Open protocols are making agentic AI easier to buy

Enterprise buyers are getting more interested in agentic AI.

They are also getting more skeptical.

The skepticism is healthy.

Most operators do not want to buy another giant platform, migrate half their stack, and hope the promised automation shows up later. They want a system that can work inside the tools they already use, complete a real unit of work, and stay governable once it goes live.

That is why one of the most important AI trends right now is not just model quality.

It is interoperability.

Why this trend matters now

The market signal has shifted from experimentation toward operational adoption.

Microsoft's 2025 Work Trend Index said 82% of leaders viewed 2025 as a pivotal year to rethink strategy and operations, and 46% said their organization was already using agents to fully automate workstreams or business processes.

That matters because once buyers move from "interesting AI tool" to "system that runs work," integration stops being a technical detail. It becomes the buying decision.

If an agent cannot read from the right systems, hand work to another system, or leave a clean audit trail across the stack, it does not matter how impressive the demo looks.

The market is moving toward open agent plumbing

Over the last year, the AI ecosystem started to standardize around the boring part that actually matters in production: how systems connect.

Two examples make the point:

• OpenAI now supports MCP-powered apps and connectors in ChatGPT, including enterprise workflows that can take write actions in company systems with approvals and admin controls.
• Google's Agent2Agent protocol is gaining traction as a way for agents to communicate across platforms, and Google said in July 2025 that the ecosystem already included support from more than 150 organizations. Google also said it contributed A2A to the Linux Foundation in June 2025.

These are not niche developer-side signals anymore.

They point to a broader market direction:

Buyers want AI systems that can plug into the software environment they already have, not force a rip-and-replace architecture just to get automation.

Why this is commercially important

Most businesses do not have a "we need more chat" problem.

They have a workflow problem.

The painful work usually sits between systems:

• an email arrives and someone rekeys data into the CRM
• a document gets downloaded, renamed, and routed for review
• an invoice gets checked against policy and manually pushed into ERP
• a lead gets enriched in one tool, scored in another, and assigned in a third

This is exactly why interoperability matters.

If AI can securely move through those systems, gather context, apply rules, update records, and escalate exceptions, it can own the workflow.

If it cannot, then the company still needs a human to glue the stack together.

That means the ROI ceiling stays low.

What enterprise buyers should infer from this trend

The next generation of serious AI vendors will look less like standalone software products and more like operational layers.

They will still use strong models. They will still need good UX. But the differentiator will increasingly be:

• how quickly they can connect to your environment
• how safely they can act inside your systems
• how well they handle multi-system workflows
• how clearly they support approvals, permissions, and auditability
• how little organizational change they require from your team

This is good news for buyers.

It means the market is moving away from "adopt our whole platform" and toward "let us automate the work that is already happening in your stack."

That is a much easier purchase to justify to operations leaders, finance leaders, and IT teams.

What to avoid

There is a buyer mistake hiding here too.

Some companies will hear "agents are interoperable now" and assume that means any workflow can be automated safely.

That is still not true.

Open protocols help with connectivity. They do not solve workflow design, exception handling, permissions, monitoring, or ownership.

A bad workflow with better plumbing is still a bad workflow.

So when you evaluate vendors, do not stop at:

• Which model do you use?
• Which protocols do you support?
• How many connectors do you have?

Ask the harder questions:

• What unit of work do you fully complete?
• What systems do you actually operate inside today?
• What happens when confidence is low or data is missing?
• Who monitors and fixes the workflow after launch?
• How do you keep actions bounded and auditable?

Interoperability makes automation easier to deploy. It does not remove the need for operational rigor.

Why this favors outcome-based automation

This trend also reinforces why we think outcome-based automation is the right commercial model.

If the technology is becoming more interoperable, buyers should not have to pay for another layer of shelfware just to experiment.

They should be able to buy the completed work:

• a routed lead
• a verified document
• a processed invoice
• a completed onboarding step

That is the cleanest way to align incentives in an interoperable market.

The vendor has to make the workflow work across the real environment, not just sell access to software that still depends on internal teams to finish the job.

The practical takeaway

The important AI trend is not simply that models got better.

It is that the market is building the connective tissue that makes agentic systems more practical inside real companies.

That changes the buying posture.

You no longer need to ask, "Which AI platform should we roll out company-wide?"

The better question is:

"Which workflow in our current stack is now realistic to automate because the tooling, protocols, and governance are finally catching up?"

That is where new budget gets justified. That is where authority gets built internally. And that is where AI starts to look less like innovation theater and more like operating leverage.

Sources

• Microsoft, "The 2025 Annual Work Trend Index: The Frontier Firm is born"
• OpenAI, "Developer mode, and MCP apps in ChatGPT [beta]"
• Google Cloud, "Announcing Model Context Protocol (MCP) support for Google services"
• Google Cloud, "Agent2Agent protocol (A2A) is getting an upgrade"

If you want to see which workflow in your stack is worth automating first, run the calculator or book a workflow audit.