Generative AI for SME Marketing: Why Going Beyond ChatGPT Changes the Results

Everyone uses it, few use it well

Over the past two years, generative artificial intelligence has entered the offices of Italian SMEs at a speed few had predicted. According to ISTAT data published in December 2025, the share of businesses with at least ten employees using at least one artificial intelligence technology doubled in a single year, rising from 8.2% to 16.4%. And the area in which this technology is adopted most frequently is precisely marketing and sales, which alone accounts for 33.1% of use cases, ahead of administrative process organisation and research and development.

Yet behind these numbers hides a paradox that anyone working with small businesses knows all too well. Most SMEs that claim to "use AI in marketing" actually do just one thing: they open ChatGPT, type "write me an Instagram post about my product" and copy the result. It is a legitimate use, but it is also the most superficial level possible, the one that produces generic content, recognisable at a glance and devoid of the brand's voice. The problem is not the tool. It is the method, or rather its absence.

This article starts from a simple conviction: generative AI delivers real results in an SME's marketing when it stops being an oracle to consult for off-the-cuff answers and becomes a component within a structured process. Going "beyond ChatGPT" does not mean abandoning it, but understanding that it is only one piece of a broader system made up of text, images, video, analysis and content governance.

The landscape: broad adoption, low maturity

The data tells a precise story. On the one hand, adoption is growing and concentrated on marketing; on the other, the depth of this use remains minimal. Throughout 2025, the Politecnico di Milano's Artificial Intelligence Observatory observed a dynamic in which large companies accelerate while SMEs fall behind, held back not by a lack of interest but by a lack of skills and a method for governing the technology.

The size gap is stark. Again according to ISTAT, in 2025 53.1% of large Italian companies use AI, compared with 15.7% of SMEs. The divide, as several analysts have noted, is not geographical but dimensional: the gap between small and large businesses weighs more than the difference between countries. And the recurring reason is not "AI isn't useful", but "we don't know how to govern it". SMEs adopt AI as a portfolio of isolated micro-initiatives, often launched by individual employees without a clear mandate and without training, rather than as a project integrated into their processes.

Meanwhile, the technology has ceased to be purely text-based. Video production with artificial intelligence rose from 18% to 41% in a single year among the companies that use it, a sign that the centre of gravity is shifting towards multimodal content: images, video, audio. This vastly expands the possibilities for marketers, but it also raises the bar: those who continue to use AI only to generate text are exploiting a tiny fraction of its potential, while more organised competitors build entire pipelines of visual content.

The result is a landscape with two types of SMEs. Those that use generative AI occasionally, obtaining marginal time savings and mediocre content. And those, still few, that have embedded it into a process, obtaining brand consistency, higher content volumes and a genuine reduction in production costs. The difference between the two groups lies not in the tools, which are accessible to everyone, but in how they are orchestrated.

Beyond text: what it really means to use generative AI in marketing

To understand what going beyond ChatGPT means, it helps to dismantle the idea that generative AI is a single thing. It is instead a set of tool families, each with different strengths and limitations, that perform at their best when they work together within a coherent workflow.

Text generation is the starting point, not the finish line

Language models such as ChatGPT, Claude or Gemini remain the heart of text production: social media posts, product descriptions, sales emails, blog article drafts, video scripts. The leap in quality lies not in the model you use, but in how you query it. A generic prompt produces generic output. A prompt that includes the brand's tone of voice, examples of past content that worked, the target audience and the desired structure produces something much closer to publishable.

The most effective practice for an SME is to build reusable prompt templates: saved frameworks for each type of content that encode the company's voice once and for all, and that anyone on the team can reuse. It is the difference between starting from scratch every time and having an editorial line built into the tool. One point remains fixed, however: final editorial control is human. AI speeds up the drafting, it does not replace the judgement of what to publish.

Images: from "nice" to on-brand

This is where many SMEs have not yet arrived. Image generators such as those available today make it possible to produce visuals for social media, banners, catalogues and campaigns without a design department. But the real innovation of 2025-2026 is not the ability to generate a beautiful image: it is directional control. The most mature tools now allow you to maintain visual consistency across multiple assets in the same campaign, to define lighting, composition and perspective with photographic precision, and to integrate brand elements directly into the generation.

For an SME this means being able to produce dozens of mutually consistent visuals while maintaining a recognisable identity, something that until recently required a photographer or a designer for every single piece of content. Beware of two pitfalls, however. The first is consistency: without a system of visual references and structured prompts, images end up disconnected from one another and weaken the brand instead of strengthening it. The second is licensing: not all tools guarantee safe commercial use, and for a company, using an image without clear rights is a real risk.

Video: the fastest-growing front

The shift in AI video content from 18% to 41% usage in a single year shows where the market is heading. Video generation tools now make it possible to create short promotional clips, product animations and social media content starting from text or static images. For an SME that has never had a budget for video production, it is a door that opens. Here too, however, value only emerges within an editorial plan: an isolated, out-of-context video is worth little, whereas a series of videos consistent with the rest of the communication is worth a great deal.

Analysis: the AI that doesn't generate, but understands

Finally, there is a use of generative AI that has little to do with content creation and much to do with strategy: analysis. Language models and dedicated tools can synthesise customer reviews, analyse competitors' communication, summarise social media comments and identify recurring themes in support requests. It is a less visible but often more valuable use, because it guides decisions instead of merely producing output. An SME that uses AI to understand what its customers are saying and what its competitors are doing gains a more durable competitive advantage than one that uses it only to write faster.

The method: from idea to published content

The underlying thesis is that generative AI works when it is embedded in a process, not when it is used as an off-the-cuff oracle. It is worth describing what this process looks like in practice within an SME, regardless of the specific tools chosen.

The first step is to define the brand's voice and rules once, in a reference document. Tone, words to use and to avoid, values, examples of successful content: all of this must be written down and then built into the prompts. Without this foundation, every piece of generated content will start from scratch and inconsistency will be inevitable.

The second step is to build a workflow for each type of content. For an e-commerce store's product descriptions, for example, the correct workflow gathers the technical specifications and differentiating features, feeds them into a structured prompt template, generates the drafts, and includes a final human review before publication. The difference from improvised use is that the process is repeatable, documented and reliable.

The third step is human review, which is not a detail but the point where quality is decided. AI produces drafts, people decide. This is especially true for technical content, where an error in a description can generate returns and complaints, and for content that touches on sensitive topics or the company's positioning.

The fourth step, increasingly important, is compliance. From 2 December 2026, European legislation introduces the obligation to label AI-generated content, so-called watermarking. An SME that produces dozens of pieces of content a month with generative tools will have to track and declare what has been produced automatically. Building this aspect into the process from the outset, instead of chasing it afterwards, avoids having to redo the work. Added to this is caution over copyright: keeping the prompts, documenting the creative process and preferring tools with explicit commercial licences is a good practice that protects the company.

What changes in the numbers

When generative AI is embedded in a structured process, the benefits become measurable. Companies that have built mature workflows report significant reductions in content production times and creative costs, with the ability to produce more variants for more channels in the same time it previously took for a single piece of content. These are not magic numbers: they depend on the quality of the method, not the power of the tool.

It is worth being honest about the starting point, however. For an Italian SME, the first return of generative AI in marketing is almost never an immediate increase in sales, but the recovery of time: hours reclaimed from the manual drafting of repetitive content that can be reinvested in strategy, in customer relationships, in analysing results. It is a real return, but an indirect one, and it must be measured for what it is. Promising rising revenues thanks to an image generator would be dishonest. Promising faster processes and more consistent content, with a solid method behind them, is realistic.

The competitive advantage is the method, not the tool

To sum up: generative AI is now widespread in the marketing of Italian SMEs, but almost always at a superficial level that produces generic content and marginal savings. The leap in value does not come from changing the tool, but from building a process in which text, images, video and analysis work together, within clear brand rules, with human review and attention to regulatory compliance.

At A126 we support SMEs precisely at this transition: we do not sell yet another tool, we build the method. We map existing communication processes, we define AI workflows tailored to the company's voice and objectives, and we integrate generative tools with the website, social media and publishing systems, including tracking and compliance aspects from the very start. The goal is for the technology to adapt to the way you work, not the other way around.

If you want to understand how generative AI can concretely fit into your company's marketing, beyond simply "asking ChatGPT", get in touch for a free consultation: together we will analyse your communication processes and define a path scaled to your size and your objectives.

A126 Corporate Advisors — Tailored digital solutions for the marketing of Italian SMEs.