T. Brian Jones is co-founder and CTO of Inventory Hero. He leads the engineering behind its Amazon data pipeline, demand forecasting, and the AI platform that lets sellers talk to their live inventory, sales, and supplier data in plain language.
Disclosure: Inventory Hero sells a paid inventory tool, so this comparison has a house interest. Amelia's honest peer set is other assistants, ours is other inventory tools, and most operators run both.
Frequently Asked Questions
Is Amazon Amelia still called Amelia?
Amelia was the codename Amazon used when it announced the assistant in beta at Accelerate in September 2024. Amazon's own current materials call the product Seller Assistant, and its Seller Central help page uses that name. Sellers and the trade press still say Amelia, so the two names refer to the same assistant inside Seller Central.
Is Amazon's Seller Assistant free?
Yes. Amazon says Seller Assistant is available at no additional cost, and its September 2025 announcement of the agentic upgrade described it as available to all sellers in the US store, with other countries rolling out afterward. There is no premium tier documented. It appears inside Seller Central rather than as a separate subscription.
Can Amelia tell me how much inventory to reorder?
It can analyze demand patterns and prepare shipment recommendations for your FBA inventory, and Amazon documents it as flagging slow movers and stock risks. Two caveats on the buying side. Seller Central does store a lead time and cost figures, but they are numbers you typed in once, not an observed distribution of how late your last ten POs actually ran. And Amazon does not document any source for your open purchase orders, your factory MOQ, or your cash position. So it can recommend how much to send to Amazon from stock you already own, while the question of how much to buy from your factory rests on inputs it either never sees or only sees as a single static number.
Does Amelia see my Shopify or wholesale sales?
No. Amazon does not document any off-Amazon data source for Seller Assistant. It is scoped to what your Amazon selling account holds, which means your Shopify, TikTok Shop, retail, and wholesale demand are invisible to it. If a meaningful share of your units move off Amazon, any days-of-cover number it gives you is measured against the wrong denominator.
Should I still pay for an inventory tool if Seller Assistant is free?
Often not. If Amazon is your only real channel and your catalog is small enough to review by hand, Seller Assistant plus a spreadsheet covers the job, and that stays true even if you import from Asia on a 90-day lead time, as long as that lead time is predictable enough that a fixed buffer holds. Paying makes sense when several of the missing inputs stack up at once: genuine off-Amazon demand, more SKUs than you can eyeball weekly, lead times that swing wide enough to need a computed safety stock, and a cash budget rationed across competing POs. Price the decision honestly. Inventory Hero starts at $79 a month, so the tool needs to be preventing something on the order of one modest stockout or one oversized PO a year before it is worth it.
Amazon's Amelia seller assistant is a free, generative AI assistant built into Seller Central that answers plain-language questions about your Amazon business and, with your permission, takes some actions on your behalf. Amelia was the 2024 codename; Amazon now ships and documents it as Seller Assistant.12 It is genuinely good at reporting what has happened inside your Amazon account. Where it gets thinner is anything that depends on data living outside that account, which is where the buying decision sits. Below is an honest read on which jobs to hire it for.
Up front, so you can weigh everything after it: we sell a paid inventory tool, which gives this piece a house interest. Two things follow from that and we would rather say them than have you infer them. Our tool cannot do most of what Amelia does, since it cannot act inside Seller Central, resolve a policy violation, or touch Account Health or compliance, and it needs setup and a data connection that Amelia does not. And the honest peer set for a paid inventory tool is other inventory tools, not Amazon's assistant. Most operators we talk to run both, as complements rather than substitutes.
Per Amazon's Seller Central help page, Seller Assistant does three things: it answers selling questions in natural language, it returns business insights across Sales Performance, Inventory Management, Listings and Catalog, Shipments and Returns, and Orders and Fees, and it takes proactive actions such as resolving some policy violations, optimizing FBA surcharges, and creating promotions.2
Amazon also documents it monitoring Account Health and alerting you to issues, automating compliance management for existing and new products, tracking inventory levels and alerting you to cost-and-growth actions, generating advertising creative from conversational prompts, and analyzing customer behavior to surface opportunities.2
On inventory specifically, Amazon's Accelerate 2025 material described it flagging slow-moving products before they incur long-term storage fees and recommending which items to leave alone versus mark down or remove, plus analyzing demand patterns and preparing shipment recommendations.4
Give it real credit for the structural advantages. It is free, it is already in the tool you log into every morning, it required zero setup or data connection, and it reads first-party Amazon account data that no third party sees as directly. Amazon reported more than 230,000 monthly users in 2025 with sellers accepting its recommended actions more than 90% of the time, which is a lot of operators finding it useful.5
"Why did my Account Health score move, and what do I do?"
Seller Assistant
Amazon owns the policy rules and the case flow. An outside tool is guessing at both.
"Is this listing compliant, and what documents do I need?"
Seller Assistant
Compliance requirements are Amazon's own; it can navigate them and in some cases act.
"Which of my FBA units are aging into a surcharge?"
Seller Assistant
Age of units in Amazon's warehouses is Amazon's own data. Nothing outside Amazon sees it sooner.
"How many units should I send into FBA this week?"
Seller Assistant
It is deciding how to move stock Amazon already holds, using Amazon's own demand signal and capacity limits, and it can stage the shipment for you. No outside tool has a better claim on this.
"How many days of cover do I really have?"
Dedicated tool
Needs every channel's velocity and every stock location, including your 3PL and in-transit units.
"Which PO do I fund this month?"
Dedicated tool
Needs open POs, factory MOQ, observed lead-time variability, and a cash budget. Those either sit outside Amazon entirely or exist there only as a static number you typed in.
Sanity-check every figure in the table against the underlying report, whichever tool produced it. Amazon's own numbers need it too, and so do ours: trailing velocity is wrong for any SKU that was out of stock, no matter who computes it (see the warning below).
That table is a scope boundary, not a failure list. Amazon built an assistant for the Amazon account, and inside that boundary it is the most direct answer available. Note that the highest-frequency inventory decision most sellers make, the weekly FBA shipment, falls on Amazon's side of the line and is free. The line falls where the question stops being "what do I move" and starts being "what do I buy."
This is where comparisons usually cheat, so let us split the inputs into two honest piles instead of one.
Genuinely absent. Amazon does not document Seller Assistant reading these, and you should assume it does not until Amazon says otherwise: your open purchase orders and their ETAs, your factory's MOQ and tooling constraints, your cash available for inventory, and your sales on any channel other than Amazon.
Present, but as a static number you typed in. Seller Central's Restock Inventory tool has a per-SKU lead time setting that you configure, and Amazon's own restock recommendations run on it. Cost inputs are the same story: Amazon's revenue and profitability surfaces accept a cost of goods figure. So it is simply not true that Amazon "has no lead time" or "has no cost." What Amazon has is the number you entered, usually once, probably a year ago. What it does not have is the distribution: how late your last ten POs actually landed, whether Chinese New Year adds three weeks, whether this supplier's variance is two days or twenty. Safety stock is computed from that variability, not from the mean, so a single typed-in lead time gets you a plan with no idea how wrong it might be. Same for cost: one static landed cost per SKU is fine for reporting margin and useless for ranking POs when freight and duty moved last quarter.
MOQ deserves its own note, because Seller Central does hold reorder settings. The MOQ that matters here is your factory's minimum order quantity and its tooling or material-lot floor, not the replenishment quantities you typed into Amazon's restock settings. Amazon knows what you told it to send in. It does not know what your supplier will accept as a production run.
None of that is a knock on the product. It is what "account-scoped" means, and a static seller-entered lead time is exactly the right design for a tool whose job is moving stock Amazon already holds. But it explains a specific failure mode: an assistant that only sees Amazon demand computes days of supply against an Amazon-only denominator, and if you sell meaningfully off Amazon, that number is optimistic by exactly the share of demand it cannot see.
If that describes you, the fix is upstream of any assistant: get every channel's velocity and every stock location into one count first, which is what multichannel inventory management is for. This is the same reporting-versus-deciding line we drew in AI vs traditional inventory management. Reporting what happened needs one system's data. Deciding what to buy needs four or five.
Take one SKU. The numbers below are illustrative and composite, built to make the arithmetic legible, not lifted from a customer account. Every figure is a seller input, not an Amazon fee.
Input
Value
In your Amazon account?
Amazon velocity, trailing 30 days
18 units/day
Yes (Reports > Business Reports > By ASIN > Detail Page Sales and Traffic)
FBA sellable on hand
1,040 units
Yes (Inventory > Manage FBA Inventory, Available column)
Units inbound to FBA
400 units
Yes (Inventory > Shipments, working plus in-transit plus receiving)
Supplier lead time, PO to FBA receipt
70 days
As a static setting you entered in Restock Inventory. The variance around it is not there.
Landed cost per unit
$6.40
As a cost figure you entered for profitability reporting. Not versioned against the freight you actually paid.
Shopify plus wholesale velocity
6 units/day
No
Factory MOQ
1,000 units
No
Safety buffer you have chosen to carry
21 days
No. This is a policy choice you make, not a number any tool reads
Cash budget for POs this month
$18,000
No
An Amazon-scoped view sees 1,440 units against 18 units/day and reports about 80 days of cover. That looks comfortable, and nothing flags.
Now add the 6 units/day moving off Amazon. Total burn is 24 units/day, so 1,440 units is 60 days of cover against a 70-day replenishment. You are 10 days short, so even if you cut the PO today you take a week and a half of stockout.
Then the sizing question. Work it the way our restock calculator does: set the target inventory position first, then net out what you already own. The 21-day safety buffer below is a separate decision from that 10-day shortfall, and the two numbers have nothing to do with each other. The buffer is what you have chosen to hold against demand and lead-time noise, and it is where observed lead-time variability would change the answer if you had it.
Order-up-to level = (lead time + review cycle + safety days) x total daily demand
(70 + 30 + 21) x 24 = 2,904 units
Order qty = order-up-to level - on hand - inbound
2,904 - 1,040 - 400 = 1,464 units
1,464 already clears the 1,000-unit MOQ, so no rounding is needed. MOQ is a floor, not an increment. Round up to a lot multiple only if your factory has a genuine tooling or material-batch constraint, and if it does, apply that multiple explicitly rather than assuming one. At $6.40 landed, 1,464 units is about $9,370, and on the 30/70 terms many overseas factories quote (yours may differ, some want 50/50 and a few extend net terms at real volume) the deposit due today is roughly $2,811, comfortably inside the $18,000 budget. You can run the cover-and-quantity half of this yourself in the FBA restock calculator.
The honest answer for both Amazon's assistant and ours: AI is very good at pulling the right numbers and proposing the decision, and it should not be the one signing the PO. That is the human-in-the-loop model we walk through in AI inventory management, and it is notably the same posture Amazon took by requiring seller permission before Seller Assistant acts.3
Where a dedicated tool differs is not intelligence, it is inputs. AI demand forecasting improves when it sees total demand rather than one channel's slice, and a reorder recommendation is only as good as the lead time and cash constraints behind it. If you want to point a general-purpose AI at your own numbers, that is what MCP is for, and we cover the setup in MCP for Amazon sellers.
Holding cost is the quiet part of this. Buying 4,000 units instead of 3,000 parks the extra cash in a warehouse that bills you monthly: standard-size FBA storage runs $0.78 per cubic foot per month off peak and $2.40 in October through December.6 An assistant that cannot see your cash budget also cannot weigh that trade.
Start with the honest headline, because it is not the one a vendor usually leads with: most sellers running one channel and a stable catalog do not need a paid inventory tool. Seller Assistant plus a well-built spreadsheet genuinely covers it. That includes plenty of people importing from Asia. A 90-day lead time is not itself a reason to buy software. What matters is whether the lead time is predictable, because a long but reliable one is handled by holding a fixed buffer and placing orders on a calendar, which is a spreadsheet's home turf.
A concrete profile of someone who should not pay us anything: 40 SKUs, Amazon-only, sourcing from one Vietnamese factory on a 75-day lead time that has landed within a week of promise on the last eight POs, ordering on a quarterly rhythm, no cash crunch. That seller should use Amelia for the weekly FBA shipment, keep one sheet for PO planning, and spend the subscription money on inventory.
Four conditions push the other way, and treat them symmetrically. None is a single-handed disqualifier, including the lead-time one:
Multichannel. Enough demand off Amazon that an Amazon-only denominator is meaningfully wrong, say 10%-plus of units.
Catalog size. More SKUs than you can actually review by hand each week. For most operators that starts somewhere around 75 to 100.
Lead-time variability. Not long lead times, unstable ones. Receipts landing three or more weeks off promise, often enough that a fixed buffer either stocks you out or bloats you.
Cash rationing. A budget that cannot fund every PO you would like to place, so orders must be ranked against each other.
Any one of these on its own is usually absorbable by a disciplined operator with a good sheet. Expect to cross the line when two land together, and to be well past it at three. Variability is worth a note because it changes the math rather than the workload, but a 20-SKU single-channel seller with swingy lead times still mostly needs a safety-stock formula and a habit of logging receipt dates, not a subscription.
Then price it, since an ROI claim with the cost redacted is not a claim. Inventory Hero starts at $79 a month, about $950 a year. Against that, one avoided stockout on a mid-volume SKU or one PO sized 1,000 units too large usually clears the bill on its own. If you cannot point to a plausible instance of either in your own last twelve months, the free assistant is the correct answer and you should stop reading here.
If you are near the line, run a test that can actually fail. The trap version is to recompute with different inputs and call divergence proof, but if you swap in cross-channel demand and a different lead time, the numbers must differ, by construction, and you have learned nothing. So hold the inputs constant and compare method instead. Take your top five SKUs, write down Seller Assistant's suggested shipment quantity, then compute the same number in the FBA restock calculator feeding it the same demand figure and the same lead time Amazon is using. Two things can happen. The answers agree within a week of cover, which is the expected result for a single-channel seller with stable lead times, and the finding is that the method is sound and you need no tool. Or they diverge on identical inputs, which means one of them is modeling the buffer or the review cycle differently, and that is worth understanding before you trust either. Only after the methods reconcile is it meaningful to swap in your real total demand, because then any remaining gap is attributable to the missing data rather than to a difference in arithmetic.
Use Amazon's Amelia seller assistant for what it is genuinely best at: account questions, Account Health, compliance, aging FBA stock, and shipment recommendations from inventory Amazon already holds. It is free, native, and there is no good reason to route those questions anywhere else, and for a lot of sellers that is the whole answer. Reach for something with a wider data scope only when the decision turns on what Amazon cannot see: your other channels, your open POs, your cash budget, and the observed spread of your lead times rather than the single number you once typed into Restock Inventory. That is a data problem, not an intelligence problem, and no amount of model quality substitutes for it. If you do go shopping, compare us against other inventory tools, which is the real peer set, not against a free assistant most of our own users keep running alongside. If you want more of your Amazon operation running on autopilot in the meantime, FBA automation covers what is safe to hand over today, and conversational analytics for FBA covers what asking your own data looks like when it spans every channel.
Amazon, "Amazon launches a powerful new generative AI-based selling assistant codenamed Project Amelia," aboutamazon.com, September 2024. Announced at Accelerate as a beta rolling out to an initial set of US sellers, built on Amazon Bedrock, covering knowledge-based questions and summaries of recent sales, units sold, and traffic, with actions and issue resolution described as coming soon. ↩↩2
Amazon Seller Central Help, "Seller Assistant, Seller Central's agentic assistant," sellercentral.amazon.com/help/hub/reference/external/GLYJTRGQCYNYAZS9 (accessed August 2026). Source for the insight categories (Sales Performance, Inventory Management, Listings and Catalog, Shipments and Returns, Orders and Fees) and the action examples (resolving some policy violations, optimizing FBA surcharges, creating promotions), plus Account Health monitoring, compliance automation, inventory alerting, ad creative, and customer behavior analysis. ↩↩2↩3
Amazon, "Amazon unveils new agentic AI-powered Seller Assistant to help independent sellers grow," aboutamazon.com/news/innovation-at-amazon/seller-assistant-agentic-ai, announced at Accelerate, September 2025. States the assistant can reason, plan, and take action with the seller's permission; available to all sellers in the US store at no additional cost, rolling out to other countries in the following months. ↩↩2
Amazon Selling Partners, "Amazon Accelerate 2025: A new era of tools and technology for independent sellers," September 18, 2025. Source for the inventory behaviors: flagging slow movers ahead of long-term storage fees, recommending leave-as-is versus markdown versus removal, and preparing shipment recommendations from demand patterns. ↩
Amazon, "2025 Small Business Empowerment Report," aboutamazon.com (accessed August 2026): "In 2025, Seller Assistant, an always-on, agentic AI expert for sellers, had more than 230,000 monthly users, with sellers accepting its recommended actions more than 90% of the time." ↩
FBA monthly storage rates, effective 2026-01-15, per Amazon's published pricing (sell.amazon.com/pricing), mirrored in Inventory Hero's reference-data registry so our articles and calculators agree: standard-size $0.78 per cubic foot per month off peak (January to September) and $2.40 peak (October to December). Confirm your own product's live rate in the FBA fee preview and your monthly storage-and-fees report, since size tier and utilization surcharges can change what you actually pay. ↩