The marketer’s AI paradox: Speed, scale and the struggle for control

Food for Thought: The AI paradox and the cost of bad data

What are the challenges and opportunities presented to marketers by AI tools? In this HotTopics Food for Thought, in partnership with Domo as part of the ‘AI to Impact’ community, senior marketing leaders discussed the rapid evolution of AI tools and the need for a solid data foundation.

‘AI to Impact’ community, senior marketing leaders discussed the rapid evolution of AI tools and the need for a solid data foundation.

The enterprise conversation around AI has matured dramatically over the last year. The debate is no longer centred on whether organisations should experiment with AI, but rather how they can integrate it in a way that delivers meaningful value without creating operational chaos.

What emerged from this discussion was not resistance to AI itself, but the opposite. Participants around the table were actively experimenting with AI in marketing and sales, and for everything from analytics and content creation to campaign reporting. Yet despite this enthusiasm, there was still a sense of tension in the air: organisations are moving faster than their people, governance structures or data foundations can support.

The AI paradox: Meet the speakers

With Doug Drinkwater moderating the HotTopics Food for Thought debate, the speakers included:

• Christie Dahmen, VP Marketing, Kore.ai
• Jo Bance, Deputy CMO, Expleo
• Leanne Chescoe, Marketing Director, Demandbase
• Malcom McLaren, Marketing Leader, Experience Everything
• Marenza Altieri-Douglas, Group VP Strategic AI, Mplus
• Merinda Hillier, EMEA VP Marketing, Tealium
• Nikki Wells, Absolute Security, Senior Director, International Marketing
• Shay Assaraf, CMO, Totogi
• Spencer Wilcox, Founder, Everybody Eats
• Jamie Morrison, Field CTO, Domo

AI as a market differentiator

Again and again, the discussion returned to a recurring contradiction.

Attendees argued that AI is creating efficiencies while simultaneously exposing weaknesses that many organisations had either ignored or not yet encountered.

They also found that teams are discovering more content, analysing more data and automating more workflows than ever before, but they are also struggling with fragmented systems, inconsistent outputs, governance gaps and also growing uncertainty over accountability.

One participant summed up the pace of change bluntly, explaining that just as one AI tool becomes embedded and approved internally, “there's something else, it doesn't work anymore or it just moves on… The speed seems to be getting faster and faster.”

That sense of instability framed much of the discussion. AI is evolving so rapidly that organisations are finding it difficult to build sustainable processes around it before the technology itself changes.

Transition versus experimentation

Jamie Morrison, Field CTO at Domo, described the challenge as one of transition rather than experimentation. Morrison’s argument is that organisations are not struggling to trial AI, but grappling with how to operationalise it.

He explained, “Certainly, what we’re seeing a lot of organisations doing is running a lot of pilots and POCs - a lot of testing things. . And there’s a challenge, or some complexity, in how you move from a pilot and get something into production and get it to be used by the business.”

That distinction between experimentation and adoption became one of the defining themes of the conversation.

The reasons for this gap are complex. Morrison pointed to three recurring conditions that determine whether AI initiatives move more successfully from proof-of-concept into operational use. The first is data foundation.

“You need that foundation in place to get good results and to build good processes and workflows on top of,” he explained, noting that marketing environments are particularly difficult because of the sheer number of disconnected tools and platforms involved.

The second is executive buy-in. They argued that AI initiatives often emerge organically from teams at ground-level, but scaling them requires both leadership alignment and strategic commitment.

The third challenge is involving users from the beginning rather than introducing AI projects to teams after they had been developed. “Bringing everyone in from day one and co-building any PoCs and initiatives,” Morrison argued, leads to significantly higher adoption.

AI paradox: Adoption and workflow evolution

Morrison’s final point strongly resonated with the rest of the participants around the table because several of them had experienced the exact opposite.

One marketing leader described internal friction caused by an AI lead who approached adoption from a purely operational perspective, rolling out changes with little to no collaboration.

“Here’s the one hour enablement recording. Do it. Why aren’t you doing it?” they explained, describing a top-down implementation style that generated more resistance rather than engagement.

What also became clear throughout this portion of the HotTopics Food for Thought discussion was that AI adoption is not just a technical rollout, rather, it is a sizable change management exercise. This became more evident when the participants discussed content generation and automation.

Several organisations had attempted ambitious AI-driven content workflows that initially appeared highly promising.

One marketing leader described building an AI agent capable of taking a single white paper and automatically transforming it into blogs, LinkedIn posts and other supporting assets. However, the project ultimately failed. The reason? The quality of outputs did not meet expectations particularly for technical subject matter experts who detected the use of weak or generic language.

More importantly, the use case itself evolved faster than the implementation.

The team initially believed that they wanted large volume of content generated simultaneously, but eventually realised that they preferred staggered outputs over time with fresh updates added in. As one participant reflected: “The use cases have evolved so fast… if you develop something, it’s really out of date.”

Speed and quality outputs

The discussion repeatedly returned to the tension between speed and quality. The participants argued that AI allows organisations to produce content and outputs at scale, but they questioned whether this acceleration is quietly degrading standards.

One marketer observed that AI generated outputs may be 80 percent complete, but that final 20 percent matters enormously when brand credibility and expertise are involved.

Another participant warned that, “volume and speed means that quality is taking a hit,” describing organisations “smashing AI into everything” simply because they can.

This concern extended beyond internal workflows to broader questions around authenticity and brand reputation.

Several participants said that audiences are already becoming highly sensitive to AI-generated content. One leader referenced comments from another industry discussion: “I can spot a chatGPT, or AI generated post, a mile away.”

One of the most striking moments in the discussion came when a participant reflected on a conversation with their teenage daughter, who “takes real offence at AI generated content because she feels you as a brand didn’t value me enough.”

This shifted the conversation away from efficiency and towards trust. While organisations are focused on scale and productivity, younger audiences may increasingly associate AI-generated communication with a lack of effort, authenticity, or care.

Governance and the AI “wild west”

This increasing concern over quality and authenticity led naturally to a broader debate around governance.

Several participants described AI-usage inside their organisations as the “wild west,” with teams experimenting freely across multiple tools including ChatGPT, Google’s Gemini, Anthropic’s Claude and Copilot from Microsoft, often without oversight or coordination.

One participant admitted that while their organisation enforced rigorous compliance around contacts and data processes, “it didn’t, say, ask me once like are you putting customer data into ChatGPT?”. The imbalance between traditional compliance frameworks and emerging AI risks became increasingly apparent.

What also emerged was a growing recognition that governance can no longer be treated as purely an IT or legal issue. AI governance is becoming a business-wide operational necessity.

Leaders discussed the need for “guardrails”, “governance” and clearer processes for validation and approval. Yet there was also acknowledgement that many organisations are still far from establishing these structures.

Who owns data?

The role of data within this conversation was deeply revealing.

AI may be the visible layer attracting attention, but leaders around the table repeatedly argued that the real bottleneck here is data infrastructure. One attendee admitted candidly:

“I can’t even find who owns our data.” Another described how organisations are discovering that AI demands a level of data structure and accessibility that previously did not exist.

Historical fragmentation, duplicated systems and unclear ownership suddenly become major operational barriers once AI is introduced.

This is where Morrison’s emphasis on data foundations comes in. The participants argued AI systems are only as effective as the data they can access, and many organisations are discovering that their existing infrastructure was never designed for this level of inoperability.

Marketing teams in particular face challenges because valuable data exists across CRMs, analytics platforms, campaign systems, customer support tools and collaboration software, often with little consistency or governance.

AI success stories

Interestingly, while most of the discussion focused on risk and complexity, participants around the table also shared several positive use cases where AI had delivered measurable value.

One marketing leader described using AI to analyse quarterly business review data in minutes rather than spending days manually compiling reports.

Others discussed AI’s success in localisation and translation, significantly reducing reliance on more expensive external agencies while accelerating delivery across multiple markets. SDR and sales enablement workflows were another area where AI appeared to deliver value, particularly where teams faced resource gaps or needed rapid scalability.

On the other hand, even these success stories often returned to a deeper measurement problem. The leaders around the table questioned how organisations should quantify AI’s value. Time-savings alone are difficult to translate into board-level metrics.

If a task that previously required four hours now takes an hour, what happens to the remaining three? One leader joked: “If something takes four hours, it takes you one hour, go and play tennis.”

The future of the AI era

Perhaps the clearest conclusion from the discussion is that organisations are still in the earliest stages of understanding what effective AI adoption actually looks like.

There was broad agreement that the future does not belong to companies that simply deploy the most AI tools. Instead, success will depend on building the right balance between experimentation and control.

Morrison’s advice? Organisations should initially focus on smaller, high value use-cases that demonstrate practical outcomes and build confidence gradually.

“Apply 70 percent of your time on how you can use AI to realise small enhancements to an existing workflow to gain efficiencies, 20 percent on how you might reinvent an existing workflow to capture gains available from AI advancements and legacy learnings… and then 10 percent apply to more strategic initiatives or innovating to do something new and capture a new opportunity.”

That framework resonated because it recognised a critical reality: most organisations are still “flying the plane while building it.”

AI to Impact:

Join the AI to Impact hub.

The AI community for executives focused on outcomes, not hype, powered by HotTopics in partnership with Domo.