Agentic AI and the Evolution of Agencies

1.0 Context: The Sputnik Moment : The Democratization of AI: Open-Source, Low-Cost, and the Rise of Reasoning Models.

The recent "Sputnik moment" of DeepSeek, marked by advancements from pre-training Level One: Conversational AI models to the emergence of Level Two : Reasoning AI Models with agentic capabilities, has set the stage for a tipping point in AI evolution. This milestone, debuted on January 27, 2024, signifies a new era in AI, further spotlighted by discussions at the World Economic Forum in Davos. On a personal level, AI is also a course I am currently pursuing has further spurred me to reflect on its’ impact. These events have collectively piqued my interest in how AI will ultimately reshape the marketing and communication industry specifically media, leading me down a rabbit hole of discovery that culminates in this journal of perspectives on the near future.

Why Does This AI Evolution Matter?

Generative AI is not only transforming us as consumers and individuals but also revolutionizing how work is executed and delivered. It is seamlessly integrating into our digital lives—be it social media (Llama, Symphony), search (Perplexity, Google), decision-making, or daily tasks (Co-pilot, Gemini, ChatGPT, DeepSeek). Even customer service is being reshaped at an unprecedented pace.

Generative AI is now a platform competing for our attention and time. It took just two months for ChatGPT to reach 100 million monthly active users (MAU), compared to nine months for TikTok and 30 months for Instagram (for reference, Threads achieved this in a record-breaking 0.2 months).

Its influence extends beyond our own individual life. The latest iteration, Agentic AI with reasoning capabilities (e.g. OpenAI o1 & Operator, and OpenAI o3-mini, launched in response to DeepSeek), is mature enough to reshape organizations. Moreover, open-source models like DeepSeek and QWen are making this technology widely accessible, accelerating its democratization. For instance, Perplexity has already integrated DeepSeek V1 into its service.

Notably, DeepSeek was developed with just $6 million—demonstrating that building custom AI models is now within reach for many organizations. And this is only Level Two of AI’s evolution. According to OpenAI, there are additional levels: Autonomous AI, Innovative AI, and Organizational AI (Levels Three to Five), each promising even more transformative capabilities.

2.0 The Agency of the Future

Through the lens of Level One and Two AI, here’s how I envision AI shaping the future of agencies:

1. Data-First Agencies: Foundation, Architecture, and Visualization

Agencies with a well-structured, data-rich foundation will lead with a competitive edge. Data will be the backbone of Agentic AI, fuelling model training and guiding reasoning. A diverse dataset not limited to audience insights, performance metrics, creative response data, commerce analytics — will be essential for enhanced performance.

Agentic AI will automate Extraction, Transformation, and Loading (ETL) of new data and process it at unprecedented speeds. Additionally, AI-driven systems will ensure data architecture and taxonomy adherence by identifying irregularities before deployment without human intervention. Moreover, manual macro drafting is no longer necessary, leave that to AI, as ChatGPT can already connect to Google Sheets. With AI-driven data transformation, agencies will be able to merge datasets seamlessly into a data lake, eliminating bottlenecks in data cleaning. Once structured, data will be instantly accessible through AI-driven prompts, reducing reliance on Excel pivots, macros and manual analysis.

2. Unlocking Consumer Behavioural Insights and Trends

Trend analysis will move beyond mere observation of signals to understanding the reasoning behind them. Agentic AI will enable real-time data cleaning, processing, and integration, transforming regionally nuanced data with auto-translation and auto-clustering. It will also synthesize data from multiple sources (e.g., social listening, e-commerce reviews, brand trackers, social policies) to detect micro-signals that inform macro trends. For example, Gemini can already analyse visual content from youtube, making sense of video data by clustering and contextualizing information.

Gone are the days of manually cleaning, clustering, and contextualizing data. AI agents will handle data acrobatics and contortions in real time, allowing strategists to focus on higher-order analysis.

3. Real-Time Responsive Personalization at Scale

While creative teams will provide guidelines, AI-driven systems will handle the creation of personalized iterations. Audience data lakes, integrated with various second-party data sources and brands' first-party data, will enable dynamic adaptation of creative assets. This is made possible as AI will not only have language capabilities but also audio visual capabilities, remix capabilities along with synthetic voice syncs and coupled with reasoning it can be on auto drive. AI will execute real-time optimizations, reducing lag time in redesigning content or targeting new audiences when engagement declines. We are already seeing baseline technologies of this via TikTok’s Sympohony.

4. Multi-Objective Optimization for Brand and Consumption

Seamless integration between brands and commerce will allow direct engagement between brand AI agents and consumer AI agents (A2A). API Partnerships with super apps and social commerce platforms—leveraging direct and proxy data on purchase behavior (value, frequency and recency )—will enable precise targeting without breaching data privacy regulations. Understand consumer data will unlock agency’s ability to help brand retain consumer in a sustainable way while expand the brand’s reach to target new consumers. By overlaying this data with brand health insights, brands will optimize for Top-of-Mind (TOM) awareness and Ad Recall simultaneously against consumption too. Do take not the word is consumption – using, eating and wearing and not just commerce / purchase.

5. Autonomous Media Planning and Execution

Building on points made earlier, along with a recent experiments demonstrate that reasoning models are already capable of designing media plans optimized against Marketing Mix Modeling (MMM). AI can generate model and refine media strategies within 30 minutes—simply through conversation. AI-driven systems can balance multiple objectives, such as reach and lifecycle, while optimizing campaign deployment by interfacing with platform setups through AI agents. These agents operate within a real-time audience response loop that continuously learns and optimizes.

6. Architect of Immersive Worlds to Unlock Experiences

It will become the norm for agencies to offer brand experiences that straddle both virtual and physical realms. "Shrooming" will be unlocked on any imaginable surface, creating new touchpoints for engagement. By linking smart city data—using Google Maps as a baseline alongside its spatial technology ARCore — with devices like Meta's smart glasses equipped with cameras, image recognition, and speakers, these experiences will become accessible to all. By the way, all of this is technically possible, as demonstrated at SIGGRAPH, where WPP and NVIDIA unveiled the technology behind 3D worlds.

7. Evolving Service Offering: Transformation Consulting and In-House AI Models

Agencies will transition into AI transformation consultants, offering AI Models as a Service (AMaaS). Custom AI models will be integrated within client ecosystems to maintain data privacy, with agencies providing periodic upgrades. This approach mirrors app update cycles which we are familiar with, allowing modular pricing structures where clients pay for base models and additional child models (akin to ChatGPT’s integration with DALL-E).

3.0 Evolving Agency Functions and Roles

AI and Agentic developments will bring about significant changes in agency roles, expanding current competencies:

1. Communication Coders & Algorithm Architects

Strategists and planners will evolve into hybrid roles requiring a blend of computational logic a splash of creativity. The role requires familiarity with emerging generative AI Agentic models, like Alibaba's Qwen 2.5 / OpenAI Operator, which already demonstrates advanced capabilities, such as booking flights. They will architect “Responsive Communication Models” that works with platform algorithms to deploy campaign and consumer’s agents who predetermines groceries and consumption choices.

These Responsive Communication Models and their plans will be designed to be auto-responsive, integrating real-time consumer insights and competitive intelligence. They will seamlessly feed into a continuous loop, adapting to market shifts—whether by driving incremental audience reach, pacing inventory based on stock levels, or being price-responsive benchmarked against competition.

2. Models & API Partnership Specialists

As campaign tactical execution such as : monitor campaigns, performance optimisation, consolidating budgets and swap assets becomes automated by Agentic AI, planners will shift their focus to optimizing AI models to deliver large scale optimisation instead of micro-optimising and managing API partnerships. API Partnership Specialists will negotiate data exchanges and integrate APIs from publishers, networks, creators, and super apps to enhance AI-driven optimizations.

3. New Roles in AI Transition

Transformation Consultants: Transitioning to AI-driven workflows will require a specialized team to identify efficiencies, customize models based on client business needs and mould an AI collaborative culture. They have a special blend of skillset in technology and change management. ?They will help both clients and teams look into the existing scope, deliverables and identify opportunities for efficiencies through Agentic AI. They will also ensure brands are able to navigate the AI world and have the agencies internally jorney hand in hand.

Forward Deployed Engineers: Inspired by Palantir, these engineers will tailor AI models to client ecosystems. They will work closely with Transformation Consultants to ensure existing models are customized, relevant to each client’s needs, and continuously updated. Each model will be uniquely designed for the client.

I am sure more roles will evolve, but these are some of the ones I foresee undergoing the most significant transformation. Meanwhile, data experts working on data lakes will remain foundational, ensuring the stability and integrity of AI-driven systems.

4.0? Irreplaceable Human Skill Sets

While AI can handle many tasks, as a generation, we must further equip ourselves with technical skills to navigate the next decade. However, some soft skills remain uniquely valuable and will continue to distinguish us from AI.

1. Randomness Over Mere Reasoning:

Critical Thinking and Creativity Reasoning models follow structured logic, but they cannot replicate human curiosity or unorthodox problem-solving.

The ability to think laterally and creatively—to connect disparate ideas and generate originality—is beyond the capabilities of reasoning models.

While reasoning may be a part of critical thinking, broader skill sets involve identifying biases, questioning assumptions, and pushing the boundaries of logic—abilities that AI models, at present, cannot replicate.

Curiosity: The Root of Innovation – While AI can hallucinate, it lacks true curiosity, the ability to wonder "what if," or the drive to explore. This remains a uniquely human trait—our ability to ask questions, discover, and dream.

Neurodivergent thinkers, in particular, will thrive in this era of innovation.

2. Cross-Functional and Cross-Industry Intelligence

As open-source AI models evolve within corporate walled gardens due to data protection and security concerns, their growth will be limited by the data they are trained on and the scenarios they optimize for.

Individuals with diverse industry experience will be crucial in bridging these gaps, fostering innovation across industries and functions.

3. Experiential Over Mere Engagement

Automation may dominate commerce and consumption, as brand agents interact with consumer agents in an A2A (Agent-to-Agent) world. This will make it increasingly difficult to introduce new brands and products.

However, to disrupt consumer choices or intervene, brands must connect directly with people. The way a brand makes someone feel through experiences—emotional resonance—will outweigh optimizations and efficiencies. The ability to craft meaningful experiences and forge deep emotional connections will be a key skill in delivering lasting brand impact and connection.

4. Authenticity in the Era of Alpha

As macro trends emphasize efficiency and speed, countertrends will emerge. Generations Z and Alpha value deep connections and imperfection, making authenticity a vital component of brand strategy.

Social, community, and cultural engagement will continue to thrive, enabled by serendipitous encounters that fuel authenticity.

Signing Off: A Student of the World and Aspiring Futurist

This article reflects my personal perspectives and is not affiliated with any organization I currently work for or may work for in the future. As a firm believer in the limitless possibilities of human imagination, I look forward to the future with curiosity and optimism.

References for Further Reading:

1.??????? A Day in the Life of a Palantir Forward Deployed Software Engineer Read more on Palantir Blog

2.??????? Advanced Language Model Reasoning: Pre-Training, Fine-Tuning, and Inference Time Techniques Read more on LinkedIn

3.??????? Agentic AI: 4 Reasons Why It’s the Next Big Thing in AI Research Read more on IBM Think Insights

4.??????? Alibaba Releases AI Model It Says Surpasses DeepSeek Read more on Reuters

5.??????? Alibaba’s Qwen Team Releases AI Models That Can Control PCs and Phones Read more on TechCrunch

6.??????? Bob McGrew: AI Agents and the Path to AGI Watch on YouTube

7.??????? Can AI Replace Human Curiosity? Read more on Forbes Tech Council

8.??????? ChatGPT on Track to Surpass 100 Million Users Faster than TikTok or Instagram: UBS Read more on Yahoo Finance

9.??????? ChatGPT Sets Record for Fastest-Growing User Base - Analyst Note Read more on Reuters

10.? DeepSeek Is Here to Stay as Microsoft, Perplexity Integrate Its Model Read more on PCMag

11.? DeepSeek Just Flipped the AI Script in Favor of Open-Source—and the Irony for OpenAI and Anthropic Is Brutal Read more on Fortune

12.? DeepSeek vs. ChatGPT: How Do They Compare? Read more on DataCamp

13.? DeepSeek-R1 vs. OpenAI’s o1: The Open-Source Disruptor Raising the Bar Read more on GeekyAnts

14.? DeepSeek: Everything You Need to Know Right Now Read more on Exponential View

15.? Flagship Models Read more on OpenAI Platform

16.? Get more out of life with AI glasses Read more on Meta

17.? Here's What DeepSeek AI Does Better Than OpenAI's ChatGPT Read more on Mashable SEA

18.? In the Age of AI, Critical Thinking Is More Needed Than Ever Read more on Forbes

19.? Make the world your canvas Read more on Google Developers

20.? Meridian is Now Available to Everyone Read more on Google Blog

21.? OpenAI Fights Back Against DeepSeek AI With Early o3-Mini Launch—Here's How It Compares Read more on Decrypt

22.? OpenAI Introduces Operator to Automate Tasks such as Vacation Planning, Restaurant Reservations Read more on CNBC

23.? OpenAI to Release New Artificial Intelligence Model for Free Read more on The Guardian

24.? OpenAI’s 5 Levels Of ‘Super AI’ (AGI To Outperform Human Capability) Read more on Forbes

25.? Reasoning Models Read more on OpenAI Platform

26.? Sam Altman’s Answer to DeepSeek Is Giving Away OpenAI’s Tech Read more on WSJ

27.? Take on Tomorrow @ the World Economic Forum in Davos: What Does Responsible AI Look Like in the Age of Agentic AI? Read more on PwC

28.? The Latest ChatGPT Statistics and User Trends (2022-2025) Read more on WiserNotify

29.? US Tech Shares Recover Some Losses From Steep DeepSeek Selloff Read more on Reuters

30.? Using Reasoning for Data Validation Read more on OpenAI Cookbook

31.? Using Reasoning for Routine Generation Read more on OpenAI Cookbook

32.? Want to Try DeepSeek Without the Privacy Worries? Perplexity AI Just Launched It on Its iOS and Web Apps Read more on TechRadar

33.? What Is Agentic AI, and How Will It Change Work? Read more on HBR

34.? WPP to create generative 3D worlds in collaboration with NVIDIA Read more on WPP