AI Adoption
·
9 min
All resources
AI Adoption
·
·
9 min read
AI in retail: 81% have tried it, 1 in 4 have integrated it (the gap is your opportunity)
In retail the AI conversation almost always starts with cost — and it's the wrong diagnosis: the tools are cheap and available, often already inside the platforms you use. Only 15.7% of Italian SMEs use at least one AI technology (up from 7.7% in 2024, versus 53.1% of large firms), and 81% make some use of it but only 1 in 4 have integrated it into their workflows. That gap is Innesti's promise in one sentence: access isn't the problem, implementation is. Where it pays off (demand forecasting, personalization, service, in-store), how to pick a single workflow and take it past the pilot, and why the barrier isn't cost but the skill to integrate — with compliance (EU AI Act, customer data) attached to every workflow.
In retail — a shop, a chain of stores, an online store — the AI conversation has a recurring flaw: it starts with cost. "I don't have the budget for artificial intelligence." It's the most common line, and it's almost always the wrong diagnosis. Because in retail, unlike nearly every other sector, the cost of the tool was never the real obstacle: the tools are cheap, available, often already inside the platforms you use. What's missing is something else — and it's exactly what makes this sector the clearest case of all.
The numbers tell a precise story. Only 15.7% of Italian SMEs use at least one AI technology (2025), against 53.1% of large enterprises: a wide gap, even though adoption has more than doubled from 7.7% in 2024. But the figure that really counts is another one: 81% of SMEs report some use of AI, and only 1 in 4 have integrated it into real workflows. The two figures don't contradict each other: they measure different things — "I tried a tool" versus "I made it part of how I work." And that gap, for a retail business, is Innesti's promise said in a single sentence: access isn't the problem, implementation is.
Cost isn't the problem (and it never was yours)
It's worth pausing here, because this is the point that changes everything. In manufacturing the barrier to adoption is openly cost, so much so that the State co-funds the work. In a professional practice AI is already included in the software subscription and sits switched off. In retail the picture differs from both: the tool costs little or nothing, it turns on the same day — what's missing is the internal skill to take it past the pilot, to design the process around it and make it scale.
Put differently: in retail the barrier isn't economic, it's about skills and process design. Most businesses "try" a tool — a chatbot, a recommendation engine, a demand forecast — keep it running for a few weeks and never really integrate it. The 3 in 4 who don't integrate didn't pick the wrong tool: they were missing exactly what we mean by "we implement it" — the workflow around the tool, not the tool.
Where AI pays off in retail
Four workflows in retail are high-volume, repetitive and rule-driven — AI's natural ground, and nearly all reachable with tools that don't require an enterprise budget:
- Demand forecasting and inventory management — instead of reorder spreadsheets kept by hand. It's the largest AI spend category in retail (22.8% of the budget): machine-learning forecasts hit 82–88% accuracy at the individual-item level, against 65–75% for statistical methods — which translates into roughly 14% less unsold stock to write down and 11% fewer stockouts. It's the workflow with the most measurable return.
- Personalized recommendations and offers — instead of generic promotions the same for everyone. Personalization brings 24% higher click-through and 18% higher average order value versus generic; 89% of marketers report a positive return from personalization.
- Customer service and inquiry deflection — instead of handling every repetitive question by hand. The documented return is around 3.5 times the amount invested, with a cost per interaction lower by 30–40%: the simple questions are handled by AI, the cases that need judgment stay with people.
- Repetitive and informational in-store tasks — instead of manual stock checks and price or information lookups. It's where in-store AI concentrates today: work that eats time but doesn't require judgment — the same "automate the grind, not the decision" pattern as every department and sector in this series.
The criterion for starting isn't "which is the most impressive," but "which takes away the most repetitive hours with the sharpest before/after": for most retailers the answer is demand forecasting — high volume, clear rules, and a return you can read in unsold stock and stockouts avoided.
The gap between "tried" and "integrated" is your opportunity
Here is the part that flips the reading, and that almost no vendor tells the right way. That only 1 retailer in 4 has integrated AI isn't bad news: it's the snapshot of an advantage still available. In a market where 81% have already tried something but three-quarters stalled at the pilot, whoever takes even a single workflow all the way to real integration pulls ahead of three competitors out of four — not by buying more technology, but by using it better.
That's why, in retail, the right question isn't "which AI tool am I missing," but "which of the ones I already have — or can get tomorrow at minimal cost — have I not yet integrated into how I actually work." Whoever reads AI as a purchase gets half of it; whoever reads it as a workflow to take past the pilot gets the point — and it's the reading on which we build the process design.
The numbers, read honestly
The picture is of a sector on the move, not a gamble. Adoption among SMEs has more than doubled in a year — from 7.7% in 2024 to 15.7% in 2025 — and in e-commerce penetration is already high: 68% of Italian businesses with over half a million euros in revenue have integrated at least one AI tool. This isn't a niche fad: it's a sector where the front of the pack has already begun to pull away from the tail.
Honesty, though, lies in not confusing "tried" with "integrated." That 81% of SMEs "use AI" doesn't mean 81% are getting value from it: it means the tools are there, often half switched on and with no workflow around them. The gap, as in every department and every sector in this series, isn't between those who have AI and those who don't — it's between those who have redesigned one process around a tool and those who leave it running on empty. The distance between the 81% who "use it" and the business that really gains is entirely there.
And compliance? Within limits, but with method
In retail, AI touches two sensitive areas: customer data (personalization, recommendations, profiling) and the automated decisions that affect them. The transparency, information and human-oversight duties of the EU AI Act, together with the already-familiar data-protection rules, apply directly to whoever personalizes offers or automates service. It's not a free zone just because the tool is cheap: clarity on how and where customer data is processed, minimization, and a documented human oversight over decisions that weigh on the customer remain the premise, not an extra.
It's exactly what our compliance overlay attaches to every workflow we design: so the value the workflow generates comes from a process that is not only efficient but also defensible. For a retailer, this isn't extra bureaucracy — it's customer trust handled with the same care as the margin.
Where to start, in practice
If you run a retail business and AI is on your radar, the sensible path is short and — surprise — almost always low-cost:
- Forget the budget, start with the workflow — the tool almost always already exists and is cheap. The first question isn't "how much does AI cost," but "which high-volume workflow has the sharpest return if I integrate it."
- Pick a single process — demand forecasting or customer service give the most legible before/after. Not a rollout across everything, one workflow.
- Integrate, don't "try" — the difference between the 3-in-4 who stall and the 1-in-4 who gain is entirely in taking the tool inside the process: who enters what, where the human decides, how the result feeds real orders.
- Measure on inventory, not on features — less stock written down, stockouts avoided, higher average order value: the number that counts is the business result, not how many checkboxes you switched on.
Before even choosing the workflow, though, it helps to know where you stand: our AI-readiness assessment helps you understand which process to start from for the most return and least friction, and which controls to put around the first workflow. If the issue is the compliance of what touches customer data, our compliance overlay explains how we attach the controls to every design.
We've turned the first step into a free, self-serve assessment: a few questions and a pointer on where to start, with which controls around it. Take the AI-readiness assessment — then, if it makes sense, let's talk.
This article is for guidance only. The adoption and return figures cited (15.7% of Italian SMEs using at least one AI technology in 2025, up from 7.7% in 2024, against 53.1% of large firms; 81% with some use but only 1 in 4 with workflow integration; 68% of e-commerce above half a million euros in revenue; forecast accuracy of 82–88% against 65–75%, with roughly 14% less unsold stock and 11% fewer stockouts; 24% higher click-through and 18% higher average order value from personalization, with 89% of marketers reporting a positive return; roughly 3.5 times the return and 30–40% lower cost per interaction in service) come from sector sources and should be read as indications of direction, not guarantees of results. Any automation that touches customer data must be assessed on the data, the controls and the context of the individual business.
Keep reading
More deep-dives on AI adoption in an SME.
From theory to your business. We graft AI in.
Want to know which department to start from in your company? The free assessment gives you a first answer in two minutes — then, if it makes sense, we talk.