Riding the AI Tsunami: The State of MarTech in 2025
The world of marketing technology is in a constant state of flux, but 2025 feels different. According to Scott Brinker and Frans Riemersma, the authors of the "State of Martech 2025" report, AI isn't just another wave; it's a tsunami. With 15 years of tracking the MarTech landscape, witnessing its growth from 150 products in 2011 to a staggering 15,384 solutions in 2025, the authors note that the pace of change has accelerated dramatically, particularly over the past six months. This rapid technological advancement contrasts sharply with the slower, logarithmic pace of organizational change, a phenomenon captured by "Martec's Law".
The report delves into how AI is transforming marketing and the underlying technology stacks. AI's evolution is moving at high speed, progressing from large language models (LLMs) that handle one-off prompts to AI Agents that turn prompts into autonomous sequences, and now towards integrated Model Context Protocol (MCP) systems that enable agents to work with real-world systems and reference external data. MCP, a new standard for AI agent communication and integration, has seen rapid adoption.
The Pervasive Sparkle of AI Adoption
AI is not just a theoretical concept in marketing; it's being actively adopted. A survey conducted for the report revealed that 87.5% of respondents reported using AI assistants in their marketing work. While passive use cases like AI search results have crossed into the early majority phase, AI agents and agentic workflows are still primarily in the innovator and early adopter stages.
The most common generative AI use cases in 2024 included content ideation (69%), content copy production (62%), and meeting transcriptions, notes, and summaries (53%). Looking at broader AI use in marketing workflows, content production (78.1%) leads, followed by audience segmentation and targeting (42.7%), and data analysis and reporting (38.5%). Personalization for email/SMS and website/mobile apps is also a significant use case.
However, integrating these AI tools into existing martech stacks remains a challenge for many, with 41% of companies facing significant hurdles. B2B companies have generally reported greater success with integration compared to B2C organizations.
Reshaping the Martech Stack: Systems of Knowledge and Context
The traditional martech stack is evolving, becoming incredibly malleable and underpinned by multiple systems of knowledge. The report introduces a conceptual model that divides the stack into systems of knowledge and systems of context.
Systems of Knowledge: These systems own and manage the core data, serving as the "shared reality" of the business. The Cloud Data Warehouse/Lakehouse is increasingly the bedrock of this domain, acting as a universal data layer for storing and distributing data. Platforms like CRM, DAM, MDM, and ERP function as sources of truth for specific business facets (customers, assets, etc.). Many CDPs or CDP-like functionalities bridge the gap, organizing data for more situational needs.
Systems of Context: These systems govern, decide, and deliver content and experiences. Traditional platforms like MAPs, CEPs, DSPs, and DXPs have typically managed context in a more fixed way, with predefined rules and workflows. The new generation of AI agents and assistants, however, creates and serves context in a much more dynamic fashion, often driven by natural language prompts.
A notable shift is occurring with Customer Data Platforms (CDPs). While CDP functionality remains crucial, the standalone CDP category has seen significant consolidation. Their capabilities are being absorbed either upstream by the cloud data warehouse (composable CDPs) or downstream into engagement platforms through mergers and acquisitions. This disruption is evident in survey data, where the percentage of B2C and joint B2B/B2C companies identifying a standalone CDP as the center of their stack dropped significantly from first to fourth place in just one year. Jonathan Moran of SAS notes that their CDP capabilities are embedded within their customer engagement platform, an approach he sees the market adopting.
An emerging layer highlighted in the stack model and discussed by interviewees is AI Decisioning. Tejas Manohar of Hightouch describes this as a newer paradigm where AI agents work towards specific business goals by deciding the best possible experience or message for each customer. This moves beyond traditional rules-based or single-model prediction, leveraging machine learning techniques like reinforcement learning to make adaptive decisions and optimize outcomes over time. The advantage of an independent AI decision-making layer, according to Hightouch, is the ability to make consistent decisions across fragmented MarTech stacks and leverage the full breadth of data in the data warehouse.
The stack also needs to consider Buyer-Side AI Agents – agents owned by customers that will increasingly interact with a business's systems to control their own context. Businesses need to design platforms that are open and interoperable to enable these third-party agents.
The Exploding Hypertail of Custom Software
Beyond the visible landscape of commercial software (the "long tail"), there's a booming "hypertail" of custom-built software. Historically expensive and requiring specialized skills, custom software creation is being rapidly democratized by low-code/no-code platforms and the accelerating power of AI. AI allows "citizen developers" to build applications and automations by simply describing what they want in natural language prompts. Even popular AI assistants are creating software programs behind the scenes for users without them realizing. This explosion of custom applications, often built by non-technical users or AI itself, adds another layer of dynamism and complexity to the MarTech environment.
Navigating Complexity: From Complicated to Complex
The introduction of AI agents and probabilistic LLM outputs is fundamentally changing the nature of the MarTech stack from complicated to complex. While a complicated system (like a traditional MarTech stack) is deterministic with predictable cause-and-effect relationships (like a Ferrari, easy to drive but hard to fix), a complex system is probabilistic, where the same input may yield different outputs. AI introduces this probabilistic behavior into marketing workflows.
Managing a complex environment requires different strategies than a complicated one. Recommendations for navigating this shift include:
Embracing composable, loosely-coupled architectures.
Enhancing observability with streaming logs and alerts to detect deviations.
Engaging in safe-to-fail experiments to understand probabilistic behaviors.
Instituting human-in-the-loop checkpoints for critical decisions.
Building cross-functional "sense-maker" pods to interpret emergent behaviors.
Investing in greater data maturity.
Promoting a culture of experimentation.
Jonathan Moran notes that AI is allowing for "probabilistic journey mapping," where reinforcement learning helps systems recommend optimal paths for individual customers based on learning, rather than forcing them down predefined deterministic journeys.
Evolving Roles and the Human Element
As AI takes on more "jobs-to-be-done," particularly in the areas of production and analysis, marketers' roles are shifting. The report suggests more bandwidth will be freed up for the higher-value, more creative work of strategy and creative.
Critically, the vision presented is not one of AI replacing humans entirely. Chris O’Neill of GrowthLoop describes AI more like a "Tony Stark Iron Man suit" – it augments human capabilities rather than automating everything in a "black box". Sara Faatz of Progress echoes this, emphasizing that AI is "still about humans building for humans" and that human oversight will remain essential, especially for data quality and autonomous AI actions. Raviteja Dodda of MoEngage believes marketing teams will focus more on strategy, broader CX, and identifying opportunities, with AI assisting in ideation and execution. The increasing technical nature of marketing operations also necessitates closer collaboration and blending of skills between marketing teams and IT/engineering departments. Greg Brunk of MetaRouter stresses the importance of both sides understanding the other's goals and constraints.
The Foundation: Data, Decisioning, and Experimentation
At the core of this AI transformation are data, decision-making, and experimentation speed. The cloud data warehouse provides the essential data foundation. A key unlock with AI is the increased ability to leverage unstructured data, such as call recordings, emails, survey responses, and social media posts, which LLMs are particularly good at summarizing and distilling insights from. While data quality remains a challenge, AI can assist, although human intervention is still necessary.
The ability to make intelligent, goal-based decisions leveraging vast datasets is moving to the forefront with AI Decisioning platforms. This shift is critical for delivering truly personalized and effective customer experiences at scale.
Finally, the concept of Compound Marketing, introduced by GrowthLoop, highlights the critical role of iteration speed and experimentation loops. By accelerating the cycle of testing, learning, and iterating using AI at every step, marketing teams can drive compounding growth faster. This focus on velocity and continuous experimentation is seen as a key differentiator for high-performing teams.
The state of MarTech in 2025 is characterized by exponential technological change driven by AI, a growing landscape of solutions (including a dynamic hypertail of custom software), a reshaping of the MarTech stack around data and dynamic context, a shift from complicated to complex operational environments, and an evolution of marketing roles that emphasizes strategy, creativity, and human oversight augmented by powerful AI tools. It's a challenging but "damn fascinating one."