Why Roy

Why a General AI App Is Not Enough

Today's leading AI apps, on paid or business plans, offer memory, connectors to your systems, browser control and scheduled cloud agents, with setup and approvals still required.

General models are powerful building blocks. Roy is the chargeback system already built from them, so you don't have to build it.

What it takes to dispute chargebacks with AI

Figure 1. The same six requirements, three ways.

	What a general AI app gives you	What you'd have to build and run	What Roy does
ERP data	Connectors that give access to your systems	Knowing which field answers each type of claim	Pulls the facts each claim needs from your ERP
Documents	Reads the files you hand it	Reliable reading of BOL packets, scanned PODs, remittances and carrier emails, matched to the right case	Reads them and matches each to the right case
Retailer portals	Browser control that can use a website	Handling for each portal, kept current as it changes, without locking out shared logins	Works supported retailer portals, designed to help avoid shared-login lockouts
Judgment	General knowledge and memory, but not the supply-chain understanding of what each document and date proves	Supply-chain knowledge of what each document and date proves, and each retailer's rules	Understands what each document and date proves and applies your retailer rules
Evaluation	Answers that can read well and still be wrong	Expert review on real cases, re-checked when a model, rule or portal changes	Reviewed by our experts on real cases, re-checked as things change
Upkeep	Scheduled cloud agents, with setup and approvals	Engineers to run it, and someone accountable when a run fails	That work is ours to run

The model is the easy part. With Roy, your team focuses on being a great wholesaler, not on running an AI-native software company.

How Roy works, and keeps learning

Roy files disputes by retailer portal or email; ROIAI One's internal analysts monitor its work and review flagged cases first.

Figure 2. Roy files when confident; an analyst checks first when it's not sure.

Roy shares the learnings across all wholesalers and increases the win rate.

Figure 3. ROI rises as Roy works more cases.

A dispute can read well and still be wrong

Two real cases. The customer is not named.

Carton shortage, Burlington, SLC load

Figure 4. The SLC load case, read three ways.

A generic chatbot model's plausible reading: Misses the SLC mark on the BOL, treats it as an ordinary carton shortage, and the dispute packet says the retailer is at fault.

What it actually means: The BOL is marked SLC (Shipper's Load and Count): the driver's signature is not a carton count, and the carrier is generally not responsible for the quantity.

With the seal intact, the shortage points to the warehouse that loaded the truck.

What Roy does: It's built to catch the SLC mark, hold back a dispute the facts don't support, and route the claim to that warehouse.

Burlington cancel date

Figure 5. The cancel date case, read three ways.

A generic chatbot model's plausible reading: Picked up after the cancel date, so the vendor is at fault.

What it actually means: The last date the retailer will accept the order. Ship-by or arrive-by depends on the retailer guide; lateness on routed freight is judged on the vendor's routing request and readiness.

What Roy does: It's built to read the date as the retailer guide defines it.

What it means for your business

Figure 6. Six outcomes of a system built this way.

Fewer wrong readings: Expertise built into Roy is designed to keep disputes off the wrong reading.
Less digging: Roy is built to pull the facts each case needs, so analysts spend less time gathering them.
Easier to check: Decisions come with reasons you can review.
One owner for submission: ROIAI One files the claim, so filing isn't left to your operations or accounting teams.
Preventing recurring chargebacks: Built to show why it happened and what to change; insights vetted by our in-house experts.
Forwarding to the party at fault: Roy is built to route the claim to the party the facts point to, such as your warehouse or 3PL.

Figure 7. The claim goes to the party the facts point to.

Who stands behind it

An operations advisor who was EVP Operations at an apparel wholesaler.
A deduction analyst who works with us, from order-to-cash at a global consumer-goods company.
Our founder, Pouya Samangouei (ex-Google DeepMind), with our ML engineers and scientists.

See what Roy does with your chargebacks

Get an assessment

Questions first? Read the FAQ.

Case studies on roiaione.com

October 2026. Statements about general AI apps reflect publicly documented features at the date of this document, and those products change over time.

Capability comparisons are ROIAI One's view, and generic readings show how a general model can misread these cases.

Descriptions of Roy reflect how it is designed to work and are not a commitment of results or a contract term.