Executive AI transformation forretail, consumer brands and e-commerce.
For retailers, direct-to-consumer brands, wholesalers and multichannel businesses. Where AI creates value when contribution per order is decided by forecasting, returns and service — and what has to change before it shows up in margin.
For businesses that sell to consumers. If you make products for other businesses, see Manufacturing.
Contribution per order is thinner than the headline margin suggests.
Once returns, delivery and service are counted, the picture changes. And most businesses have already bought AI licences without any of it moving — technology without transformation, which is the most common way this money gets wasted.
Forecasting is guesswork at the edges
Get it wrong upwards and cash sits in stock. Get it wrong downwards and you lose the sale entirely.
Returns eat the margin
The cost of processing, restocking and writing down returns is often larger than anyone has calculated.
Service volume scales with orders
More growth means more contacts, and the cost line grows with the revenue line.
Product content is a bottleneck
Listings, descriptions and marketplace variants delay getting product live and earning.
Promotions are hard to judge
Volume moves, but whether it moved profitably is usually decided long afterwards.
Working capital sits in stock
Availability and cash pull in opposite directions, permanently.
Three things we hear in almost every conversation.
In the words people actually use. If none of these sound like you, that is useful to know in the first ten minutes.
“Returns are eating the margin and we handle them all the same way.”
The loaded cost of a return is almost always higher than the business thinks it is.
“Every extra order brings an extra service contact.”
Growth doubles the cost line unless the work changes.
“Product sits in the warehouse for three weeks before it’s live.”
Stock bought, cash committed, and nothing earning until the listing exists.
Five places where contribution is made or lost.
Each one has a different lever. AI is only worth doing where it moves one of them.
Planning demand
Forecasting, range, buying. The lever is availability against stock turn.
Getting product live
Content, listings, pricing. The lever is speed to earning.
Acquiring the customer
Marketing, promotion, channel mix. The lever is cost per acquisition.
Fulfilling the order
Picking, delivery, exceptions. The lever is cost to serve.
Serving and retaining
Enquiries, returns, repeat purchase. The lever is contribution per customer.
The repeatable work that carries the money.
This is the level at which change actually happens. Not “the business” — these processes.
Demand forecasting and range planning
Buying and replenishment
Product content and listing creation
Pricing and promotion
Order fulfilment and delivery exceptions
Customer service enquiries
Returns triage and processing
Marketplace and channel management
Patterns worth examining, stated as questions.
Whether any of these applies to your organisation is exactly what the paid work establishes. We would rather call these patterns than dress them up as proof.
Demand forecasting support
Could forecasts account for promotions, seasonality and channel behaviour together, rather than being adjusted by hand?
Product content generation
Could listings and descriptions be drafted from your own product data for a person to approve, cutting time to live?
Returns triage
Could returns be assessed and routed on arrival — resell, refurbish, write off — instead of being handled uniformly?
Customer service enquiries
Could routine questions on orders, delivery and stock be answered from your own systems, so the team handles real exceptions?
Personalised engagement
Could customer communication reflect what someone actually bought and returned, rather than a broad segment?
Promotion analysis
Could the profit effect of a promotion be understood in days rather than at the end of the quarter?
Each one is marked for evidence. Research-backed means we see it in the sector. Practitioner-validated means people who do this work have confirmed it. Nothing here is claimed as true of your business until your own people have said so.
Notice what is not on this list: anything that starts with a tool. The question is always which number moves.
Buying, service and content have to change together.
These processes are connected. Changing one without the others usually moves cost rather than removing it.
The buyer’s job changes
From building forecasts to challenging them. Judgement moves to the exceptions, which is a different skill.
Returns policy becomes a commercial decision
Triage only pays if someone owns what happens to each outcome, and has authority to change the policy.
Content approval has to be defined
Brand voice and accuracy still need a human owner. Decide what ships without review and what does not.
Product and stock data has to be clean
Forecasting and content are only as good as the underlying record. Usually the real first project.
Service exceptions need an owner
When routine contacts are handled automatically, what remains is harder and needs a service level.
The measures change
Not “contacts deflected”. Forecast accuracy, contribution per order, returns cost and repeat purchase rate.
Numbers already on your trading report.
We set these before anything is built. A baseline taken beforehand is evidence. One reconstructed afterwards is an argument.
Forecast accuracy
By category and by channel.
Stock turn and availability
Together, not separately.
Returns rate and cost
Per return, fully loaded.
Contribution per order
After returns, delivery and service.
Service contact rate
Contacts per hundred orders.
Time to live
Product received to earning.
Repeat purchase rate
And contribution per customer.
Working capital in stock
And weeks of cover.
You probably already own most of what a first case needs.
Microsoft 365 Copilot for content and analysis work. Copilot Studio agents for defined steps such as order enquiries or returns triage. Power Platform and Dynamics 365 where commerce and service workflow have to change. Azure and your data estate where product, stock and order data need to reconcile across channels.
We sell none of it, resell none of it and build none of it. Your technology partner does that, and keeps the work.
Every AI programme needs both capabilities. We do the commercial one and never compete for the implementation work.
Forty-five minutes, in the language of contribution per order and stock turn.
Executive AI Opportunity Assessment
Where AI could create value in an organisation like yours, and whether it is worth going further.
Executive AI Briefing
One day with your leadership team to agree where to act first, and record why.
AI Accelerator
Thirty days: the work redesigned, the case built in your numbers, the critical piece proven.
Our retail and consumer view is a working model. It gets sharper with every engagement, and we will always tell you which parts are evidence and which are still assumption.
Find out where AI pays across your channels.
Forty-five minutes with your leadership team, and a straight recommendation either way.
Book a free assessment