Two forces are colliding in marketing this year. First, the martech universe is still expanding, not shrinking. Scott Brinker’s 2025 landscape counts 15,384 products, up 9 percent year over year, and he notes that AI is lowering the barrier to create new software even as older categories consolidate. Second, customer expectations for content and instant service keep rising. Adobe’s latest India cut shows 84 percent of marketers expect content demand to grow fivefold by 2027, with nearly all saying demand has at least doubled already.
Below are the five shifts that matter now, with what industry leaders are actually saying in public.
1) Agents are moving from chat sidekicks to full operators
The story of 2025 is agents that can plan, call tools, and act across systems. Microsoft’s CTO Kevin Scott frames it as an “agentic web” built on open standards so agents can talk to services directly. “It means that your imagination gets to drive what the agentic web becomes,” he said, backing the Model Context Protocol and calling for interoperability.
Enterprise apps are going the same way. At Davos, SAP CEO Christian Klein called agentic AI “a big next step,” outlining paired sales and supply chain agents that coordinate in real time so “the sales agent is not making any kind of promises, which the supply chain can’t fulfill.”
Meta is pushing the idea to consumers. Mark Zuckerberg’s July letter set a clear direction: “Meta’s vision is to bring personal superintelligence to everyone.” He also argued that glasses will become “primary computing devices.”
Why it matters for marketers
Agent workflows will move from suggestions to semi autonomous execution, particularly in merchandising, pricing, creative testing, and lifecycle operations. Your stack will need clear guardrails, audit trails, and well defined permissions long before agents get free rein.
2) The center of the stack is changing
CDPs are getting squeezed between warehouse native data and engagement platforms with embedded identity and decisioning. Brinker’s 2025 analysis shows more teams naming their cloud data warehouse or engagement layer as the stack’s “center,” while CDPs slide as capabilities shift to those layers.
At the same time, stacks are expanding again. “Yes, their stacks are expanding,” Brinker told MarTech. “And I don’t think it’s unreasonable to lay the credit or the blame for that on AI.”
Why it matters
Plan for warehouse first patterns, composable activation, and decisioning that sits closer to the channel. Tool counts may go up before they come back down, so integration discipline and naming conventions will save you real money.
3) From content explosion to content supply chains
India is an early signal. Adobe’s research shows 84 percent of Indian marketers expect content needs to grow fivefold by 2027. Adobe’s CEO Shantanu Narayen put a sharper point on how AI changes the work: “India’s next growth as an economy won’t be in software code but in creativity.”
Why it matters
Generative tools are only step one. The operational win is building a supply chain for content, with governance, reuse, and authenticity signals. Expect more brands to adopt content credentials to verify provenance as synthetic media scales.
4) Safety, data quality, and risk triage are becoming frontline marketing skills
Agentic approaches heighten the blast radius of bad data and weak controls. Practitioners are getting pragmatic. In a MoEngage panel this July, SBI Securities’ Vipul Sharma advised teams to “start with those use cases [that are low risk] and then maybe move to the other ones where you feel there might be some hiccup.”
Why it matters
Adopt a risk ledger. Classify use cases by regulatory exposure, user harm potential, and model brittleness. Mask PII, constrain tools an agent can call, and log every action. The teams that scale AI safely will ship faster, not slower.
5) Build or buy is bending toward build again
Survey data shows a resurgence of homegrown components, with teams crediting “new generation AI tools” that let non developers ship simple apps and automations. Brinker’s landscape piece calls out a “hypertail” of instant software created by pros and “citizen developers,” enabled by AI.
Why it matters
Expect more internal mini apps and agents stitched to your warehouse and messaging rails. This does not replace your platforms. It augments them with thin, business specific layers that move at the speed of the brief.
What this means for your 2025 roadmap
Anchor on a clear operating model for AI. Define owners for data, model governance, and agent access. If responsibilities are vague, incidents are inevitable.
Choose one or two agent use cases to productionize. Pick contained, measurable workflows, for example churn reactivation or price testing, and wire them end to end, including rollback and human approval steps. Use the low risk first rule of thumb from SBI Securities.
Recenter data. If activation is still bottlenecked by brittle pipelines, prioritize your warehouse schema, events taxonomy, and consent state before you add more AI on top. The stack center is drifting toward data plus engagement for a reason.
Treat content like a supply chain. Build libraries, templates, and review paths. Adopt content credentials to protect creators and reassure audiences as synthetic assets scale.
Invest in talent, not just tools. The strongest executive takeaways remain human centric. As GroupM’s Prasanth Kumar put it, “the true value of AI lies in its ability to enhance human creativity and strategic thinking.”
