Somewhere in your company right now, someone is copying a number out of one system and pasting it into another. Maybe it’s an invoice total moving from your accounting software into a CRM note. Maybe it’s a customer address that has to be typed into three different places because none of your systems talk to each other automatically. This kind of quiet, manual stitching-together of software is one of the most expensive things a growing business does, and almost nobody puts a line item on it.

Over the past few weeks, a cluster of announcements from some of the biggest names in business software has made it clear where the industry thinks that problem gets solved next. The short version: nobody serious is trying to sell you a replacement for your ERP or CRM. Instead, they’re racing to build a smarter, safer front door into the systems you already have, one that lets AI agents reach in, do useful work, and leave an audit trail behind them.

AI agents are moving in, not tearing down

The clearest example landed in early September, when Stuut, an accounts-receivable automation company, deepened its relationship with Microsoft to bring autonomous order-to-cash agents to enterprise customers through the Microsoft Marketplace. What’s notable isn’t the funding or the marketplace listing — it’s the design choice. Stuut’s agents handle collections outreach, cash application, and dispute resolution, but they do it “within the customer’s existing ERP stack rather than replacing it,” working alongside SAP, Oracle, NetSuite, and Microsoft Dynamics 365 rather than asking a finance team to rip and replace anything.

That pattern — agent on top, existing system underneath, left untouched — is showing up everywhere integration vendors are talking right now. It’s a quiet admission that the ERP and CRM your business already runs on isn’t going anywhere. What’s changing is who, or what, gets to interact with it.

The plumbing is having its moment

The less visible but arguably bigger story is happening one layer down, in the connective tissue between systems. In mid-August, MuleSoft (Salesforce’s integration arm) shipped a round of tools — Agent Scanners, MCP Bridge, and Agent Broker — built specifically to take existing APIs that were designed for human developers and make them discoverable and usable by AI agents “in just a few clicks,” without rewriting the underlying services. Around the same time, Kong pushed out an update extending its AI Gateway to govern agent-to-agent traffic alongside ordinary API calls, and Apigee expanded its ability to auto-generate a governed AI access point directly from an existing API, with no new code required.

All of this activity centers on a technical standard called MCP, short for Model Context Protocol, which has quickly become the common language AI tools use to ask enterprise systems for data or to trigger an action. You don’t need to understand MCP’s mechanics any more than you need to understand TCP/IP to use the internet. What matters is what it enables: instead of every software vendor building a custom, one-off connection to every AI assistant, systems can expose a single, governed doorway that any properly authorized agent can use. That’s the difference between hiring a locksmith for every new employee and issuing one badge that opens the doors they’re allowed to walk through.

Governance just became a board-level conversation

The clearest sign that this has moved beyond an IT concern came on September 3, when UiPath restructured its executive team and added Yazdi Bagli — Kaiser Permanente’s Executive Vice President of IT and Enterprise Business Services — to its board of directors. Bringing in a veteran enterprise IT operator, rather than a purely financial or strategic appointment, signals that the questions boards want answered about automation and integration are shifting from “does it work” to “can we govern it, audit it, and trust it at scale.”

That shift matters for businesses far smaller than the Fortune 500 companies these vendors are chasing. Every one of the announcements above is really about the same underlying worry: once you let software agents reach into your financial and customer systems, who’s watching what they do, and can you prove it later if a regulator, auditor, or customer asks?

What this means if you’re running a growing business

You don’t need a Fortune 500 integration budget to benefit from this shift, but you do need to ask the right questions before you sign anything. A few worth putting to any software partner or vendor pitching you an “AI-powered” integration:

• Does this tool work with the ERP, CRM, or accounting system I already have, or is it quietly asking me to migrate everything to a new platform to get the AI features?
• What can I see, after the fact, about what an AI agent actually did — which records it touched, what it changed, and why? If the answer is “not much,” that’s a real risk, not a minor gap.
• Who is responsible if an agent makes an incorrect change to customer or financial data — the vendor, the integration layer, or us? Get that answer in writing before go-live, not after an incident.

If your current systems have decent APIs, there’s a good chance you can get real value from agent-based automation without a rebuild. If they don’t, that gap — not a lack of AI features — is probably your actual bottleneck.

The takeaway

The businesses that get the most out of this next wave won’t be the ones that buy the newest AI-branded software. They’ll be the ones that already have clean, well-governed connections between the systems they run their business on, so that when a useful agent comes along, plugging it in is a configuration decision instead of a six-month project. If you’re not sure how well-connected your own stack actually is, that’s a worthwhile conversation to have before you’re pressured into one by a vendor demo.

If you’d like a second opinion on where your systems stand, or want help thinking through how AI agents could fit safely into your existing ERP or CRM, reach out to us at info@kodevox.com or visit our contact page. We’re happy to talk through what makes sense for your business, with no obligation attached.

Sources and further reading:

• Stuut Deepens Microsoft Relationship to Bring Autonomous Order-to-Cash to Enterprise Customers
• MuleSoft ’26 Mid-Year Release: Create AI Experiences
• Introducing Kong’s Enterprise MCP Gateway for Production-Ready AI
• UiPath Announces Executive Leadership Changes, Appoints New Board Member

— The Kode Vox Team