AI and E-commerce for Omnichannel SMEs: Personalisation, Recommendations and Dynamic Pricing Without a Giant's Budget

You don't have to be Amazon, but the comparison is with Amazon

When a customer shops online, the benchmark is not the shop across the street: it is Amazon. That is where they have learned to expect spot-on recommendations, searches that understand what they are looking for, pages that seem built to measure. For an SME that sells both in-store and online, this expectation is bad news and an opportunity at the same time. Bad, because no small business can compete with a giant's technological infrastructure. An opportunity, because for the first time the very AI tools that power those giants have become accessible even to those running an online shop of ordinary size.

The numbers describe a mature market that is still largely unexplored on the AI front. According to the B2c eCommerce Observatory at the Politecnico di Milano, Italian e-commerce surpassed 54 billion euros in 2025, growing at double digits. Yet, again according to the Observatory, only 38% of Italian e-commerce businesses use at least one artificial intelligence solution, and just 15% have integrated it into the processes that really count. Translated: the majority of Italian online shops, including a great many SMEs, are leaving on the table a competitive advantage that more organised competitors are already picking up.

This article focuses on three concrete levers, the ones every omnichannel SME's roadmap should assess: personalising the experience, product recommendations and dynamic pricing. The first two are almost always a good investment. The third is powerful but treacherous, and it is worth explaining why.

Personalisation: treating every customer as if you knew them

Personalisation is the heart of it all, and for an omnichannel SME it has a particular meaning. Those who sell both in-store and online have something pure digital players often lack: a real relationship with customers, made of faces, habits, preferences observed at the counter. AI makes it possible to bring that same attention to the online channel too, where otherwise every visitor would be a stranger.

Concretely, personalising means adapting the experience to the behaviour of the individual person. A customer who has already looked at winter coats may find, on their next visit, scarves and gloves brought to the fore. The homepage can change according to what that visitor has explored in the past, emails can propose different content depending on purchase history, the very layout can reorganise itself around the interests shown. This is not science fiction: these are functions now available even in tools within SMEs' reach.

The advantage is measurable. Research by McKinsey estimates that omnichannel personalisation can increase revenue by between 10% and 15%. On the direct marketing front, personalised emails show significantly higher open rates than generic mailings, a sign that customers respond when they feel recognised. And here lies the delicate point for an SME: the value does not come from the technology in itself, but from the data that feeds it. An omnichannel e-commerce business that manages to unify what it knows about a customer from both channels, in-store purchases and online browsing, builds knowledge that no generic algorithm can replicate. It is exactly the kind of advantage that AI amplifies rather than replaces.

Recommendations: the expert shop assistant, multiplied

Product recommendations are the most visible and most profitable form of personalisation, the "products you might like" and "customers who bought this also bought". They are the expert shop assistant who, in a good shop, suggests the right pairing or the accessory you need. AI does the same thing across thousands of customers at once, analysing browsing, purchases and preferences to propose what genuinely makes sense.

The data on this front is among the most solid in the sector. According to the B2c eCommerce Observatory at the Politecnico di Milano, Italian e-commerce businesses that adopt AI-based recommendation systems report increases in conversion rate on the order of 15-35% and in average order value between 10% and 25%. These are wide ranges, because the result depends heavily on data quality and product assortment, but the direction is clear: recommending well sells more and raises the value of each order.

For an omnichannel SME, recommendations work across three different moments. On the product page, by suggesting alternatives and complements. In the cart, by proposing the addition that completes the purchase without being intrusive. And post-purchase, with emails that propose the right product at the right time. One precaution matters more than any other: recommendations must be relevant, not insistent. A system that fills every page with disconnected suggestions achieves the opposite effect, it irritates and lowers trust. Here too, the prerequisite is the quality of the behavioural data: without a meaningful history, the algorithm is just guessing.

It is worth adding a note of realism, because inflated figures circulate online on this topic, with promises of revenue increases of 300% or the like. These are numbers to take with enormous caution, almost always referring to exceptional cases or to large marketplaces with vast inventories. For an SME, the reasonable expectation is a real but gradual improvement in conversions and average order value, not a doubling of revenue.

Dynamic pricing: powerful, but to be handled with care

The third lever is the one that requires the most attention, and it is right to treat it with the honesty it deserves. Dynamic pricing consists of automatically adjusting prices based on variables such as demand, competition, seasonality and stock levels. The theoretical advantages are concrete: some analyses estimate improvements in gross margins on the order of 5-12% compared with manual pricing, as well as the ability to clear stock with automatic discounts on slow-moving products and to stay competitive by monitoring competitors in real time.

So much for the theory. The practice, for an SME, is more delicate. The first risk is customer perception. When people notice that a price changes according to the moment, the device or the browsing history, they may feel treated unfairly, and trust in a small brand is far more fragile than trust in a giant. The most striking case is recent: the tickets for some concerts sold with dynamic prices that more than doubled within minutes sparked controversy and even the attention of the European competition authorities. An SME cannot afford reputational damage of that kind.

The second risk is technical. Automating prices without a solid data base means amplifying errors at scale: if margins are already tight or the history is insufficient, the algorithm can lead to selling below cost or triggering price wars to the bottom. And for an omnichannel company there is an added pitfall: consistency across channels. Different prices between the website, marketplaces and the physical shop can confuse the customer and cannibalise the sales of one channel in favour of another.

This does not mean giving it up, but tackling it with clear rules. The best practices indicated by experts are simple and non-negotiable: set minimum and maximum price thresholds below which the algorithm can never go; limit the frequency of changes so as not to disorient; start from real costs and not just from the market; and always maintain human supervision. Dynamic pricing, in short, gives its best when it is trained on the specifics only the entrepreneur knows, the seasonality of their own products, the dynamics of their own local market, and not entrusted blindly to a generic algorithm.

Beyond the three levers: the uses that help omnichannel

The three main levers do not exhaust the picture, and for an SME that lives between the shop and the web there are other uses of AI that deserve a mention, because they touch precisely the point of contact between the two worlds.

The first is demand forecasting and inventory management. An omnichannel company has a classic problem: the same product serves both the physical shelf and the website, and getting stock wrong means either tying up money or losing sales. AI, by learning from the purchase history across both channels, can estimate the demand for individual products and reduce both stock-outs and overstocking. The value emerges when the model learns from your own data and your own sales cycles, not from demand in the abstract.

The second is intelligent search on the site. The search bar is one of the most critical points of any e-commerce business: if the customer does not find what they are looking for, they leave. An AI-enhanced search understands the meaning of what the user types, tolerates typos and synonyms, and appreciably reduces the abandonment rate while improving product discovery. For a catalogue of even medium size, it is one of the improvements with the most favourable effort-to-result ratio.

The third is conversational assistance. Here, though, a caveat applies: a generic chatbot that parrots standard FAQs adds little, because similar solutions are now everywhere. The value is there when the assistant knows the catalogue and the customer's history and can genuinely guide them, steering those who do not know what to choose towards the right product, exactly as a competent shop assistant would in-store.

Where to start, in order

Lined up, the three levers suggest a priority on their own. For an omnichannel SME taking its first steps, the most sensible order starts from product recommendations, which offer the fastest and most measurable return and carry minimal risks. Next comes experience personalisation, which builds value over time by drawing on the unique knowledge of one's own customers. Dynamic pricing comes last, and only when you have solid data and a human oversight capable of supervising it.

There is then a principle that runs through all three and is worth keeping in mind. An SME's competitive advantage does not come from adopting the same tools as the giants, which are now within anyone's reach, but from using them on its own data and its own knowledge of the market. A generic algorithm makes a shop more like its competitors; an algorithm fed with the real relationship you have with your own customers sets it apart. It is the difference between chasing and standing out.

One final practical point concerns accessibility. None of these technologies any longer requires large-enterprise investments: there are tools, plugins and integrations within the reach of an ordinary-sized e-commerce business, often built on the main platforms already in use. The real obstacle is no longer the cost of the software, but having tidy data and a clear strategy on what you want to achieve.

To sum up: AI in e-commerce is no longer a luxury for giants, but for an omnichannel SME the value lies in choosing the right levers in the right order, starting from recommendations and personalisation and treating dynamic pricing with the caution it requires. The biggest risk is not adopting AI, but adopting it without a method, replicating mechanics already seen instead of making the most of what makes your own business unique.

At A126 we help SMEs that sell both online and in-store to put exactly this into practice: we start from the company's real data and objectives, identify the AI levers that deliver concrete results and integrate them with the platform and systems already in use, without chasing the hype and without upending what already works. If you want to understand which of these opportunities really makes sense for your shop, contact us for a free consultation: together we will analyse your e-commerce business and define the right priorities for your scale.

A126 Corporate Advisors — Artificial intelligence at the service of your e-commerce, without chasing the giants.