Welcome to 2025! It will be the Year of AI agents In Marsch, in word and did. This is what I think that means …
From the moment there was more than one software product in the technological battery of a company, there is a leg to determine who the boss was.
If the technological battery were an orchestra, what product would the driver be?
Whicever one directed the data flow between the applications and determined what function in which the application was called when, literally, orchestra the technological battery, it would be a position of great power.
That is why it was a strategic error for marketing suites in the beginning of 2010 to minimize integrations with other products in the Burgewoning Panorama Martech. Other products that remain necessary or important. Having struggled to be the driver would have a leg to recognize the value of other applications. They did not do it because to conceive that point.
Like Julia Roberts’s character Pretty woman He rebuked, “big mistake. Great. Huge!”
The failure of suites to confiscate the baton created an opening for a completely new category of products known as “IPAA” -Interación-Platforma as a service to go to the stage and build multimillion-dollar companies, such as Mulasesoft, Workato and Zapier. IPAAS adopted the diversity of technological batteries and trained people to address digital operations at a level above individual products.
The suites finally realized that they became a spoken on the edge of the graphic of the pile instead of the center in its center. Many changed to platform strategies and pursued construction applications ecosystems. Those who did not slide into irrelevance. Salesforce acquired Mulasesoft. Hubspot acquired feet.
But the barn’s door already opened, and the cows never realized the house. To this day, IPAA products, many of which have become “business automation platforms”, retain a strong presence in most marketing technology batteries. They are coonductors of the orchestra. For some cases of use, the marketing platform calls the shots. For others, the Née Ipaas business automation platform.
Ironically, the story was almost repeated a few years later. CDPS-Customer-Arose data platforms between 2015-2020 because most marketing suites/platforms were not yet open or flexible enough with their data models. This great CDPS is an opening similar to IPAAS, and stamped it.
But marketing suites/platforms were not so resistant to respond this time. The majority acquired a CDP or CDP -type capabilities built for the 2020s (some excess compensated by Buychasing or Building multiple CDPS, but that is a bit inside baseball). Meanwhile, in a surprise turn, cloud data stores/lake houses interrupted the entire CDP landscape. However, that is a story for another day.
Which takes us to 2025. Modern Martech Stack has multiple platforms competing for being the orchestra director in the “center” of everything. Many marketing operations teams try to standardize one to automatic workflows through their tools set, Althhehegh, whose platform they choose: CRM, MAP, CDP, CDW, etc.) varies from one company to another. But almost 1/3 also uses an IPAAS/Enterprise automation platform.
Complex? A bit. Management management? More or less.
But prepare. You are about to make it more complex with – battery roll – AI agents.
While there is a lot of debate about what qualifies as a “true” agent on how much decision -making autonomy has a characteristic that almost all agents share: Use of the tool. They can interact with another software, either through API or simulating the clicks of a human user and writing, similar to RPA. They apply this use of the tool to read or write data from different sources, take measures in other applications or make applications to other agents. The liter is the vision of the networks of multiple agents.
AI agents can assume tasks, processes or even more open objectives. They are automatic digital operations in different products in their technological battery in an integrated way.
Except for the objective observation turn, that sounds familiar, right?
The agents of IA reverberate with the echoes of the IPAA and workflow automation platforms. In fact, many of the suppliers in these categories have become platforms for the construction of AI agents. The difference? The workflow automations were agreed with the deterministic rules and steps that were programmed by a human. IA agents apply their own intelligence to decide what tools to call when, without an explicit plan dictated by humans.
The existing platforms of Marsch Saas are also either dragging their feet to the party of the AI agent. Salesforce has balanced his narrative pendulum until agentforce. Hubspot has its agents Breeze AI, and co -founder Dharmesh Shah launched Agent.ai, “the professional network for AI agents.” Microsoft has incorporated Copilot Studio to build dynamic agents. Braze, Bloomreach and Zeta have also announced their own AI agents.
Now, if we were simply dealing with the existing IPAA/Automation suppliers and the Ecosystem -oriented SAAS platforms that are transformed into AI agents platforms, that would be enough. In the short term, digital operations will become a tangle of old deterministic workflow automation intermingled with new intelligent agents automation. We will have to migrate to new forms of architecture of marketing operations processes accordingly. But we will obtain the rise of the agents that do more work autonomously, reducing the resumes of hand -kept rules. (Hallelujah, right?)
However, existing SAAS and IPAA/Automation companies are No The only players in the well. Multiple new categories of AI agents have been joined, each with its own intelligent version of individual productivity and/or the automation of organizational processes.
Here is a quick tour of the other taxpayers of the agents on the previous map …
The Main Motors of Anthrope, Google, Meta, Openai, et al., All development support agents, with “calls of functions” for the tool as a basic primitive. The recent “computer use” capacity of Anthrope and the model’s context protocol are exciting steps forward in the fundamental agent capabilities.
Marcos of developers of agents such as Autogen, Autogpt, Camel AI, Greiptape, Langchain and Flame Index help facilitation companies to build their own custom agents. If you can code with APIS and with developer tools with AI as a cursor. Someone Tin code: You can build your Ownered ia agents in a matter of days or hours.
Why even more helps build your own AI agents? There is a completely new category or constructors of low/low code agents: Crewai, Lindy, Reavence.AI, Sema4.ai, Smythos and Tektonic.AI, to name just some. The limit between the thesis and the evolved workflow automation platforms is extremely confusing. I would merge them, except it is worth pointing out how many new Native Builder Platforms of AI agents have emerged in this category.
At the other end of the spectrum, Anthrope, Google, Microsoft and OpenAi are incorporating the use of tools in their end user products such as Claude, Chatgpt, Copilot and Gemini. More and more, the use of Gemini within Google’s work space or in co -pilot in Microsoft 365 products, such as Excel and equipment, will perform automation similar to agents for people. There will be an unwinding overlap between personal productivity and commercial processes automation.
The independent automation tools based on the browser such as Barts and Axiom.AI will also be part of the mixture. Whatever the web -based software that a user has access, can create AI (or pseudo agents) agents to automatically their work with those applications, even if the suppliers of those products have not done anything free to enable that.
There is a special category of agents that have emerged to take advantage of learning reinforcement, a automatic learning sub-rama of autonomous personalized campaign and marketing learning programs of Ly-Mize. Offerit is a pioneer in this space (Disclosure: I have an advisor for them). Aampe and CDP Hightoouch compound have also entered the category, which is known as the “decision of AI”.
Finally, but definitely no less important, there is a massive wave of tools of the specific roles agents, especially for sales and customer service. Representative AI agents include 11x, AISDR, Artisan, Momentum, Directie.Ai and Rox. The AI agents of the client’s success include Decagon, Mavenagi and Sierra.Ai. Of course, existing SAAS platforms compete with their own specific roles agents, such as Hubspot, Intercom, Salesforce and Zendesk.
Do the different AI agents have in common?
All are automating interactions, more and more autonomous, with the products and platforms in their pile. Everyone thinks that he is orchestrating his piece of symphony. But will there be a driver who sees throughout the orchestra and harmonizes all these players? Wait for intense supplier competition for that position.
I looked, orchestrating the orchestradors will be the great challenge of the great operations, in 2025 … and during the rest of the decade ahead.
