Accenture Song’s Mark Curtis: Generative AI Needs Human Creative Direction

Generative AI can make customer interactions more personal and help creative teams move faster. But if brands rely on it to produce ideas without human direction, they risk turning out work that looks much like everyone else’s. That is the central tension in Mark Curtis’s advice: use the technology to extend creative work while keeping people responsible for the choices that give a brand its character.

Curtis, Accenture Song’s global sustainability and thought leadership lead, discussed these themes in connection with the Accenture Life Trends 2024 report. His experience spans media, digital, mobile and design. He co-founded Fjord, which Accenture acquired in May 2013, and previously led work on the metaverse at Accenture Song.

Personalization depends on better interactions

One shift Curtis sees is that people are moving from conventional search toward conversational ways of finding information. That change could give brands new ways to understand what customers are asking for and tailor the responses, products or services they provide.

The opportunity reaches across customer service, marketing and design. Natural language interfaces can make it easier for people to explain what they need, while generative systems can help businesses adapt content and interactions to those needs. Curtis argues that improving these touchpoints matters as customer loyalty weakens and companies look to customer experience as a source of durable growth.

Personalization, though, is not valuable simply because it is technically possible. It needs to solve a customer’s problem or make an interaction more relevant. A system that generates fluent answers but leaves people struggling to get help would add another layer of technology without addressing the underlying experience.

The report’s figures point to that gap. Curtis says 47% of consumers feel undervalued because of poor customer service, and a third of respondents said it takes more effort to get help or reach customer service. Generative AI could streamline some communications and track customer sentiment, but businesses still need to judge whether these changes make service easier for people.

Creative speed can come at the cost of distinction

Generative systems can produce concepts on demand, giving creative teams more material to work with. Curtis’s caution is that these systems tend to draw toward familiar patterns. Without time, budget and skilled creative direction, the output may not stand apart or reflect a brand’s identity.

He describes generative AI as an “averaging engine”: it can return plausible options, but those options may converge on what is already common. In a crowded market, a brand that accepts the first generated concept may end up with work that feels interchangeable. Curtis warns that replacing craft with automated output could raise the baseline for mediocre work while making distinctive ideas harder to find.

His proposed approach is to treat generated concepts as starting points. Creative professionals can select, reshape and develop them into finished work, while technologists contribute as co-creators. That pairing may help teams produce high-quality work more efficiently without handing over the judgment that determines whether an idea fits the brief or resonates with its intended audience.

For that process to work, companies need a clear sense of their brand identity and culture. They also need briefs that explain how the brand can adapt across different interactions and which creative risks are acceptable. Curtis offers three practical guideposts: protect the quality of the work and its value to the audience, keep the brief in view as the tool presents more options, and understand how platforms and discovery tools can push different teams toward similar results.

Trust requires attention to data and model behavior

Curtis identifies several areas businesses should consider when putting AI into use: data security, rising carbon emissions, copyright and model hallucinations. The interview does not prescribe a technical control for each risk, but its broader point is that responsible use needs to be built into how systems are created, advanced and deployed.

A privacy-first approach should keep sensitive information from being used unethically. Curtis also calls for transparency about AI use and for algorithms to be developed with data that is as impartial as possible, reducing the chance of unintended bias. These measures matter to customer trust because personalization depends on information about people and can shape what they see or receive.

Employees are another important checkpoint. Curtis says they should be able to raise concerns at any stage, so problems can be identified and addressed promptly. He cites research finding that 35% of consumers believe everyone shares responsibility for discussing healthy and responsible technology use. That finding places the conversation beyond technical teams: people affected by these systems also expect businesses to consider their impact.

Workplace gains come with disruption

For employees, Curtis frames generative AI as a way to augment human talent and automate tedious tasks. He cites Accenture’s Pulse of Change report, in which close to 9 in 10 organizations said generative AI contributes positively to the employee experience, while 68% said it improves employee productivity.

Those figures describe organizations’ reported views; they do not establish that every deployment benefits every worker. Curtis also says AI is likely to replace extensive parts of current white-collar jobs, a shift companies should expect to draw debate. The practical challenge is to use saved time to support strategic thinking and difficult problem-solving, while recognizing that automation can change the work people do and the roles businesses need.

Simpler service, despite more technology

More than 25% of people in the Life Trends findings feel technology has made life more complex. Curtis says companies should be deliberate about adding tools and train employees to use them effectively. Generative AI can analyze information across platforms, but that capability only helps when staff can apply it to a clear business or customer need.

People’s relationship with technology can also be mixed: 41% of those who use it often say it both complicates and simplifies their lives. That tension makes real-time personalization potentially useful, since brands can adjust content or products as preferences change. Serving individuals directly could offer more precise experiences than relying only on broad customer groups.

Yet the aim should be to reduce friction. Conversational and multimodal systems may support more tailored interactions, but companies need to check whether customers can get what they need with less effort. Curtis’s advice points to a practical test for adoption: AI should make the experience more useful, keep human creative judgment in the process and give employees room to raise concerns. That is where speed and personalization can translate into value without making every brand sound alike.