- Should you really worry about your brand when using AI?
- What happens when only IT manages AI governance?
- How widespread is the problem in Italy and worldwide?
- What changes in practice: who does what in the new model?
- How does an AI brand kit work in practice? Three Italian scenarios
- Which tools let you set up an AI brand kit?
- What the AI Act says: are there legal obligations to meet today?
- Why it's in the news now: is the Italian market really exposed?
- How to get started: concrete steps to build marketing AI governance
- What mistakes are most common when building AI governance for marketing?
- How do I know if my AI governance system is working?
- So, is it worth building marketing AI governance in your company?
- How we see it at SHM
- Frequently Asked Questions
- AI governance from IT and Legal doesn't cover the brand: you need a specific layer for marketing, tone of voice and visual identity.
- 84% of large Italian companies already have generative AI licenses, but only 9% have structured governance (Osservatorio AI Polimi, 2025).
- The solution isn't to block the tools, but to build 'safe enclosures': configured brand kits, approved prompts and human review workflows before publication.
- From August 2, 2026, the EU AI Act requires labeling of AI content on matters of public interest, unless there is effective human review with editorial responsibility.
A CEO posts an AI-generated image without going through the creative team. The result is an off-brand post with obvious visual artifacts and a tone that contradicts years of communication. The reputational crisis hits within hours. It's not an isolated case: according to IAB, over 70% of marketers have already experienced a similar incident.
Should you really worry about your brand when using AI?
Yes, and sooner than you think. The risk doesn't come from an external attack, but from a colleague who publishes AI content without going through marketing.
Anyone in the company can open a generative AI tool and produce text, images or video in seconds. The speed is real. The problem is that the quality of the output depends on who is prompting the tool, and it's almost never the brand manager.
The damage isn't just aesthetic. Off-tone or visually wrong content tells customers the company doesn't control its own voice. And that perception is hard to fix.
| Your situation | Brand risk | why |
|---|---|---|
| No AI policy in the company | High | Anyone publishes anything |
| IT/Legal policy but no marketing rules | High | The brand isn't covered by the security rules |
| Brand kit configured in AI tools | Medium | Reduces errors, but human review is still needed |
| Approval workflow before publication | Low | Marketing sees the content before it goes out |
| Brand kit + human review + differentiated policies (internal/external) | Very low | A complete system, compliant with the AI Act |
What happens when only IT manages AI governance?
IT protects data and security, not tone of voice or the exact color of the logo. These are two different problems that need different solutions.
IT policies focus on three things: keeping company data out of public models, approving software and respecting privacy. These rules are essential, but they don't answer questions like: does this post sound like us? Does this image use our palette?
Legal, for its part, checks copyright and regulatory compliance. It doesn't assess whether an article's tone is passive-aggressive rather than authoritative, as described in the case reported by MarTech.
The result is a gap: the AI tool passes the security review, but no one has decided how to use it to create brand content. The first colleague in a hurry fills that gap.
Having a formal AI governance lead doesn't mean that person handles brand. In most cases, the role sits in IT or Legal. Marketing stays out of the room where the rules are decided.
How widespread is the problem in Italy and worldwide?
A lot more than companies admit. Italian and European data show a huge gap between the adoption of AI tools and control structures.
The AI Observatory of Politecnico di Milano found in 2025 that only 9% of large Italian companies have AI governance with defined responsibilities and alignment with business goals. Meanwhile, 84% have already purchased at least one generative AI license.
It means most companies are using powerful tools without a control system. It's like handing out warehouse keys to every employee with no log of who goes in and what they take.
On consumer trust, Agenda Digitale reports a Meltwater/YouGov study showing that 32% of global consumers would trust a brand less if it uses AI-generated content, compared with only 15% who say they would trust it more.

What changes in practice: who does what in the new model?
IT and Legal remain responsible for security and compliance rules. Marketing needs to add a specific layer for brand, tone and content review.
- 1 · IT / LegalTool approvalThey assess the security, privacy, copyright and regulatory compliance of AI tools. They decide which ones are allowed in the company.
- 2 · MarketingBrand kit setupThey program the visual and verbal guidelines into the approved tools: color palette, fonts, tone of voice, approved vocabulary, validated prompts.
- 3 · MarketingInternal/external separationThey set different rules for internal content (pitch decks, updates) and external content (campaigns, social, press releases). The latter require human review.
- 4 · Cross-functional teamsWorking inside the safe enclosureSales, product and HR use AI tools within the templates and prompts approved by marketing. They can move fast without risk.
- 5 · MarketingReview and approvalBefore external publication, a marketing team member checks that the content respects the brand and takes editorial responsibility (required by the AI Act).
The key is separating the internal sandbox from external production. A colleague preparing an internal presentation needs freedom and speed. Whoever publishes on company channels needs an approval process. Applying the same rule to both leads to one of two worst outcomes: either everything gets blocked, or everything gets through.
How does an AI brand kit work in practice? Three Italian scenarios
An AI brand kit is a set of instructions preloaded into the tool: palette, tone, approved prompts. Whoever uses it can't step outside the boundaries set by marketing.
Many enterprise AI platforms let you configure custom brand kits: you upload the guidelines, define approved prompts and train custom models on your past communications. The result is that a salesperson drafting an email gets something that already sounds like the company, not generic text.
| Company type | Risk without an AI brand kit | Benefit with an AI brand kit | Human review needed? |
|---|---|---|---|
| Fashion e-commerce (20 people) | Social posts with the wrong tone, images outside the palette | Drafts already aligned with the brand, fewer rounds of revisions | Yes |
| Professional firm (consulting, 8 people) | LinkedIn articles with a tone that's too informal or generic | Content consistent with your positioning, ready to polish | Yes |
| Manufacturing company (150 people) | Product sheets generated by different departments with incompatible styles | Unified format, correct technical terminology | Yes |
| Retail chain (50 stores) | Local promotions with contradictory messages across locations | Approved templates, freedom only on local details | Yes |
Which tools let you set up an AI brand kit?
Several enterprise tools support brand kits and custom models. The choice depends on what you already use and how much you want to build control into your workflow.
| Tool | Native brand kit | Trainable custom model | Workflow integration | Suitable for |
|---|---|---|---|---|
| Adobe Firefly / Express | Yes | Yes | Medium-high | Creative teams, companies on the Adobe stack |
| Canva AI (Magic Studio) | Yes | Partial | High | SMEs, non-technical teams |
| Microsoft Copilot (365) | Partial | Yes | High (Office) | Companies already on Microsoft 365 |
| ChatGPT Enterprise / custom GPTs | Partial | Yes | Medium | Marketing and sales teams |
| Jasper AI | Yes | Yes | High | Content marketing teams |
No tool removes the need for human review before external publication. The brand kit reduces errors and speeds up work, but final editorial responsibility stays with a person. This isn't just good practice: from August 2, 2026, it's also a legal requirement.
What the AI Act says: are there legal obligations you must meet today?
Yes. From August 2, 2026, Article 50 of the EU AI Act requires AI content on matters of public interest to be labeled, unless there is effective human review with editorial responsibility.
EU Regulation 2024/1689 (AI Act), in force from August 2026, introduces transparency obligations for anyone using generative AI in their professional activity. In short: if you publish AI-generated content on matters of public interest (politics, health, economy), you must disclose it. The only exception is when the content has undergone real human review, with a person taking editorial responsibility for it.
For companies doing content marketing, this means the approval workflow isn't just good brand management practice: it's a legal requirement. Publishing AI content without review risks fines and, above all, a credibility crisis far harder to handle than an off-palette post.
AI governance overlaps with other risks we've already covered in the AI and process automation cluster: from managing AI agent permissions to the growing autonomy of models, as in the case of GPT-5.6 Sol deleting files on its own.
| Content type | AI Act obligation | what to do |
|---|---|---|
| Text on topics of public interest (health, economy, politics) | Labeling required if AI-generated | Human review with editorial responsibility, or an "AI-generated" label |
| Synthetic images and video (deepfakes) | Labeling required | Always disclose the AI origin |
| Internal content (reports, presentations) | No specific obligation | Good practice: note the use of AI in the metadata |
| Social posts, newsletters, blog articles | Required if the topic is of significant public interest | Human review recommended in every case |

Why it's in the news now: is the Italian market really exposed?
Yes. Adoption of AI tools has exploded in a year, but control systems have stood still. The gap between use and governance has never been this wide.
Last year, the AI Observatory of Politecnico di Milano recorded a jump from 53% to 84% of large Italian companies with generative AI licenses. That's a 31-point increase in a single year. Over the same period, the share of companies with structured governance stayed at 9%.
The result is a market where almost everyone uses the tools, but almost no one has set clear rules for using them in external communication. According to IAB, fewer than 35% of marketers plan to increase investment in AI governance over the next 12 months, even though over 70% have already experienced an incident.
On LinkedIn, 41% of long-form content appears to be written entirely by AI, a figure that fuels a saturation felt by readers and platforms alike. The audience reaction is already underway: the problem isn't in the future, it's here now.
How to get started: concrete steps to build marketing AI governance
You start by separating internal content from external content, then build the brand kit in approved tools, and finally define who approves what before publication.
- 1 · Marketing + ITMap the tools in useTake an inventory of all the AI tools used in the company, including those not formally approved (so-called "shadow tools"). Find out who uses them and for what.
- 2 · MarketingSeparate internal from externalDefine two levels: internal content (presentations, drafts, reports) with more freedom; external content (social, website, email, press releases) with a mandatory approval process.
- 3 · MarketingBuild the brand kit into your toolsUpload visual and verbal guidelines into the approved AI tools: palette, fonts, tone of voice, vocabulary, examples of correct and incorrect content.
- 4 · MarketingCreate approved prompts and templatesPrepare a set of validated prompts for the most common use cases (social posts, emails, product pages). Anyone who uses them gets results that are already on-brand.
- 5 · MarketingDefine the review processDecide who approves external content before publication, and how quickly. Document the review for AI Act compliance.
- 6 · Marketing + IT + LegalTrain your teams and update your policiesCommunicate the new rules to every team that uses AI. Update company policies to include marketing-specific responsibilities.
What mistakes are most common when building AI governance for marketing?
The most common mistake is treating AI governance as an IT-only problem, or building rules so rigid they make the tools useless.
We've seen similar dynamics in other areas of AI governance. As explained in our article on AI agents and access to company data, the problem is never the tool itself, but the lack of clear rules on who can do what. The same goes for brand content.
| Common mistake | Consequence | How to avoid it |
|---|---|---|
| Delegating everything to IT without involving marketing | Security rules without brand protection | Marketing at the governance table from day one |
| A single policy for internal and external use | Either everything is blocked or everything gets through | Two distinct levels with different rules |
| Outdated brand kit | AI content that uses the old logo or an outdated tone | Review the brand kit every time the guidelines are updated |
| No documented approval process | Legal risk (AI Act) and no traceability | Review logs, even a simple one |
| Blocking unapproved tools without offering alternatives | Teams secretly use unsafe tools | Offer approved tools that are easy to use |
| Training only the marketing team | Other departments keep using AI with no rules | Cross-team training for everyone, even if brief |
How do I know if my AI governance system is working?
There are concrete indicators to monitor: frequency of brand incidents, content review time, and the share of teams that have been trained.
| Indicator | How to measure it | Positive signal |
|---|---|---|
| AI-related brand incidents | Monthly count of off-brand or withdrawn published content | Zero, or steadily declining |
| Average review time for external content | Hours between AI draft and final approval | Under 24 hours for standard content |
| % of external content with documented review | Approval log / total publications | 100% for official channels |
| Adoption of approved tools | % of teams using tools set up with the brand kit vs. shadow tools | Over 80% of active users |
| Satisfaction of non-marketing teams | Internal survey every six months | Teams find the templates useful, not limiting |
| AI Act compliance | Quarterly check with Legal | No unlabeled publications on significant public topics |
A governance system that works is invisible: content goes out aligned, teams move fast, and marketing doesn't spend its time fixing other people's mistakes. If instead marketing becomes a bottleneck or teams work around the rules, the system needs rethinking, not tightening.
It's also worth keeping an eye on how the capabilities of AI agents evolve. As analyzed in our article on OpenAI and the cyber risks of AI governance, today's rules may not be enough for tomorrow's tools.
So, is it worth building marketing AI governance in your company?
It's worth it if you already use generative AI tools across several departments, you've had at least one off-brand piece of content in the last six months, or you regularly publish on external channels. The reputational and regulatory risk is real, and a control system costs far less than a crisis.
Somewhere in between if you're in an early adoption phase with a few tools and a small team. You can start with simple rules (two levels, internal and external, plus a set of approved prompts) and add structure as usage grows.
Not yet if you use AI only for internal work with no external publishing, and you have a marketing team that reviews everything before it goes live. In that case the risk is already managed, but keep an eye on regulatory developments.
How we see it at SHM
For us, AI is a tool that should give people their time back, not replace human judgment on the content a company puts out into the world. A well-configured brand kit and a clear review process don't limit creativity: they free it, because people already know where they can go and where they can't.
We worked on a similar case with a national association that needed to manage content across very different channels, with teams that weren't communication specialists. The solution wasn't to block the tools, but to build a system (architecture, an AI assistant trained on the guidelines, an approval workflow) that lets anyone produce consistent content without going through the central team every time.
If you're evaluating how to structure AI governance for your company's marketing, we can help you start in the right place: from the strategic integration of AI into existing processes, to the development of virtual assistants trained on your guidelines. If you also want to understand how to train your internal teams, our AI training for companies program is the most practical starting point. And if AI agents and access permissions worry you, our article on Prime Intellect and enterprise AI agents offers a useful picture of where the market is heading.
Frequently Asked Questions
Yes. IT policy covers security, privacy and approved software, but it doesn't define how to use those tools to create brand content. You need an additional layer: brand kits configured in the tools, approved prompts and a review process before anything is published externally.
There are two types of risk. The first is reputational: off-tone or visually wrong content that customers notice right away. The second is regulatory: from August 2026, the EU AI Act requires labeling of AI content on matters of public interest, unless there is documented human review with editorial responsibility.
For a company with guidelines already in place, setting up a brand kit in the main tools takes one to three weeks of work. The longest part is validating prompts and training teams. If the guidelines don't exist yet, that's where we start.
Free versions don't let you configure brand kits or custom models, and they often lack the privacy safeguards needed for company data. For professional use with brand content, enterprise or business versions are the minimum standard. Always check with IT before you choose.
The IAB data is useful: over 70% of marketers have already experienced an AI incident, but fewer than 35% plan to invest in governance. Framing the risk as the cost of a reputational crisis (PR management, lost customers, regulatory fines) makes the case more concrete than any statistic.
It applies to any company that publishes content on external channels using AI tools. For SMEs, the system can be very simple: two levels (internal/external), a set of approved prompts and one person who signs off before publication. You don't need a complex structure to reduce risk significantly.
The real question isn't whether to use AI for content, but who in the company takes responsibility for what goes out.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.