I’m as skeptical in regards to the newest “shiny object” announcement as anybody else. And simply getting back from Cannes Lions, firms each giant and small had main information to share; the approaching of agentic synthetic intelligence was the discuss of the city.
Proper now, we’ve bought advertising groups working 80 to 120 completely different instruments and platforms. Half the time, they’ll’t even discuss to one another simply—they’re siloed. APIs come to the rescue and produce the mandatory platform-to-platform connectivity wanted to energy in the present day’s adtech ecosystem.
However APIs don’t remedy for cross-platform orchestration. If you wish to run a easy viewers activation marketing campaign throughout a number of channels, immediately you want a workforce of engineers simply to make it occur.
So, with everybody on the Croisette racing towards AI brokers that promise to automate advanced workflows, how can companies really begin AI-enabling their tech stacks?
That hope is the place the Mannequin Context Protocol is available in. MCP isn’t going to encourage screaming headlines about epochal change on your profession. There gained’t be any convention periods about MCPs changing your job as an award-winning artistic. However the potential could be very actual that MCP is rising as the usual that can make our agentic advertising future a actuality.
The muse of connectivity
APIs are the OGs of connectivity. They’ve been round because the Fifties (actually).
APIs are used to attach in the present day’s tech stacks, however they clearly weren’t constructed with AI brokers in thoughts. Contemplate a typical marketing campaign workflow: You need to construct an viewers in your buyer information platform, push it to The Commerce Desk and Meta, get reporting again, optimize based mostly on efficiency, observe conversions, and ship a abstract to your consumer. Sounds easy sufficient. However in actuality, you’re coping with completely different price limits, various information buildings, and about six completely different locations the place the entire thing can collapse.
Now, think about attempting to get an AI agent to deal with that workflow. With APIs, you’d want to show the agent each platform’s particular necessities, each error situation, each little gotcha that comes with every integration. That doesn’t make enterprise sense, particularly if you find yourself speaking about 80-120 platforms in your tech stack.
MCP flips this fully. As an alternative of brokers needing to grasp the technical complexity of each platform, they convey by way of the protocol servers that present context (the “C” in MCP) about what every system can do and learn how to work with it. Consider it as giving your agent a extremely good interpreter for each platform in your stack.
What agentic orchestration really appears to be like like
The promise right here isn’t nearly simplifying integrations. It’s about lastly enabling the sort of automated workflows we’ve been speaking about for years.
I may write right into a chat interface proper now, “Take our high-value buyer phase from final quarter’s marketing campaign. Run it on social and programmatic channels. Intention for a return on advert spend of greater than 4 occasions, and ship me updates each week on the way it’s performing.” An agent geared up with the fitting MCP connections may theoretically execute that whole workflow with out me touching one other platform.
Once more, this isn’t 5 years sooner or later. The foremost LLM suppliers have already adopted MCP as the usual: OpenAI, Gemini, and Claude are all supporting it. The longer term is now.
Why this time needs to be completely different
What I like about MCP in comparison with each different “revolutionary” integration answer we’ve seen is the modularity. You’re not locked into some vendor’s ecosystem or compelled to rebuild the whole lot to make it AI-enabled or if you add a brand new platform.
Managing API-based connections at scale takes actual sources. MCP provides a brand new connection layer that makes those self same APIs usable by brokers throughout a number of platforms, extending the worth of APIs, not changing them.
Give it some thought from a sensible standpoint. For those who’re working an enterprise tech stack, you may’t wait for each vendor to construct native AI integrations. However you should utilize MCP because the bridge between your current APIs and these new agentic capabilities.
The guardrails we really need
You would possibly very nicely be questioning whether or not we’re going to have AI brokers working free with our media budgets. Look, I get it. The concept of autonomous techniques making real-time selections with consumer budgets ought to make you nervous.
It comes right down to safeguards. MCP works by way of current API frameworks, in order that they inherit the guardrails you’ve already bought in place. Plus, you may layer on extra controls. Consumer profiles additionally restrict what brokers can do. Necessary approval checkpoints are an important safeguard for giant selections, and there are audit trails for the whole lot.
The aim isn’t to remove human oversight. The position of execs’ judgment is the purpose. As an alternative of manually shifting recordsdata between techniques and babysitting integrations, you have to be setting strategic parameters and making high-level optimization calls whereas brokers deal with the operational duties.
Begin small. Perhaps it’s simply automating viewers file transfers. Then perhaps it’s fundamental reporting aggregation. You crawl, then you definately stroll, then you definately run.
The selection
What most individuals aren’t occupied with but is that MCP isn’t nearly connecting your personal platforms. It’s about including a brand new layer of connectivity that different firms can plug into. The query is whether or not you’re constructing that orchestration layer or simply hoping another person builds it for you.
We must be smarter about this device. Select your MCP suppliers fastidiously. Begin with clear use circumstances. Construct governance frameworks earlier than you want them. And please, for the sake of your sanity, don’t attempt to automate the whole lot without delay.
That is the benefit of these early days. We are able to strategy MCP thoughtfully and create one thing that really works. Or … we will rush in blindly and create one other costly mess that requires armies of specialists to keep up.
