[← All posts](/blog)
For most of the last decade, analytics was a tab. Your customer ran their business in the app, and once a week they clicked into “Reports” to see how it went. The dashboard was a courtesy, bolted on after the core workflow shipped.
That model is ending. In the last twelve months, vertical software companies kept making the same move without quite saying it out loud: analytics is not a feature anymore. It is the product, and it is becoming the main way customers interact with the application.
Three companies making moves in analytics #
In October 2025, Mews, the hospitality operating system used by 12,500 hotels across 85+ countries, acquired DataChat, a generative analytics company. Mews did not buy a reporting tool. It bought a way for hotels to ask questions of their data in plain language and act on the answers. Founder Richard Valtr put it plainly: hotels generate enormous data from guest interactions, and “the true opportunities lie in unlocking that data to power smarter, more intelligent operations.”
A few months later, Procore, the construction management platform, acquired Datagrid, an agentic AI company, to “eliminate data silos” and surface insights across a customer’s entire technology stack. Same logic, different vertical. The company that already owns the workflow is buying the layer that turns the workflow’s exhaust into answers.
And in ecommerce, Polar Analytics has built a business doing exactly this from day one. Polar is an ecommerce-specific analytics platform with hundreds of pre-built metrics (CAC, LTV, ROAS), used by 4,000+ brands including Allbirds, Volcom, and Jones Road. It is not a dashboard brands check on Fridays. It is where they decide what to buy, who to target, and what to email, with reported outcomes like 36 to 50% lower CAC and hundreds of hours saved on reporting. Analytics is the wedge, the workflow, and the value, all at once.
This is not three isolated bets. Analytics and data management was one of the two most active categories in SaaS M&A in 2025 (per Software Equity Group), driven by buyers who want intelligence embedded in the core workflow rather than stitched on afterward.
The interface is already moving to agents #
If the value is moving into the data, the interface is moving with it. In the last few months, vertical platforms have shipped ways for AI agents to reach their data directly. Shopify rolled native MCP support into every store, so an agent can read and act on live product, order, and customer data. Procore is building Agentic APIs on top of the Datagrid infrastructure it just bought. Veeva launched AI Agents across its life sciences cloud, built to interoperate over MCP. The horizontal giants are moving too, with Salesforce making its hosted MCP servers generally available and HubSpot’s letting an agent pull pipeline reports. Anthropic, which created the protocol, reports more than 10,000 active public MCP servers as of early 2026. MCP (the Model Context Protocol) is just the plumbing, but the direction it points is the whole story. Customers are starting to interact with applications by asking an agent, not by clicking through six dashboards. Toast is pushing the same idea inside restaurants: surface the trend, suggest the menu change, instead of waiting for an operator to go find it. When that becomes the interaction model, analytics stops being a screen. It is the starting point of your customers creating value.
Why analytics became the pillar #
When something this consistent happens across hospitality, construction, and ecommerce in the same year, it is worth asking why. We see three reasons, and they map cleanly onto what customers actually want.
1. Customers want their data inside their own workflows
The old deal was: we’ll show you a chart. The new deal is: give me the number so I can act on it where I already work. Hotels want occupancy and rate intelligence feeding their pricing decisions. Brands want CAC and inventory signals feeding their ad spend and reorders. The value is not in viewing the data. It is in moving it into the decision. DataChat’s CEO described the goal as agents that “understand intent, reason across data sources, and act autonomously.” An app that lets customers access and understand their own data, then carry it into their own workflows, becomes part of how they run the business instead of a place they visit.
2. Analytics gives you far more surface to demonstrate value
A workflow tool proves its worth when someone uses it. An analytics layer proves its worth every time someone asks a question, and there are far more questions than there are workflows. Each metric a customer cares about, each benchmark, each “how am I doing versus last quarter,” is another surface where the product shows up and earns its keep. That is why platforms that already own the workflow keep reaching for the analytics layer. It multiplies the number of moments where the product is visibly valuable, and it raises switching costs at the same time.
3. Analytics is your best product-feedback loop
This is the one people underrate. When customers explore their own data through your app, they tell you what they care about without filling out a survey. The metrics they build, the questions they ask repeatedly, the cuts they save, these are a live signal of what matters to them and where your product should go next. An embedded analytics layer is not just a way to deliver value. It is a way to learn what value to build, straight from how customers interact with their numbers.
Agentic analytics only works on a governed foundation #
Here is the part that does not show up in the press releases. Making analytics the core of a vertical app is hard, and the hard part is underneath.
The moment you let a customer (or an agent acting for them) ask open questions of their data, you inherit a governance problem. Agents writing raw SQL against schemas they do not fully understand hallucinate metrics. The definition of a core number drifts, and “revenue” or “active customer” quietly means three different things depending on who asks. A dashboard with one wrong number is an embarrassment. An agent that confidently returns wrong numbers, inside the workflow, at the moment of decision, is a liability.
So the bet every one of these companies is really making is on the layer between the app and the data: a place where every metric, dimension, and piece of business context is defined once and enforced everywhere, so that whatever sits on top, an MCP server, a dashboard, an agent, an embedded report, sees the same governed numbers. That is the unglamorous infrastructure that makes embedded analytics trustworthy enough to be a core pillar instead of a liability.
Building it is not trivial: tenant isolation, user management, row-level security, and maintenance of semantic definitions are some of the key problems developers face when building agentic analytics into their app. The pattern is consistent: agentic analytics and the semantic layer underneath it are being treated as one thing, because the first does not work without the second.
Where we sit #
This is the layer we build at Motley. SLayer, our open-source semantic layer, lets a team define metrics and context once on top of their warehouse, then expose them to internal agents, AI clients, and the apps they ship.
With MCP servers, the agent is whatever the customer is already using, Claude, Cursor, an embedded assistant inside a vertical app. Our job is to make that agent trustworthy on the customer’s data, so a vertical app can put analytics at the center of its product without betting the company on a model guessing what “ARR” means.
Mews, Procore, and Polar are early signals of a shift that is going to run through most of vertical software. The companies that win will be the ones whose analytics customers actually live in, and that only holds up on a foundation where every number means one thing. The feature era of embedded analytics is over. The product era is just starting.