Alkemi Wants to Be the Data Scientist Your Slack Already Needs

The Macro: Everyone’s Drowning in Data Requests They Can’t Actually Answer

Here’s a thing that happens at basically every company above 20 people: someone in sales or marketing or ops posts a question in Slack, a data person either answers it three days later or ignores it entirely, and meanwhile a decision gets made on vibes. This isn’t a people problem. It’s a tooling problem. Data tools are built for analysts. Everyone else is just guessing.

The fix everyone keeps reaching for is “make data accessible to non-technical users.” The graveyard of products that tried this is enormous. Tableau, Looker, Mode, Metabase. They’re all powerful. They’re all also kind of annoying to actually use if you’re not someone who enjoys setting up dashboards on a Sunday afternoon.

The newer generation is going AI-first. You’ve got tools trying to put a natural language layer on top of warehouses, BI tools bolting copilots onto their existing interfaces, and a bunch of smaller players building purpose-built agents. The pitch is always some version of “just ask your data a question.” The execution gap between that pitch and reality is where companies keep tripping.

Slack is the interesting wedge here. According to multiple sources tracking the platform’s growth, Slack had roughly 38.8 million daily active users in 2024. That’s a massive captive audience of people who are already having work conversations. If you can make data answers show up in the same thread where the question was asked, you’ve removed one of the biggest friction points in the whole workflow. The context is already there.

That’s the actual opportunity Alkemi is chasing. Not another dashboard. A data presence that lives where the conversation already is. Whether any of these tools actually deliver on that at enterprise scale is a genuinely open question, and I’ve been skeptical of a few that claimed to. Superset’s approach to managing AI agents is a useful reference point for how fast this category is moving.

The Micro: A Data Analyst That Lives in Your Slack Threads

Alkemi’s core product is a Slack-native AI agent that connects to your company’s actual data sources and answers questions in plain English. You ask something like “what drove pipeline last week” and it returns a chart or summary directly in the channel. No context switching, no ticket to the data team, no waiting until Thursday.

Under the hood, the product connects to Snowflake, Google BigQuery, and Databricks, according to their MCP server listing on GitHub. That’s a deliberate choice targeting companies that are already operating at some level of data maturity. They’re not going after the Google Sheets crowd. This is for teams that have a warehouse and just can’t get value out of it fast enough.

The flagship product they’ve built is something called DataLab, described as a secure AI-native workspace that connects to governed data from those same warehouse sources. The Slack agent appears to be the conversational frontend sitting on top of that infrastructure. So the intelligence isn’t just a wrapper around a generic LLM. It’s (reportedly, based on how they describe it) working with structured, governed data products rather than just prompting against raw tables.

That distinction matters more than it sounds. Governance is the thing enterprise customers actually care about. Any AI tool touching financial or customer data that can’t answer “how do you know this is accurate” is going to get blocked by IT before it ever reaches the people who wanted it.

It got solid traction on launch day, which suggests there’s real demand for this kind of thing from the people paying attention to this space.

The question I’d push on is latency and reliability. Conversational BI sounds great until the query takes 45 seconds and the chart comes back wrong. Connor Folley, co-founder and CEO according to LinkedIn, has framed Alkemi’s mission as bridging the gap between data and decisions. The bridge metaphor is doing a lot of work there. Bridges have to hold weight consistently, not just on good days.

The Verdict

I think the Slack-native angle is genuinely smart, not just as a distribution play but as a product philosophy. The best enterprise tools are the ones that meet people where they already are. Forcing behavior change is the fastest way to get abandoned by a company that paid for an annual contract.

The warehouse-first approach (Snowflake, BigQuery, Databricks) narrows the addressable market but probably increases the chance of actually working well. That’s a trade-off I respect more than trying to boil the ocean.

What would make this work at 30 days: real usage in actual Slack workspaces with evidence that non-analysts are the ones asking the questions, not just data teams validating outputs.

What would make this fail: if the answers are wrong often enough that people stop trusting it. One bad number in a sales meeting and this tool gets turned off. Trust is the whole product.

I’d also want to understand the competitive moat here. The same way security-focused AI tools need to demonstrate defensibility, Alkemi needs to show why a well-resourced BI incumbent can’t just ship a Slack integration and eat their lunch.

The idea is good. The execution is where this either becomes something people rely on or becomes a neat demo that nobody actually uses.