Every few years the marketing technology industry swings between two poles. It fragments into best-of-breed point solutions, each brilliant at one job, until the cost of stitching them together becomes intolerable. Then it consolidates into suites that promise one login and deliver mediocrity at everything. We are at the start of another consolidation, and the organising principle this time is unusual.
The previous consolidations were organised around data. Bring every record into one warehouse, one customer view, one dashboard. That was the CDP era, and it produced enormous value for large enterprises and enormous invoices for everyone else. Its limitation was always the same: consolidating data does not consolidate work. Teams still opened six interfaces to act on what the unified view told them.
Organised around action, not storage
The emerging model is organised around the doing. Instead of asking where the data lives, it asks who performs the task. Software agents connect to the same platforms a marketer would open — ad managers, analytics, content systems, commerce back ends — and carry out the work directly: reading performance, drafting the analysis, making the routine adjustment, publishing the page, scheduling the post.
This is a meaningful architectural difference. A dashboard is a read layer. An agent is a read-and-write layer with judgement in the middle. Categories such as marketing agency software are being rebuilt on that premise, and the products that result look less like reporting tools and more like a member of staff with API access and a rulebook.
Why now
Three things converged. Ad and analytics platforms finished exposing rich, structured APIs. Language models became reliable enough to interpret multi-source performance data and explain their reasoning in a form a professional would accept. And the cost of running that interpretation continuously fell far enough to make always-on supervision economically sensible for accounts spending thousands rather than millions.
None of those alone would have been sufficient. Together they moved a whole class of work from “possible in principle” to “cheaper than the status quo.”
What buyers should actually interrogate
The category is young and the marketing is ahead of the substance in places. Four questions separate serious products from demonstrations.
Does it write, or only read? Many tools described as agents are dashboards with a chat box. Ask specifically which actions the system can execute in which platforms, and what happens when a platform rejects the call.
Where do the guardrails live? Meaningful autonomy requires configurable boundaries — spend ceilings, protected campaigns, action types requiring approval, quiet hours. If those are absent, the choice is between no autonomy and reckless autonomy.
Is there an audit trail? Every automated change should be recorded with the triggering signal and the reasoning. Without it, nobody can explain the account’s history, and that becomes a real problem the first time a client or a finance team asks.
How is client data isolated? For anyone managing multiple brands, cross-contamination between accounts is not a theoretical risk. Ask how contexts are separated and what the vendor does with the data.
The likely shape of the next few years
Point solutions will not disappear — specialised creative tools and measurement platforms will keep their place. What is likely to erode is the middle layer: the reporting products, the alerting products, the workflow products whose entire function was to move information between systems and prompt a human to act. When the software can act, the prompt is redundant.
For marketing teams the practical advice is unglamorous. Do not rip out a working stack for an emerging category. Do start by pointing an agent at a well-instrumented account in read-only mode and comparing its analysis with your own for a month. That comparison tells you more about whether the technology is ready for your context than any vendor demonstration will.
Consolidation cycles reward the patient. The teams that come out ahead are rarely the earliest adopters — they are the ones who understood the new organising principle first and picked their moment.