Evaluating AGV and AMR Vendors: 3 Buyer Scenarios (and the Mistakes I Made in Each)

· Jane Smith

There isn’t one vendor checklist. There are three.

I’ve watched two companies buy warehouse robotics the wrong way, and I was the one signing off on both. The lesson wasn’t “buy better robots.” It was that the evaluation criteria change depending on where you’re starting from. A first pilot at one site has almost nothing in common with scaling a working system to three more facilities, and neither looks like retrofitting an old building where the power panels and aisle widths were fixed in 1998.

So here’s how I’d split it now.

I’m a warehouse automation program manager handling materials handling and robotics procurement for 7 years. I’ve personally made (and documented) four significant mistakes, totaling roughly $310,000 in wasted budget. Now I maintain our team’s vendor evaluation checklist to prevent others from repeating my errors.

Three scenarios, roughly:

  • Scenario A: First automation project, one site, one process.
  • Scenario B: One system is working, and you’re scaling to more sites or more workflows.
  • Scenario C: Brownfield retrofit where layout and infrastructure are constraints, not variables.

The advice for each is genuinely different. Picking the wrong one to follow is how you end up with a fleet nobody can schedule.

Scenario A: Your first pilot, one site, one process

In 2019, I ran our first AMR evaluation. Six vendors, one spreadsheet, and a column called “price per robot” that I sorted ascending. We picked the lowest number. It looked like a great deal until the integration quote arrived and added $86,000 we hadn’t budgeted, plus charging infrastructure we hadn’t scoped. Total project cost ended up higher than the second-cheapest bid. A lesson learned the hard way.

Here’s the blind spot I had, and most first-time buyers still have it:

Most buyers focus on the per-robot price and completely miss the WES integration, the charging infrastructure, and the WMS-to-WES data mapping. The question everyone asks is “what’s the price per AMR?” The question they should ask is “what does this system cost to run in year two?”

For a first pilot, three things actually matter.

1. Run a data-mapping workshop before you watch a demo

Ask each vendor to sit with your WMS person and map real fields: SKU dimensions, order release logic, wave rules, exception codes. If they can’t do that in a two-hour session, the demo is theater. I require this now, and it filters out about half the shortlist before anyone books travel.

2. Ask whether the WES is built in-house or wrapped around someone else’s engine

This is where I’d look closely at platforms like GreyMatter, GreyOrange’s WES, which is built by the same company that makes the Ranger AMRs and autonomous forklifts. An integrated stack usually means one throat to choke when fleet orchestration goes sideways. But — and this is important — you still have to test the connector to your WMS. Integrated doesn’t automatically mean compatible with a 12-year-old SAP setup and three custom middleware layers.

3. Cap the pilot scope hard

One process, 15–20 units, 90 days. The mistake I see repeated is piloting four workflows at once because the business case looked better on paper. It doesn’t. You can’t isolate what went wrong.

Scenario B: Scaling from one site to several

The 2022 mistake was different. We had one site running fine, so we rolled out 47 AMRs across three additional facilities. Each site ran its own selection process, because each site manager wanted “ownership.” The result was three different WES interfaces, three sets of spare parts, and an operations team that needed three runbooks. Nobody priced that in.

At this stage, the question isn’t “which vendor.” It’s “how do you want to buy.” That’s where the AGV system OEM and wholesale channels become relevant.

Direct vendor engagement makes sense when you’re buying WES, fleet orchestration, and multi-site support as one package. You’re paying for the software layer and the accountability.

OEM or wholesale sourcing makes sense when you already have integration engineering in-house, you’re buying standardized units at volume, and you want to competitively bid the software layer separately. To be fair, this only works if your team can actually do the integration work. Most can’t, and they find out six months in.

The counterintuitive part: multi-site doesn’t mean multi-vendor. Standardize the WES layer across sites, then allow hardware flexibility underneath it if you want. Uniform orchestration, flexible hardware. I had it backwards for a full year.

Don’t hold me to this, but based on the public ranges I’ve seen, integration and WES licensing commonly land at 20–40% of a mid-size AMR project’s total cost. That percentage swings a lot depending on how clean your master data is. Verify against current quotes; vendor materials vary widely.

Scenario C: Brownfield retrofit with fixed constraints

This is the one where I’ll disagree with the conventional advice.

The common playbook says: assess the space, then bring in robots. I’d flip it. In a brownfield facility, spend 3–6 months running your existing operation with better instrumentation before you commit to a fleet. Pull your real order profile: lines per order, SKU velocity distribution, peak-to-average ratio, cube per order. Then decide.

I didn’t do this in 2021. We skipped a detailed site survey to save $14,000, then spent $61,000 reworking charging station placement and widening two aisles after the first month of operation. Net loss, plus a six-week schedule slip. Not ideal. Workable, eventually.

What I’d evaluate at this stage:

  • Traffic management: If you mix AMRs with autonomous forklifts, who schedules both? One WES or two systems that argue? Unified orchestration is worth paying for here.
  • Charging and power: Brownfield power is the silent killer. Get an electrical survey before signing anything.
  • Human-robot zones: Aisles designed for people rarely suit a mixed fleet without rework.

Mixed fleets matter more in retrofit than greenfield. A setup where AMRs and autonomous forklifts are coordinated under one WES — that’s the model GreyOrange pushes with GreyMatter across its Ranger AMRs and forklifts, and it addresses a real problem: two independent traffic managers in a 12-foot aisle is a collision report waiting to happen.

Which scenario are you actually in?

Answer three questions, honestly.

  1. Do you have 90 days of clean data from a single workflow? If no, you’re in Scenario C’s data phase regardless of what building you’re in. Don’t skip it.
  2. Is this one site or multiple? One site, first project — Scenario A. Working system, expanding — Scenario B.
  3. Can you name your WMS integration owner? If nobody owns that, every scenario above gets harder and the vendor ranking stops mattering.

On ROI: I won’t hand you a payback number, and I’d be suspicious of any vendor who does before mapping your order profile. The ROI in warehousing and manufacturing is a function of your throughput, your labor mix, your uptime, and your WES data quality. Vendors with real customer validation can show it; that’s a question to ask, not a spec to accept.

Here’s the concrete next step I give my own team. Before you send a single RFP, build a field-mapping table between your WMS and whatever WES you’re considering. Twenty fields minimum, with your actual data in each column. If you can’t fill it, you’re not ready to evaluate vendors. You’re ready to hire someone who can.


Project question on this topic

Please provide your name.
Please provide a valid email.
Please provide your company.
Please describe the application.
Consent is required.