you talk to any vendor.
2. Time your current OCR on your worst invoice. Pull one recent invoice with a long line-item table or no explicit due date. Time how long it takes your current platform to extract it cleanly and how much manual cleanup you do after. That number is your benchmark.
3. Audit your approval triggers. List every change inside a bill that currently requires approval. Sort into "actually needs review" and "blanket policy that no one questions." Cut the second list. The goal is conditional controls, not universal ones.

Final thoughts

"Before, I used to spend an hour or two just creating a batch for payments, and now it's like a couple of ticks and payments are done."

That's what 80% time savings actually looks like in practice. Not a slide but a workflow that used to consume the better part of an afternoon, collapsed into a handful of clicks. The reason Staci now runs facilities and real estate alongside AP isn't because she got faster. The process did.

Start with one fix.

See how Ramp fits in

You'll get the most out of accounts payable automation if you redesign the process before you swap the software. With Ramp Bill Pay, you can replace manual data entry, all-or-nothing approval chains, and one-bill-at-a-time execution on the same platform you already use for card spend and ERP sync.

Common questions AP managers are asking

How do I know if my current OCR is the bottleneck?

Open the 3 messiest invoices you received this month, the ones with long line-item tables, missing due dates, or non-standard formats. Time how long it takes your platform to extract them cleanly and how many fields your team had to fix by hand.

If more than a third of the data needed manual cleanup, your OCR is the bottleneck, not your team's speed. The fix is comprehension-grade extraction, not adding another reviewer.

When should auto-coding stop and suggested coding take over?

Hard-code only the vendors where the GL category never changes, such as recurring software subscriptions, your office landlord, or a consulting partner on a flat retainer. Everything else should run through suggested coding so the system learns your team's decisions over time. Trying to hard-code every vendor breaks the moment a line item deviates.

What approval triggers matter for fraud prevention?

The four preapproval toggles available in Bill Pay settings cover vendor changes, payment amount, payment scheduling, and payment details. Those four cover the changes that have actual financial consequences.

Routing every coding tweak or note edit through CFO approval slows the workflow without reducing risk. The point of conditional controls is to gate the changes that move money, not the changes that move metadata.

How long does an AP rollout like this actually take?

Staci did the groundwork (stakeholder interviews, channel mapping, the deck to her team) before evaluating any vendor. That's why the Ramp implementation moved quickly once she chose it.

The honest answer is that the platform install is the short part. The longer part is getting every stakeholder to agree on what the workflow should look like. Teams that skip the conversations spend the time later in implementation.

Where does AP automation help with 1099 filing?

Every vendor payment you run through Bill Pay software flows straight into your 1099 workflow. Ramp fills in vendor details, maps categories to 1099 boxes, and calculates totals for you. When you're ready, you file directly with the IRS and eligible states—without leaving Ramp.

You can read more about the 1099 filing feature.