Since its inception, we all predicted that AI was going to shake up telco retail – we just didn’t foresee quite how much.
For the last two years, the industry has been preparing for AI to make our existing software smarter - better recommendations inside the POS, sharper forecasting inside the inventory tool, slicker insights inside the merchandising dashboard. The assumption was that AI would be a feature layered into the applications we already use.
But with the emergence of Model Context Protocol (MCP), it’s becoming clear that AI isn’t just a software upgrade – it will now be the primary orchestrator of daily tasks, ushering in the era of true AI Retail Operations.
In this edition of Telco Talk, we’ll dive into the implications of MCP, what AI Retail Operations look like, and how telco leaders must prepare for this transformational change.
The Emergence of MCP
For the last two years, "AI in telco retail" has mostly meant suggestion. Recommend a plan. Summarize a ticket. Draft a response. The agent could talk about the work, but it couldn't do the work - because every BSS, POS, inventory, and contract system had its own integration model, and stitching agents into them was a custom job every time. The AI knew what should happen. It just couldn't make it happen without a human in the middle.
That's the bottleneck MCP breaks. Introduced by Anthropic in late 2024 and rapidly adopted by OpenAI, Google, Microsoft, and Amazon, MCP is an open standard often called the "USB-C for AI" - a universal way for agents to discover and use the tools, data, and actions of any system that supports it.
In telecom, TM Forum has declared MCP a foundational requirement for AI-driven BSS automation, and industry providers are adding MCP support to their roadmap.
Translated into retail terms, MCP exposes three things:
- Tools: Actions like "create order," "reserve stock," "issue trade-in credit".
- Resources: Data like customer profile, store inventory, and plan eligibility.
- Prompts: The guardrails that keep an agent inside policy.
Any agent that speaks MCP can use any system that speaks MCP. The connector tax disappears. That's the moment AI stops being a co-pilot and becomes the operator.
What AI-Driven Retail Operations Actually Look Like
Once the platform is MCP-native, AI is now where the work happens. Management and operations functions move inside the agent, with humans approving the decisions that matter.
Five retail operations get rebuilt first:
- Replenishment: The AI agent examines POS sell-through data, promo calendars, and supplier lead times and builds out appropriate stock transfers and POs. Now, the merchandiser reviews the plan instead of building themselves.
- Promotion & Offer Design: AI agents generate new tariffs, bundles, and packages from natural-language briefs - launching complex promotions in hours instead of weeks.
- New Product Launches: The typical weeks long process to launch new products simply becomes an agent-orchestrated workflow that the launch team approves.
- Pricing & Margin: Agents continuously monitor competitor pricing, margin erosion, and inventory aging, so they can propose changes for the analyst to sign off on.
- Customer Journey Orchestration: Complex sales, such as those simultaneously involving a trade-in, number porting, plan changes and home internet add-ons, can be assembled in plain language – and in real time - from any channel.
The Foundation: Integration Is The Real Key
Here's the part most AI conversations skip: agents are only as good as the platform underneath them.
An agent that can't see real-time inventory across stores can't replenish properly. An agent that can't read a customer's full account context can't orchestrate a journey. An agent that can't write back to the catalog, the contract engine, and the order system in one transaction can't launch a product. Point AI at a fragmented, partially-instrumented stack and you'll get fragmented, partially-correct results - at machine speed.
This is why the winners of AI-driven retail won't be the carriers with the flashiest AI demos. They'll be the carriers whose platforms were already doing the unglamorous work: API-first architecture, a unified data model, consistent customer and product objects across POS, inventory, contracts, and self-care, and orchestration that treats the customer journey as a first-class entity instead of a sequence of disconnected screens.
MCP is what lets the agent act. Integration is what lets the agent act correctly. The order matters.
Why Human-in-the-Loop Is an Architecture, Not a Feature
The right model for AI-driven retail operations isn't "AI replaces the merchandiser." It's "the agent assembles, the human approves." That sounds like a process choice, but it's actually an architectural one.
Done right, governance lives in the platform layer between the agent and the system. The agent uses the same APIs every other application does - just through an interface designed for it. That's how you stay compliant with PCI for payments, KYC for activations, e-signature law for contracts, and GDPR/PIPEDA for customer data, even as agents take on more of the work.
Time To Take Stock
Are you ready for AI Retail Operations? Here are three questions every telco retail leader should be asking this quarter:
- Which systems in our retail stack expose MCP servers today, and which have a roadmap? (e.g. POS, inventory, contracts, self-care, kiosks, WMS, etc.)
- Is our data and integration foundation actually ready for an agent to act on it - one customer, one product, one inventory truth across channels?
- Where do we want the human in the loop, and where don't we?
The carriers who answer those questions in 2026 will be running a fundamentally different retail business by 2028 - one where merchandisers, buyers, analysts, and store ops leads no longer operate software, but direct agents that operate the business for them.
Final Thoughts
AI Retail Operations isn’t some far-fledged dream. The protocol is here, and the use cases are real and deployable today. The question isn't whether the action moves out of the software and into the agent. It's whether your platform is ready to let it.
That's the future of telco retail. And it's already being built.
