The first time I walked into our parts yard at 5 a.m. and watched a senior tech dig through three shelves for a single stuffing box packing set, I knew the spreadsheet wasn't cutting it. He found it. Forty minutes later. The pumper on location had already called twice.
That morning is basically why we ended up rebuilding the whole thing around AI. Not because somebody read a Gartner report. Because a guy named Travis was tired.
Here's what actually happened when we switched, what worked, what didn't, and a few things I'd tell anyone in oil and gas before they start.
The Yard Was a Mess and Nobody Wanted to Say It
We had a CMMS. We had bin locations. We had a parts clerk who knew where everything was if you asked nicely.
The problem was none of that survived a busy week.
Here's what was actually going on:
- Two of the same VFD sitting in a back row because nobody knew we had one
- Beam pump bearings stocked for a well type we retired in 2022
- Glycol pump seals out of stock for six weeks while three units waited
- Inventory counts that matched the system on January 1 and nothing after
- Techs pulling parts and writing it down "later"
Later never came. Later is a myth in field operations.
A real parts yard before any structured tracking. Closer to the norm than anyone wants to admit.
What AI Actually Does Here (It's Not Magic)
When our ops manager said "AI for parts," half the yard rolled their eyes. Fair. There's a lot of noise around this.
What it really does is boring in the best way. It looks at your work order history, equipment runtime, failure patterns, and consumption rates, then hands back a forecast that isn't based on a gut feeling from 2019.
A few things it started catching for us almost immediately:
- Seasonal demand on heater treater elements we always under-ordered in November
- A specific brand of pump packing that failed 40% faster on one pad
- Reorder points that were set once and never touched since the yard opened
The shift from "we usually order this much" to "based on the last 14 months and current runtime, you'll need 6 in the next 21 days" is the whole game.
If you want a deeper walkthrough of how that demand modeling actually works under the hood, Equipt put together a solid breakdown on how to digitize spare part management that I sent to our purchasing lead before we kicked off.
Connecting Parts to the Equipment They Actually Live On
This is the part I underestimated.
A bearing isn't just a bearing. It's a bearing on Unit 14, which is a Lufkin 320, which had a gearbox rebuild in March, and has been running hotter than its peers since May.
When parts records sit in one system and equipment history sits in another, you lose all of that. The tech sees a part number. He doesn't see that the last three times this bin was pulled, the unit was back down within 90 days.
Once we tied parts usage to asset history, two things happened:
- We stopped replacing symptoms. The bearing wasn't the problem. The alignment was.
- We started seeing which suppliers' parts actually lasted on our equipment, not in a catalog.
That's where a connected oil and gas asset management setup earns its keep. One asset, one history, one parts trail. No more cross-referencing three tabs to figure out why a pump keeps eating seals.
The Overstocking Problem Nobody Talks About
Everyone worries about stockouts. Nobody talks about the $180,000 in slow-moving inventory aging on the back rack.
We had:
- Specialty valves for a workover rig we sold in 2023
- An entire shelf of fittings for a metric-spec pump line we phased out
- Filters in three sizes when techs only ever pulled one
AI sorted this in about a week of usage analysis. It flagged anything that hadn't moved in 18 months and cross-checked it against current active equipment. The list was uncomfortable. We moved or wrote off about a third of it.
The cash that freed up went into actually stocking the fast-movers properly.
Same yard, four months in. The shelves aren't fancier. They just match what the field actually pulls.
Where Maintenance Strategy Fits In
You can't talk about parts without talking about how you do maintenance. If you're still running pure reactive, AI forecasting won't save you, because your demand signal is chaos.
We had to clean up our maintenance approach first. If you haven't sat down and worked out where you actually fall on the preventive vs predictive vs proactive maintenance spectrum, the parts side will keep fighting itself.
Quick version of what we learned:
- Preventive gives the forecast something stable to chase
- Predictive tells you the part is about to be needed before the failure happens
- Proactive changes which parts you stock in the first place
The yard inventory shifted as our maintenance posture shifted. That's normal. That's the point.
What Surprised Me Most
A few things I didn't expect going in:
- Techs adopted the mobile scanning faster than the office staff adopted the dashboards
- The biggest savings weren't on big-ticket parts, they were on the small consumables nobody tracked
- Suppliers negotiated better terms once we could show them clean usage data
- Our parts clerk became more valuable, not less, because she was finally working on planning instead of hunting
The fear that "AI will replace the parts guy" was the opposite of what happened. The parts guy got promoted.
A Few Things I'd Do Differently
If I were starting again tomorrow:
- Spend two full weeks cleaning master data before turning anything on. Bad data in, bad forecasts out.
- Get the field techs in the room for the bin layout, not just the warehouse lead.
- Don't pick a tool that only talks to itself. The whole point is connection.
That last one is why we ended up settling on a single platform instead of bolting three tools together. When parts, work orders, asset history, and maintenance plans live in the same place, the AI has something real to work with. Equipt.ai was where we landed because their oil and gas software treats the yard as part of operations, not as a separate inventory problem to solve later.
Final Thought
The first morning after we went live, Travis walked in, scanned a bin, grabbed the right packing set in under a minute, and left without saying anything.
That's the whole win. Nobody talked about AI. Nobody talked about dashboards. A guy got a part and went back to work.
If you're in an oilfield yard right now wondering whether this is worth the lift, that's the bar. Not the demo. Not the slide deck. Whether Travis gets out the door faster on a Tuesday morning.
Top comments (0)