Connect London: The Real Work Behind AI Adoption

AI can help us create and deliver content faster. But how do we scale it without losing sight of quality, accountability and the people we're ultimately creating it for? That was the central question at the second edition of Connect 2026 in London.

On 22 October, around 40 senior leaders from some of the world's leading brands across media, entertainment, technology and other global industries gathered at The Lanesborough Hotel in London for the second edition of Connect 2026. Building on the first event in Dublin, they explored the practical challenges of scaling AI responsibly, moving beyond its capabilities to consider the governance, accountability and operational decisions needed to make it work.

From governance and accountability to workflows and human oversight, one message came through clearly: adopting AI is not just about choosing the right tools but rethinking how work gets done.

Opening the event, Florian Faes, Managing Director of Slator, set the scene for a discussion that quickly turned to the realities of AI adoption. Asked to describe their company's AI strategy in one word, audience members offered responses including fragmented, cautious, disjointed, panicky and unclear.

AI Governance and Accountability in Content Operations explored why organizations are struggling to turn AI ambition into consistent practice. Acolad's Head of Global Solutions, Petra Angeli, joined industry professionals to examine the challenges of establishing clear accountability, effective governance and meaningful human oversight.

The discussion highlighted a fundamental tension: large organizations need governance that is consistent enough to manage risk, yet flexible enough to accommodate different business units, markets and operational needs. A single, rigid approach rarely works for everyone, but fragmentation without shared safeguards creates its own risks.

• Accountability must be clear. When something goes wrong, responsibility can become blurred across business teams, legal, security and the people approving AI-generated content. Organizations need clear ownership and someone empowered to stop a process when necessary.

• Policies need to be part of everyday work. A framework that employees must interpret for every decision isn't doing enough. Governance needs to translate into practical guidance, clear responsibilities and workflows people can follow.

• Human oversight must be meaningful. Asking someone to approve an AI output is not enough if they lack the training, context or time to assess it properly.

The key takeaway is that AI governance cannot be separated from operational reality. It must account for how work is actually performed, who makes decisions and what is needed to ensure those decisions are sound.

In his presentation, Acolad's Chief &  AI Technology Officer, Stéphane Cinguino, challenged the idea that the debate around AI should be about humans versus machines. The more useful question is how AI changes the entire content system, from creation to delivery.

His banana ketchup analogy brought this to life. During the Second World War, tomato shortages in the Philippines led to ketchup being made from bananas. The intended outcome remained the same, but the ingredients and process changed.

Organizations face a similar opportunity with AI. Rather than simply adding it to existing processes and expecting efficiency gains, they can rethink how work gets done.

Cinguino also explored the concept of a “sandwich workflow”: humans provide the context and direction, AI helps produce the output, and humans review it for quality, accuracy and brand fit.

The principle is not to keep people involved for the sake of it. It is to design workflows in which AI and human expertise each contribute where they add the most value.

This requires combining two complementary forms of expertise. Domain specialists understand the business, its complexities and its risks, bringing the experience and judgement needed to assess whether an output is fit for purpose. AI specialists understand the technology's capabilities and limitations, and where it can create meaningful improvements.

Neither perspective is sufficient on its own. Overestimating AI can lead to unrealistic expectations and unnecessary risk; underestimating it can mean missing opportunities to fundamentally rethink how work is done. Progress depends on bringing these perspectives together and allowing each to challenge the other's assumptions.

The same principle applies to AI tools. Different teams have different needs, and no single platform will necessarily meet them all. The priority is to establish governance, visibility and accountability across the tools in use, without sacrificing the flexibility needed to do the job well.

Consent needs to be established before a performer's voice is cloned. Compliance requirements need to be considered before content is released. And teams need a way to use specialist tools without losing visibility or accountability.

The risk is not only regulatory. AI-generated content can also raise questions about authenticity, trust and brand identity. Audiences may welcome AI when it improves an experience or supports useful content, yet react negatively when its use feels inauthentic or disconnected from a brand's values.

This makes it important to distinguish between using AI to improve content and allowing AI-generated output to define the audience experience without sufficient consideration. The technology may make production easier, but it cannot determine on its own what a brand should sound like, how a message will be received or whether a particular creative choice is appropriate.

Global requirements matter from the outset, too. Local regulations, cultural expectations and language-specific risks are much harder to address when treated as final checks rather than built into the original plan.

For organizations working across markets and languages, this means combining shared safeguards with an understanding of local needs. It also means extending appropriate standards to external partners and vendors, not just internal teams.

Across the discussions, three practical principles emerged:

• Stop assuming AI will solve everything, working in silos, or measuring success only in speed and cost savings.

• Change how work is designed. Involve the right people early, establish guardrails, clarify responsibilities and build meaningful human review into everyday workflows.

• Decide based on value, business impact and accountability. Be clear about who owns the outcome, where human judgement is essential, and when AI isn't the right choice.

  These principles point to a broader shift in how organizations approach AI. Rather than treating it as a standalone technology initiative, they need to consider its implications across the entire content lifecycle, from strategy and creation to review, distribution and audience experience

  One idea connects all three: the purpose of content has not changed, even if the way it is created and delivered has. Content still needs to inform, entertain, protect and connect with people. AI can change the process, but it does not remove the need for trust, cultural understanding or responsible decisions.

The second Connect event reinforced that responsible AI adoption isn't a one-off policy exercise or a technology decision. It requires organizations to understand how work actually happens, give teams the time and tools to adapt, and keep governance aligned with a changing landscape.

The goal is to augment human expertise, helping people work more effectively while recognizing where human judgement remains essential and where AI shouldn't be used.

With further Connect events planned across Europe and the US, this conversation will continue. As AI reshapes how content is created and delivered, bringing industry leaders together to share experiences, challenge assumptions and explore practical solutions will be key to making AI work responsibly.