The Five-Figure AI Mistake: How to Avoid Automating More Workflow Pain
The AI rush is creating expensive shortcuts around broken workflows—and small aviation companies are paying for them.
Everybody’s in a hurry to show a return on AI. Ownership wants a plan. Boards want a story. Software companies want new revenue. Providers need customers to buy the new widget, agent, assistant, or automation before anybody asks whether it belongs in the workflow.
The companies most exposed look like yours and mine: five, ten, or twenty users, trying to scale and small enough that a bad technology purchase hurts. A 500 user company isn’t changing its operation overnight because somebody gave a chatbot a clever name. A growing ten-user company may spend five or six figures after being promised more RFQs, better pricing, or dramatic productivity, then quietly not renew when the business doesn’t change. The provider keeps the revenue. The customer keeps the lesson.
That isn’t an AI strategy. It’s a gold rush, and the easiest money often comes from selling tools to people afraid of being left behind.
The customer comments below come from real conversations. They're anonymized and lightly cleaned only for readability.
AI Pressure Is Creating the Wrong Buying Question
The market keeps asking, “What are we doing with AI?” That’s the wrong first question. The right question is, “What decision can’t we make well today, and why?”
If the problem’s bad data, AI will read bad data faster. If the problem’s a broken approval path, an agent may move the same confusion more efficiently. If customer, inventory, pricing, certification, and vendor information live in different systems, a conversational interface doesn’t reconnect the transaction. It gives the fragmentation a friendlier front door.
One customer said it plainly:
“There’s a lot of manual work that is still done, despite all the tools that are available through the ERP and through other systems that we have.”
Parts supplier, anonymized
That company didn’t lack technology. It lacked return from the technology it already had. Another layer wouldn’t repair the operating chain. It would add another invoice, implementation, and place for context to get lost.
The board may hear that the company “deployed AI.” Ownership may hear about new AI revenue. Neither tells the operator whether a quote went out faster, margin improved, a late part was caught, or an auditor can reconstruct the transaction.
Automating a Broken Workflow Does Not Fix It
AI’s made it easier to build. It hasn’t made it easier to understand the exceptions, controls, and commercial judgment inside an aviation transaction.
Parsing an email is useful. So is voice-to-text. Turning a spoken request into an old database query may save a few clicks. But none of those is a decision engine. They move information. They don’t determine whether it’s complete, the inventory is available, the documentation matches the customer’s requirement, the vendor is approved, the price protects margin, or the next action should be allowed.
That’s why putting an agent on a five-year-old workflow can be such an expensive illusion. The same handoffs and bad assumptions remain. The agent simply runs between them faster.
One repair operator explained why its current-system warnings had become meaningless:
“We get those warnings all the time in this system. We just blow past them… If we do it the way they want us to do it, it doesn’t let you mark it up properly.”
Repair-station operator, anonymized
That isn’t a missing alert. It’s a workflow that conflicts with the business, so experienced employees bypass the control. Repeating the warning with AI won’t solve it. The system must understand the outside-service transaction, preserve cost and trace, support the markup, and make the right path easier than the workaround.
If Your ERP Needs Five More Services, Ask What You Bought
An ERP is supposed to run the operation. Yet companies buy an ERP, an RFQ service, a marketplace assistant, an email parser, reporting software, an integration layer, and a spreadsheet to reconcile what still doesn’t agree. Every integration creates another data boundary and another place to copy, delay, or lose the truth.
One aviation operator described the stack in plain language:
“We would be operating from email—that is one platform. Calls, inbound and outbound, that’s another platform. And the software that keeps our inventory… what we would love is if all of that would be able to sync.”
Aviation parts supplier, anonymized
Another owner put it even more plainly:
“I have a Google Sheet nightmare going on, because I have 27 tabs.”
Owner of a growing aviation supplier, anonymized
This is the part software pricing hides. The subscription is visible. The employee copying a vendor response isn’t. Neither is the salesperson searching an inbox, the buyer reconciling a spreadsheet, or the manager rebuilding the story before an audit. You pay for the software, then pay your people to compensate for what it doesn’t do.
When a customer asks a simple question, your team searches seven places for the answer. That isn’t an integrated business. It’s a group project between vendors.
The Return Disappears in the Work Nobody Measures
The strongest proof comes from watching what customers do.
One operator said:
“We’re spending 90% of our time finding the part, entering all the information, and going through that sequence all the way to shipping - instead of doing sales.”
Aviation distribution operator, anonymized
A repair shop described the same loss on the production side:
“I can’t have people spending two or three hours trying to create all this, with all the possibility for making mistakes, just to get an 8130 on a piece part we’re using in another unit.”
Repair-station operator, anonymized customer transcript
Another company described the margin consequence of a vendor quoting one amount and invoicing another after the unit had already been received and sold:
"You’re quoted five, and you’re billed six… the unit is in, you’ve sold it, it’s gone, and now you’re dealing with an adjustment.”
Aviation supply and repair operator, anonymized c
And a sourcing team described the cost of fragmented outreach in two sentences:
“We missed that one part… and that’s the important part.”
“We do a blast out to everybody, and it raises the market price of the part.”
Part 135 and Part 145 sourcing team, anonymized
That’s the real ROI calculation: selling time lost to administration, labor spent rebuilding documentation, margin repaired after the sale, critical lines missed, and purchasing behavior that moves the market against you. If the new AI tool doesn’t materially change one of those outcomes, it isn’t an investment. It’s entertainment with an implementation fee.
A Decision Engine Must Do More Than Produce an Answer
A useful decision engine doesn’t merely tell you a part exists or repeat what a report already says. It brings together the evidence required to decide what should happen next.
For an RFQ, that means more than extracting a part number. It means understanding available and reserved inventory, condition, trace, certificates, customer history, prior pricing, vendor responses, alternates, margin requirements, lead time, and the probability the transaction can be completed. It should show why one option is stronger, know who can approve the action, and keep the decision attached to the transaction so the next person doesn’t rebuild the context.
One manufacturer asked what every operator should answer:
“I thought this was going to take an hour. How long did it take? And how much did it really cost me?”
Aviation manufacturing operator, anonymized
That’s a decision question. The value isn’t a dashboard showing the job closed. It’s knowing whether the estimate was wrong, where the cost changed, what should be priced differently next time, and whether the same pattern is happening elsewhere.
Before spending a penny, ask the vendor: What decision becomes better, faster, or safer? What evidence supports the recommendation? Does the action return to the governed transaction? If a human still has to visit five systems, compare the answers, and re-enter the result, you didn’t buy intelligence. You bought a more polished handoff.
Why We Chose Embedded Intelligence for ERP.Aero
Customers have been just as clear about the desired state:
“It’s the little things. It’s all interconnected. You don’t have to fill out the same information twice, or multiple times.”
Part 135 and Part 145 operator, anonymized customer transcript
That’s the standard I had in mind. We didn’t want us to add a generic assistant beside the ERP and charge customers to repair the separation. We started with the operating chain.
​iRFQ handles supported inbound RFQ and vendor-response formats by converting the information into structured ERP records where the quoting work already happens. Known transaction steps can move through governed rules without waiting for someone to remember every status change. RFQs, vendor quotes, inventory, history, documents, and commercial context stay connected to the same workflow.
​ELIA. the Embedded Layer of Intelligence in Aviation, is built inside that transactional foundation. It uses the company’s own ERP data to surface connected context, relevant history, evidence, and the next controlled step. I don’t want AI to run the company. I want the person making the decision to see the pricing, margin, sourcing, customer, vendor, inventory, documentation, and operational consequences without rebuilding the story by hand.
That distinction is everything. Automation should handle known movement. Intelligence should improve judgment. The ERP should preserve permissions, workflow, and audit history. When those pieces live together, the system can help the business act. When they’re scattered across add-ons, the user remains the integration layer.
I’m not arguing every outside service is worthless or every company needs more AI. Some specialized tools solve specialized problems. Sometimes the system you already have is good enough, and the smartest decision is to spend nothing. But if somebody’s asking for five or six figures because a new agent can speak, summarize, search, or move data between broken workflows, ask the only question that matters:
Will this help my people make better, smarter, faster decisions inside the operation - or am I paying another company to make our fragmentation look modern?
If the answer is not obvious, keep your money.
About Ralph Merhi
I'm Ralph Merhi, CEO of ERP.Aero. I didn't enter aviation as a software executive. I came through the industry, from sales and customer success to subject-matter expertise and leadership, and eventually found myself in software because I was tired of watching good operators compensate for systems that never fit the work.
I joined ERP.Aero because I saw a team willing to solve root causes instead of decorating symptoms. My goal is to help operators see more clearly, make better decisions, and stop accepting friction as the price of doing business.
About ERP.Aero
​ERP.Aero is an aviation-native ERP and operational platform for suppliers, distributors, MRO and repair organizations, and related aerospace operations. It connects quoting, sourcing, inventory, certification and compliance, repairs, fulfillment, finance, reporting, integrations, and embedded intelligence in one operational system.
iRFQ converts supported inbound RFQs and vendor responses into structured ERP records. ELIA adds guided intelligence inside the same transactional foundation, helping users see relevant history, evidence, and the next controlled step without rebuilding the transaction across bolted on answers. One screen. Built for aviation.
​