Konnect Insights Built Its Front Office on Model Context Protocol. That Cuts Both Ways.
Only 12.3 per cent of SaaS products ship verified official MCP support, and the analysts tracking the standard warn that connectivity commoditises the layer above it.
Konnect Insights expanded Front Office OS on 28 October, a layer that sits above a customer-facing stack and connects social, contact centre, CRM, marketing automation and field service systems into one place where a leader can ask a question and act on the answer. The architecture is the interesting part. Rather than building proprietary connectors, the company wired the platform to the Model Context Protocol, the open standard Anthropic published in November 2024. That is a defensible choice, and it is also a choice that makes the company’s own position harder to defend.
Both statements follow from the same property of the standard: MCP makes every system equally reachable, including the ones whose data Konnect Insights currently sits on top of.
What Front Office OS actually is
The product has three verbs. Ask lets a leader query in plain language across social, contact centre, CRM and campaign data. See pushes prebuilt intelligence to named roles through a component called Konnect Pulse. Act, through KRC Triggers, monitors connected data and executes across systems: an airline spotting disrupted passengers in calls and social posts and sending them lounge offers on WhatsApp, or a bank pausing promotional messages to customers with unresolved complaints.
The company describes a three-layer architecture, and the description is precise enough to be worth setting out. The systems layer holds the existing platforms unchanged, each publishing a connector that declares what it can return and what it can do. The connection layer is MCP. The intelligence layer is where the chief executive, marketing chief and customer experience head operate. Ownership, permissions and licensing stay where they are. Konnect Insights does not replace Genesys, Salesforce or MoEngage; it reads from all of them and acts through all of them.
Worth being precise about what the standard carries, because the answer explains both the opportunity and the risk. The Model Context Protocol defines how a server describes the tools it offers, what inputs each expects and what it returns, and how a client discovers and calls them. It does not define what those tools mean, what data model sits behind them, whose permissions apply at the point of execution, or what an acceptable action looks like in a given business. Those remain vendor and customer decisions. That division is why a small company can connect to Genesys and Salesforce in weeks, and also why the resulting connections are only as governed as each endpoint allows.
Why MCP is the right bet
The case for the Model Context Protocol is arithmetic. Without a common protocol, an intelligence layer above a multi-vendor stack is a custom integration project that never finishes, and adding the tenth system costs roughly what adding the first did. With one, integration effort grows linearly rather than quadratically, which is the point Boston Consulting Group made when it called MCP a deceptively simple idea with outsized implications. Organisations implementing it report 40 to 60 per cent faster agent deployment.
The governance has also matured faster than most standards manage. Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation, in December 2025, with OpenAI and Block as co-founders and AWS, Google, Microsoft, Bloomberg and Cloudflare as platinum members. By March 2026 the protocol had produced 97 million monthly SDK downloads and more than 10,000 active production servers, with roughly 150 members steering it.
The adoption numbers are smaller than the noise
It is worth separating the standard’s momentum from vendors’ actual implementations, because the gap is wide. Cyclr tracked 816 SaaS products and screened 23,830 entries in the official MCP Registry, comparing 2025 with 2026. Products with verified official MCP support rose from 47 to 100, which is a doubling and also a share of just 12.3 per cent. Fifty-seven per cent of the businesses in the sample had no verified official MCP server at all.
Stacklok’s survey of 300 senior technical leaders, conducted in December 2025, is consistent with that picture. Only 12 per cent reported broad production use; 26 per cent sat in limited production, 30 per cent in pilot and 26 per cent still planning. Nearly half ranked MCP adoption as a top-five technology priority, which tells you about intent rather than deployment. Both samples skew toward early adopters, so the true market figure is likely lower still.
The commoditisation problem
Here is the difficulty with building a business on a standard, and the analysts covering it have already said so plainly. Verdantix’s assessment is that as MCP connectivity commoditises, value shifts from connection to governance and action, that MCP support becomes a baseline buyer expectation rather than a differentiator, and that vendors face a growing risk of value disintermediation as agentic harnesses become the primary interface to enterprise software.
Read that against Konnect Insights’ own architecture and the tension is visible. If the Model Context Protocol makes every system reachable by every intelligence layer, then the layer above becomes easier to build for everyone, including the platforms underneath it. Salesforce runs MCP in Agentforce 3. Microsoft exposes Dataverse as an MCP server to Copilot Studio. Google supports it natively across Cloud Run, Cloud SQL, Spanner and Workspace. Avaya ships it in production in Infinity with Databricks governance and sits on the foundation’s steering body. Konnect Insights’ defence is the thing MCP cannot standardise: the cross-brand benchmark data, the public social corpus no internal system sees, and an evidence base accumulated across 500 brands.
The customer experience category is where this will be tested first, because it is unusually fragmented and unusually exposed. A typical stack runs a contact centre platform, a CRM, a marketing automation tool, a social listening service and a field service system from five different vendors, none of which sees the whole customer. That is the gap Konnect Insights addresses, and it is a real one. It is also the gap each of those five vendors can see, which is why Avaya built MCP support into Infinity and why Salesforce put it in Agentforce.
The security barrier
The other constraint is the one every MCP deployment meets. In Stacklok’s survey, security concerns topped the obstacle list at 64 per cent, ahead of implementation cost at 40 per cent and legacy integration complexity at 38 per cent. The measures enterprises apply are conventional and telling: role-based permissions at 62 per cent, audit logging at 55 per cent, zero-trust access control at 53 per cent.
The standard is addressing this directly. The Enterprise-Managed Authorization extension became stable in June 2026, letting an identity provider provision MCP server access centrally so users inherit connected servers on first login, with Okta as the first supported provider and Anthropic and Visual Studio Code as clients. That solves the repeated consent prompt, which was the most visible friction. It does not solve the harder question of what an agent should be permitted to do once it is inside.
What Konnect is not saying
Three claims need substantiation before anyone plans around them. The digital war room within the first week is an implementation claim with no named customer. The on-demand quarterly board pack is a striking capability described in a single clause. And the commitment not to train AI models on customer data is exactly the right promise and exactly the kind that should be contractual rather than promotional. The release names no customer, publishes no outcome metric and gives no pricing for a layer that sits above systems the customer already licenses.
What to watch
Three markers. The first is whether official MCP support becomes a procurement checkbox in 2027, which Gartner’s projection that 75 per cent of API gateway vendors and 50 per cent of integration platform vendors will support it natively suggests is already happening. The second is whether the platforms underneath start charging for the privilege of being connected, since a commoditised connector layer invites tolls. The third is whether Konnect Insights publishes customer outcomes, because a benchmark corpus is only a moat if someone can see what it produced. A fourth, less visible, is whether the foundation’s roadmap delivers the audit trail enterprises need: its March 2026 plan named audit trails, SSO-integrated authentication and configuration portability as the predictable enterprise problems, which is an unusual admission for a standards body and a useful signal of where the friction sits.
Three years ago the vendor that owned the connector owned the customer, and everyone selling enterprise software knew it. A custom integration into a bank core was a moat, because it took eighteen months and nobody wanted to build it twice. That moat is gone, and the companies that built businesses on it now compete on what survives its removal: proprietary data, accumulated benchmarks, and the quality of the judgement applied to them.
What it signals
The expansion marks a stage where a mid-sized vendor in Mumbai builds on the same open standard as Microsoft, Salesforce and Google, and does so credibly. That is new, and it is what a genuine standard looks like: not a vendor alliance but a floor everyone stands on. The strategic question it raises is the one every layer-above-the-stack company now faces. Being reachable is no longer the product. What you do with the reach is, and that is a harder thing to describe in a release than a connector.