Skip to main content
ftes.aiftes.ai
E-commerce · 30 roles · 9 departments

The money is already in your store.

Nobody is employed to go and get it. These are the roles that do — cash stuck in stock, margin that was never really there, returns coming back for fit, discounting nobody can see. They work inside the store you already run, and nothing they do happens without your approval.

Read-only to start. No card. Nothing to type.

Frozen

Cash you already spent

Stock sitting at landed cost while the decision to move, mark down or hold waits in a spreadsheet.

Leaking

Margin leaving quietly

Discount stacking, shipping subsidy, returns and carrying cost — each one a cost your platform does not load in.

At risk

Revenue about to not happen

Proven sellers running out, size runs broken, and best customers already past their own buying rhythm.

Four jobs your store pays for and nobody holds

None of these is a failure of effort. They are the jobs that never got a job description, because they sit between the tools and nobody has an afternoon free. Each one has a role here, and each role is judged on one number.

  • 01

    Nobody is employed to get the cash out of the stockroom

    Every unit you are holding was paid for with cash that is now unavailable, and it keeps costing you while it sits — capital, storage, the risk it ages out, and the buy you could not make instead. The decision to transfer it, mark it down or hold it is a real weekly decision with real money on both sides of it, and in most stores at this size it is made once a quarter by whoever has an afternoon free.

    The role that owns it

    Cash Officer

    Judged on: Cash released from stock

    Hire this role
  • 02

    Nobody is employed to work out what you actually earn

    Your platform shows revenue minus what you paid the supplier and calls it gross margin. It does not know what the freight cost, what the duty and the brokerage cost, what the discount actually given was once codes stacked, what that product's own return rate takes back, what the shipping and the payment fees took, what the customer cost to acquire, or what carrying the unit cost for every day it sat. Load all of it and the ranking of your catalogue changes.

    The role that owns it

    Controller

    Judged on: True contribution margin, to the order line

    Hire this role
  • 03

    Nobody is employed to find out why it came back

    A return is not one cost. It is the outbound shipping you already paid, the return shipping, the handling, the support contact, and the unit itself — which may not resell at full price, or at all. The reason codes that would tell you which styles are doing this, and which vendor runs a size small across a whole range, are sitting in your returns platform, unread, because reading them is nobody's job.

    The role that owns it

    Fit Officer

    Judged on: Return rate and return cost

    Hire this role
  • 04

    Nobody is employed to count what the discounts really took

    You know the headline rate of each code. You do not know the realised rate — what came off once codes stacked, staff discounts applied and shipping was subsidised, on which products, in which shop, by whom. A discount decision takes one click and its cost lands three weeks later inside a number nobody decomposes.

    The role that owns it

    Pricing Officer

    Judged on: Realised discount rate by product, shop and code

    Hire this role

Four jobs, four roles, and each one reports to you. Start with whichever is costing you the most.

Get started

Why the tools you already run cannot close them

What your stack is genuinely good at

  • Taking the order, holding the catalogue, moving the money
  • Showing you revenue, sessions, conversion and what each channel spent
  • Running the email and the flows you already wrote
  • Processing the return once the customer has started it
  • Getting the parcel out of the building

What none of it holds

  • Landed cost — what the unit actually cost you by the time it reached the shelf
  • Supplier terms, lead times and what the last shipment really did
  • Warehouse performance — where orders get stuck and what that costs
  • Helpdesk history — what customers are telling you, at product level
  • Return reason codes, and the fit verdict hiding inside them
  • Till data from the shops, and the stock decision that spans both
  • Marketplace orders, sitting outside the platform entirely

None of this is a criticism of the tools. Each is good at the half of the job that ends with a chart. The half that begins after the chart — deciding, preparing the work, doing it inside limits and proving what it earned — is not a feature any of them is built to have.

Nothing here replaces your platform. It reads what your platform cannot see, and brings you the decision.

Get started

This is what it looks like at 7am

Not a dashboard you have to interrogate. A page that already did the work overnight and brought you the few things worth your signature — each one with the money attached, the evidence it was computed from, and a button.

Worked example

Good morning — Thursday

Yesterday · 614 orders · compared with the same day last week

Needs attention

  • Returns waiting more than three days58 returns · oldest 105 days
  • Orders not packed in five days214 orders · across three locations
  • One shop converting below the others17 points below the best-performing shop

Today’s decisions

Send managers their overdue-order lists — $88,654 unshipped, over 48 hours, across 291 orders

291 orders crossed the 48-hour mark. The lists are ready, split by location.

Computed from synced orders

  • Online / unattributed · 199 orders · $48,656
  • High-street store · 44 orders · $27,854
  • Outlet store · 30 orders · $9,481
  • Airport store · 18 orders · $2,663

paid, not cancelled, fulfilment pending over 48h

Review & sendViewSkip todayPrepared overnight · nothing sent yet

Move stock from storage to the shops that are selling it — $5,022 ready to sell once transferred

Three best-sellers are stocked out at shops while units sit in storage.

Inventory and 28-day sales velocity

  • Ribbed knit dress · black → High-street store · 38 units · sells 3 a week
  • Ribbed knit dress · black → Airport store · 26 units · sells 4 a week
  • Wide-leg trouser · navy → Outlet store · 31 units · sells 2 a week

selling locations at or under 2 units; storage holding over 10

Review & sendViewSkip todayPrepared overnight · nothing sent yet

Win back the best customers who have gone quiet — 3,170 customers past their own usual gap between orders

Each one is quiet by their own buying rhythm, not by a rule of thumb applied to everybody.

Customer order history

  • Top fifth by lifetime value, three or more orders
  • Silent for twice their own usual gap, minimum 30 days
  • Message drafted per segment, nothing sent yet

drafted for your review — no message leaves without approval

Review & sendViewSkip todayPrepared overnight · nothing sent yet
  • Read-only connection to your store
  • Data refreshes on a schedule, and every answer says when it last synced
  • Every decision is logged and auditable — nothing acts without your approval

A worked example, with illustrative figures. On your store the roles, the decisions and the numbers are your own.

That page is built from your own store overnight. The first one arrives the morning after you connect.

See it on your store

Revenue minus product cost is not a margin

It is a comforting figure that ignores everything that happened between the factory and the customer's wardrobe. The Ledger loads every line below onto each order line, then rolls it up by product, variant, size, colour, vendor, channel, location, campaign, code, customer and cohort.

Worked example
One order line✦ = your platform does not load this in
  • Sold at$120.00
  • Supplier cost, at the rate on the day−$38.00
  • Inbound freight, insurance and handling — not loaded in by a store platform−$4.10
  • Customs duty — not loaded in by a store platform−$3.40
  • Brokerage, per customs entry — not loaded in by a store platform−$1.90
  • Shrinkage provision — not loaded in by a store platform−$0.60
  • Discount actually given, after stacking — not loaded in by a store platform−$18.00
  • Returns, from this product's own rate — not loaded in by a store platform−$12.40
  • Outbound shipping−$7.20
  • Payment fees−$3.10
  • Cost of acquiring the customer — not loaded in by a store platform−$14.00
  • Carrying cost, for every day it sat — not loaded in by a store platform−$2.80
  • Contribution$14.50

A worked example on one order line, with illustrative figures. The cost lines are the real ones the Ledger loads.

Costs are dated
Changing a cost today never silently rewrites last quarter's margin. The figure you signed off stays the figure you signed off.
Coverage is always shown
If only part of a category's cost is known, the page says which part. A missing source is named as missing — nothing is ever displayed as zero when the truth is that we could not read it.
Cost data is the upgrade, not the entry fee
Landed cost needs freight invoices and supplier terms from you. Nothing about the first findings depends on it — where a finding would be sharper with it, the page says exactly which data and what it would reveal, without locking the finding away.

Your catalogue ranked by what each product actually earns, not by what it appears to earn.

Find your real margin
The org chart

9 departments. 30 roles. Hire the ones you need.

Read it like an org chart, not a feature list. Every role has a job description, a scope of authority, a manager (you), and one number it is judged on. Dismiss any of them in one click.

01

Office of the Owner

Judged on · Decisions taken and landed · 2 roles

02

Finance

Judged on · True margin, cash released, dollars recovered · 3 roles

03

Merchandising

Judged on · Sell-through and markdown avoided · 5 roles

04

Product Quality

Judged on · Return rate and return cost · 3 roles

05

Customer

Judged on · Repeat rate and owned-channel revenue · 5 roles

06

Demand

Judged on · Contribution per channel, citation share · 7 roles

07

Operations

Judged on · Shipping as a share of revenue, promise kept · 3 roles

08

Retail

Judged on · Sales per trading hour · 1 role

09

Supply

Judged on · On-time delivery and vendor quality · 1 role

+

Available to every industry

The marketing roles with a full workspace today · 3 roles

Compare every role on the Workforce page

Every role starts at Level 0: it proposes, you approve, and every claim carries its evidence.

Inside your stack

It works in the systems you already run.

Connections are granted per organization and per property, with the permissions you choose, and can be revoked at any time. Integration credentials are encrypted at rest.

  • Shopify
  • Shopify Plus
  • WooCommerce
  • BigCommerce
  • Adobe Commerce (Magento)
  • Amazon Seller Central
  • Walmart Marketplace
  • eBay
  • Etsy
  • Klaviyo
  • Mailchimp
  • Gorgias
  • Zendesk
  • Google Merchant Center
  • Google Analytics 4
  • ShipStation
  • Recharge
  • Yotpo
  • Stripe

We tell you when we were wrong

Every tool in this market gives advice, and none of them comes back to say whether the advice worked. That is why so many get cancelled in a budget review by somebody who never used them. Three mechanisms, working together, so you never have to take our word for anything.

  1. Before

    The prediction is declared

    A recommendation cannot even be displayed unless it states, in advance, what it expects to happen, over what period, measured how, and with what confidence. A recommendation without a prediction is not shown at all.

  2. After

    The measurement runs itself

    When the horizon arrives the system runs the declared measurement, records the actual result, computes the error, updates that role's accuracy score, and publishes it — whether or not the news is good.

  3. Monthly

    The value ledger posts

    A running, double-sided record: opportunity identified on one side, value realised on the other, each entry tied to a specific action and its measurement. Plus the figure most vendors would never show — forgone, the value of the opportunities you declined or let expire.

Four rules that keep it honest

  • Only measured value is ever billable. Never identified, never predicted, never estimated.
  • Where a range exists, the ledger records the low end.
  • Opportunity converts to realised — it never adds to it. Double counting is prevented structurally, not policed.
  • The method is published, and you can dispute any entry.

Causal, or labelled otherwise

Where a holdout is practical we use one — withholding a recommendation from a randomly chosen control group, so the difference is causal rather than coincidental. Where it is not practical the measurement is labelled observational, plainly, on the page. We do not call a correlation a result.

Every recommendation states what it expects before it runs, and reports what happened after.

Get started

It starts by watching. It earns the rest.

Autonomy is earned per role and per action type, on measured accuracy. It is never a settings toggle you flip on the first afternoon.

LevelWhat the role may doHow it is earned
L0WatchReads and reports. Touches nothing.The default, from day one
L1RecommendProposes a specific action with the money attachedOnce the ledger is verified against your own books
L2Act on approvalPrepares the action completely; you press one buttonA full cycle at L1 with its predictions landing
L3Act within limitsActs alone inside written limits — value cap, category, hours, blackout datesA further cycle at L2, and you sign the limits
L4AutonomousActs alone; you read a daily summaryYour decision only. Never automatic. Reversible instantly.

Most managers run most roles at Watch or Recommend indefinitely, and that is a perfectly good outcome. The higher rungs exist for the repetitive, low-value, high-volume work where a human approval adds nothing but delay.

Five guarantees, on every action, at every level

  1. 01Nothing acts without being declared. A role cannot perform an action type that is not in its register.
  2. 02Limits are written, not implied. An action that would exceed one is refused and queued — never quietly trimmed to fit.
  3. 03Everything is logged immutably: who, what, when, on what evidence, under whose approval, with what prediction, and what happened.
  4. 04Everything reversible is reversible in one click, and anything that is not reversible looks different at the moment you approve it.
  5. 05One button stops everything — per role or globally — and the state is visible everywhere.

Start every role at watch-only. It reads, it reports, and it touches nothing until you say so.

Start at watch-only

What it replaces, and what it does not

Likely to replace

  • Standalone lifetime-value and cohort tools
  • Inventory forecasting and replenishment add-ons
  • Spreadsheet open-to-buy, cash and margin models
  • The hours a week spent reconciling one dashboard against another

Likely to reduce

  • Attribution and analytics platforms, kept for creative and channel work
  • Agency retainer hours spent on reporting and reconciliation
  • Custom reporting and business-intelligence seats
  • Returns analytics add-ons

Will not replace

  • Your commerce platform
  • Your helpdesk
  • Your email and SMS platform
  • Your returns portal and your third-party logistics provider

The third column is the important one. A consolidation claim with nothing in it is not believed by anyone who has bought software before.

Who this is not for

Saying this plainly saves both of us a month. If one of these is you, the audit would only confirm it, and we would rather tell you now.

  • Stores below roughly two million in revenue holding little stock. The numbers are too small to pay for the work, and a spreadsheet is genuinely fine at that size.
  • Pure dropshippers holding no inventory. Half of this is about the cash inside stock. Without stock, half the value is gone.
  • Anyone who wants a customer-support bot. That job is solved and sold well by others — we read the helpdesk, we do not replace it.
  • Anyone who wants software to run the business without them. Every role here works for a human, reports to a human, and stops when a human says stop.

The questions operators actually ask

Is this just another dashboard?
No. Every screen either supports a decision or proves an outcome, and the headline figure is not a metric — it is money released, measured. If a screen did nothing but report, it would not be here.
My commerce platform's own AI will do this for free. Why pay?
For the parts it can see, largely yes, and we say so. It does not hold your landed cost, supplier terms, warehouse performance, helpdesk history, return reason codes, till data or marketplace orders — and it has no commercial reason to tell you that a decision made inside it cost you money.
AI makes things up. How is this different?
Every number comes from a tested query against your own data. The AI chooses which calculation to run and writes the sentence; it is structurally incapable of inventing a figure. Every answer names the systems it read and when they last synced.
I cannot let software touch my store.
Then do not. Every role starts at watch-only and stays there until you decide otherwise. Each step up is earned on measured accuracy and signed by you, every action is reversible, everything is logged, and one button stops all of it.
How do I know it is worth the money?
Because it tells you, monthly, in dollars, with last month's accuracy beside it and a drill-down to every decision behind the figure. Only value it actually measured counts, and you can dispute any entry.
We already have someone doing this.
Then they get their evenings back, and they get numbers they can defend in a meeting. The workforce does not replace an operator's judgement — it removes the assembly work that stops them using it.

Start one role at watch-only

Pick the job costing you the most — cash in stock, the margin you cannot see, returns coming back for fit — and let that role read your store and report. It proposes, you approve, and it tells you afterwards what it earned.