Reimagining the marketing organization in the age of generative AI

The speculation about ChatGPT and generative AI's impact on the marketing organization runs in two directions. One says the technology will replace the majority of marketing professionals, who will be left to walk dogs and fix air conditioners. The other says it will be an accelerant – better creative, stronger audience connections, differentiated talent. The truth, as it usually is, sits between the two: the technology introduces new capabilities to automate certain tasks while preserving others for the [human] marketing professional.

I reviewed the academic literature on AI and machine learning's potential to automate professional tasks, then compared and ranked those functions against the modern marketing org. A mere 7% of it is strongly resistant to automation.

The Functions Least Likely to Be Automated

Strategy and new brands. Category strategy and innovation. Brand management.

The more strategic, less brand-aligned functions scored highest in defensibility against automation. Strategy and new brands held the largest concentration of highly defensible tasks. Brand management and category strategy and innovation each held a smaller but significant share of highly defensible work.

The reason isn't mysterious. These roles make decisions about future investments, with risk and perception in play, and they generate new ideas – work that depends heavily on social and creative intelligence.

The Functions Most Likely to Be Automated

Insights. Media. Internal agency.

The roles that scored lower in defensibility were concentrated on day-to-day growth or management of the existing business, including receiving and maintaining brand positioning. Internal agency, insights, and media roles are the ones I'd expect to be less resilient as AI and machine learning are further deployed in the marketing process.

How I Got There

I started with a meta-analysis of the academic literature, which converged on six capabilities least susceptible to automation:

1. Perception. Tasks that require complex human perception – assessing a landscape or situation, identifying proximity and properties, and making decisions based on that perception – are too difficult for machines or algorithms to replicate.
2. Creative Decisions. Generative AI has proven valuable in the creative brainstorming process and in speeding up the execution of creative or messaging, but it can't replace actual creativity – "the ability to come up with ideas or artifacts that are novel and valuable."
3. Social Tasks. Tasks that require social intelligence, like negotiation and persuasion, require an iterative process of assessment of and reaction to human emotion. Some algorithms can replicate aspects of this process, like predicting likely responses, but they fail to reproduce the real-time analysis and reaction of a human.
4. Development of Positioning. Technology-enabled segmentation and targeting can certainly inform positioning, but strategic positioning that communicates a value or message that resonates at a personal, emotional level with consumers remains a uniquely human ability.
5. High-level Product Decisions. Research suggests that high-level product decisions based on assessments of consumer behavior and identity are too difficult to automate; low-level product tasks like brand logo design can be supported by AI.
6. Entrepreneurship. Creative entrepreneurship – the process of identifying, pursuing, and capturing new opportunities for business innovation, creation, and growth – requires more creative intelligence and dexterity than AI can currently provide.

To understand how these tasks distribute across an actual marketing org, I ran them against a large global beverage company. Using ChatGPT, I generated 73 unique job descriptions covering the company's 233 marketing positions.

I then classified each of those descriptions against the six difficult-to-automate functions, giving each a rating of 1 (low requirement), 2 (medium), or 3 (high).

Averaging the ratings produced a final score for each job, and a scale to evaluate the org against:

• Jobs scoring 2.5 or above: high defensibility against automation.
• Jobs scoring between 1.51 and 2.49: medium defensibility.
• Jobs scoring below 1.5: low defensibility.

Low and medium defensibility doesn't necessarily mean these jobs will be replaced by AI. It does mean that many of the functions currently associated with these roles could be improved or streamlined using AI.

The average score across the full marketing org was 1.67. The distribution:

• 16 jobs (7%) scored 2.5 or above.
• 99 jobs (43%) fell between 1.5 and 2.49.
• 118 jobs (50%) came in at 1.5 or below.

This research only scratches the surface of how the gradual evolution of technology and the marketing org will affect one another, but it does point at where we'll need to rethink, realign, and reskill:

• Enterprising marketing departments will invest in areas like media and insights, where AI offers the greatest potential for automation and ROI. They'll simultaneously look at how to take some automatable responsibilities off the plate of professionals in human-intensive roles, freeing up more of their time for creativity, strategy, and innovation.
• We need to research and consider how this breakdown of defensible roles affects different segments of the workforce – across tenure, education, gender, and race.
• We need to seriously weigh the costs of automation against the benefits. All of these jobs, regardless of their scores, are good jobs. Replacing employees with AI may benefit the bottom line, but it can do long-term damage to the workforce by eliminating the valuable early-career experience that prepares professionals to take on the more complex tasks that can't be automated later in their careers.

From Defending the Budget to Defending the Job

Forrester recently estimated that AI will replace 7.5% of agency jobs by 2030. I think the impact is going to be much more substantial.

There's more to the story than the raw distribution above. Underneath the unsettling proportions of replaceable marketing work sit some worrying indicators.

Start with gender. The 136 women in the organization tended to hold jobs with less defensibility – an average score of 1.63 – than the 73 men, who averaged 1.73. The defensibility gap is smaller than the salary gap in this organization, where men's average salary was $161k and women's was $131k. The conclusion is unflattering: women in this org aren't only underpaid, they're also disproportionately holding jobs with fewer defenses against replacement.

The story doesn't get better when you look at tenure and education. Of the 217 people in jobs with medium-to-low defense against AI replacement, 107 had over ten years of experience, and 65 had MBAs. According to Forbes, the average cost of an MBA is $61,800. Multiply that out, and the org has spent roughly $4 million on additional education for jobs that may not survive the next reorganization.

Roll it all up. The annual salary cost of the 99 medium-defensibility jobs is $15.4 million; the 118 low-defensibility jobs cost $12.8 million. Eliminating half the jobs from each category would yield $14.4 million in annual savings – a 10-basis-point improvement to net income.

From a strategic point of view, this is an exciting moment. Companies that lean into artificial intelligence will have a hiring advantage for high-potential talent, a competitive edge against their peers in product differentiation and unit costs, and the ability to unlock new, previously unreasonable ways of doing business. That much is obvious to anyone who has even dabbled with these tools, and it's supported by a growing body of academic research.

From a humane point of view, it's more complicated. Each slice of that graph is a livelihood, and the chart couldn't be a better depiction of inequality if it tried: most of those jobs have a limited defense against replacement by a sufficiently well-trained computer.

So, Now What?

I didn't do all of this research to prove out a bleak future. I did it to get clarity on the pitfalls and the wins sitting in plain sight.

And the analysis somewhat misses the point. We should automate the work that gives us little joy and where humans have a limited advantage over a computer. We should replace what we automate with better, fulfilling, growth-oriented work.

• Innovations that serve underserved and underrepresented communities.
• Breaking the systemic problems and patterns in place in advertising and marketing.
• Developing and launching new – and more sustainable – products.
• Creating deeper relationships with customers and partners.
• Supporting psychological safety with and for peers.
• Coaching teammates, building new skills and capabilities.
• Impacting cultures and society toward forward change.
• Deep learning and research.
• Going home on time, and focusing on health and wellness.

Most of us only have a playbook suited for a pre-AI world, and our job descriptions reflect that playbook. The good news is that playbooks and job descriptions are well within our control. Better, more future-ready versions are within our grasp. We just need the will to make the leap.

Four New Structures

Beyond directly rewriting the job descriptions that are under threat to emphasize growth, fulfillment, and humanity, leaders can adopt new structures to solve the problems that AI will create. I see four:

• Multi-Tenure Teams, solving the talent drain.
• AI Councils, keeping the business in line.
• Platform Teams, bringing up the level.
• Coaches, keeping the human side in touch.

Bad jobs for incoming top talent produce a weak long-term leadership succession plan. As I've shown above, AI will hollow out the roles new graduates have historically filled. Those new grads are more likely than anyone to be power-users of AI tools, but they'll be joining firms that don't have much of a place for them – most of the routine tasks that have provided training opportunities for junior talent will be automated or handed to a computer entirely. Top talent will simply stop considering roles that don't offer immediate access to intra- and entrepreneurship, strategic leadership, and innovation.

In the early days of social media, there was an uptick in reverse-mentorship, where junior employees coached senior executives on the digital revolution. Companies are going to need to go beyond mentorship and develop Multi-Tenure Teams, where new employees, senior executives, and middle managers break hierarchical boundaries and work as a single, cohesive team.

The result: new grads get better jobs. Companies maintain their talent pipelines. Tenured employees get a diffusion of knowledge from the younger generation.

In the mid-aughts, companies struggled to write effective social media policies. Legal departments are still struggling to handle "Shadow IT," the tendency for business units to acquire and use their own software systems outside of their company's governance framework. This issue will compound dramatically as generative AI explodes – and helps internal non-developers build their own tooling.

Companies will need markedly better governance forums than the ones they rely on today. Traditionally, governance has focused solely on data integrity and security. There's an opportunity now to create AI Councils that offer more holistic guidance, balancing business needs and ground truth.

To pull this off, AI Councils will need greater proximity to operations, with more council members coming from junior ranks rather than the executive layer, and from operating units rather than HQ. They'll also need more inclusive, digital tools for vetting, deciding on, and communicating new rules and regulations – tools that go beyond MS Word, PowerPoint, and traditional intranets.

Generative AI tools will eliminate the idea that an entire function requires centralization. Instead, we'll see dedicated teams turning their skills into software. These Platform Teams will take functions like "Marketing Insights" and break them down into individual pieces – a Concept Screener for assessing the validity of marketing ideas before they receive budget allocation, an Audience Optimizer for improving a media plan's targeting.

Every department I've observed has five to ten human-intensive tasks that require intense, inter-team agreements to spin up but are necessary to the proper functioning of the business. I've seen this administrative overhead consume around 70% of a team's available working time inside large organizations. Platform Teams turn those tasks into simple, digitized workflows that take a few clicks instead of a multi-quarter agreement.

In the short term, this lets distributed or decentralized teams maintain quality and business alignment without paying the administrative tax. In the long term, these capability-as-a-service implementations become a core differentiator for the firms that had the guts to invest in them today.

Like the previous structures, the coach role isn't 100% new. But with the changes ahead, it will need to – and finally be able to – expand into a first-class, insourced member of the firm. Coaches will be essential for bridging the gaps between people, forming and scaling high-performing teams, and developing the human side of the business.

I see an increased emphasis for coaching in three domains:

1. Leadership Coaches, focused on nurturing the leadership skills of the new workforce, ensuring a smoother transition into roles of responsibility and influence. Leadership coaches will guide individuals in decision-making, conflict resolution, and strategic planning, ensuring the cultivation of a robust leadership pipeline for the organization.
2. Team Coaches, focused not just on inter- and intra-team dynamics but on the relationship between AI tools and human teams. They'll work on enhancing collaboration, communication, and problem-solving skills, ensuring that the human-team-AI integration is smooth and productive.
3. Skill Coaches, focused on the continuous development of technical and soft skills – ensuring the workforce stays abreast of the evolving technological landscape while maintaining the human touch in their interactions and operations.

As we enter an era where the research tells us many roles are replaceable, the role of coaches becomes irreplaceable – and probably powered by a knowledge-and-coachee database that lets fewer coaches work with more of the workforce.

Wrapping Up

For leaders, this is not the time to adopt a wait-and-see approach. Generative AI is already actively in use across the corporate world, from approved and intentional approaches like Bain's OpenAI alliance and Expedia's ChatGPT integration to grassroots applications at the edges of the firm. These tools are already driving labor disputes, like the recently concluded months-long Writer's Strike.

There are pessimistic views of the impact of this new technology on our workforce, ranging from minimal returns on investment to drastic job loss. I take a different view. I hope we'll land in a future where AI has helped us build dramatically more productive workplaces, and where it has helped us elevate the human side of work.

But as with any new technology, the future we get relies on the decisions we make.

Let's dig in.

Earlier versions of this analysis were originally published in three parts on Black Glass and at Forbes.