Slack's AI Strategy: How the Messaging Leader Listens to Its Own Users in 2026

TL;DR#

Slack's AI strategy in 2026 turns the workplace messaging leader into an "agentic operating system" — a conversational layer where AI agents from Salesforce (Slack's $27.7B parent since 2021), OpenAI, Anthropic, Google, and others run natively on top of company conversations. Slack reaches roughly 47.2 million daily active users across 750,000+ organizations and 77% of the Fortune 100, and more than 1.5 billion messages cross the platform every day. Slack AI launched in February 2024, and by Dreamforce 2025 the company had shipped channel summaries, AI search, huddle notes, a rebuilt Slackbot, and a real-time search API with Model Context Protocol support. Yet the company that turned async conversation into a category still learns what its own users want largely through in-app surveys, NPS prompts, and feature-request forms — instruments that flatten the very conversation Slack champions. This is the core tension in Slack's product feedback approach: a conversation company researching its customers with checkboxes. Conversational AI interviews, the approach Perspective AI is built on, close that gap by letting users explain the "why" behind a request in their own words.

What is Slack's AI strategy in 2026?#

Slack's AI strategy is to make Slack the conversational home for enterprise AI — the place where agents live, get context from company conversations, and act on work without the user leaving the channel. CEO Denise Dresser framed it bluntly at Dreamforce 2025: "It is not that we work 'in' Slack, but that Slack works 'for' us." Rather than bolt a chatbot onto a chat app, Slack repositioned the entire product as an agentic operating system that unifies CRM data, third-party agents, and company knowledge behind one conversational interface.

That ambition rides on a platform most workplaces already live in. Slack serves an estimated 47.2 million daily active users and 79 million monthly active users, spread across more than 750,000 organizations, with 77% of the Fortune 100 on board, according to 2025 Slack statistics compiled by DemandSage. More than 1.5 billion messages move through the platform daily. When a company with that footprint declares AI is the future of its product, the strategic question stops being "should we use AI" and becomes "how do we learn fast enough to build the right AI." That second question — how Slack does customer research and product feedback at this scale — is where the story gets interesting.

How Slack uses AI today#

Slack uses AI today through a stack of native features built to compress the noise of high-volume messaging into action, plus an agent platform that lets outside AI tools operate inside Slack. The progression from 2024 to 2026 shows a deliberate escalation.

• Slack AI (February 2024). Slack brought generative AI features to enterprise customers, then extended them to Pro and Business+ plans as a $10-per-user-per-month add-on, per Computerworld's coverage. Core capabilities: channel and thread summaries, AI-powered search that answers questions in natural language, and daily recaps.
• AI search, huddle notes, transcriptions (2025). Slack expanded AI search to pull from files and connected apps, and added huddle notes that auto-generate a canvas of takeaways, action items, and transcripts when a call ends, as TechCrunch reported in July 2025.
• Agentic OS and the agent platform (Dreamforce 2025). Slack opened a real-time search (RTS) API and a Model Context Protocol (MCP) server so partners — OpenAI, Anthropic, Google, Perplexity, Writer, Notion, Cursor, and others — could build agents that run natively on Slack's conversational data.
• Slackbot reborn (early 2026). Slack rebuilt Slackbot from the ground up as a personalized AI companion that draws on a user's messages, files, and calendar across Slack and connected systems like Google Drive and Salesforce, shipping 30+ new capabilities.

The connective tissue is Salesforce. Salesforce acquired Slack for $27.7 billion in 2021, and by the fourth quarter of fiscal 2026 it reported 2.4 billion "Agentic Work Units" delivered across Agentforce and Slack — tasks completed by AI agents — growing 57% quarter over quarter, according to Salesforce's FY2026 Q4 results. Slack is no longer just a chat app inside Salesforce; it's the front door for Salesforce's "digital labor" thesis.

Why customer research is the hidden bottleneck in Slack's AI strategy#

Customer research is the hidden bottleneck in Slack's AI strategy because the company is making bigger product bets — agents, an agentic OS, a rebuilt Slackbot — than ever before, while its primary listening instruments still compress users into structured fields. The faster a product ships AI, the more it depends on understanding intent, and intent is exactly what surveys lose.

Consider how a product organization at Slack's scale typically gathers signal. There are in-app pulse surveys ("How satisfied are you with the new search?"), NPS prompts that produce a 0–10 score, feature-request forms and upvote boards, and analytics on what gets clicked. Each instrument captures something real. None of them captures the sentence that actually changes a roadmap: "I turned off channel summaries because they kept surfacing the wrong thread, and now I don't trust the AI to know what's important." A five-point scale records dissatisfaction. It never records the reason — and the reason is the roadmap.

This is the pattern Perspective AI describes in detail in why AI-first research cannot start with a web form: forms front-load effort, flatten nuance into dropdowns, and fail precisely at the messy, high-value moments where a user says "it depends." The same failure mode shows up in why conversations beat surveys for real customer research. When the product you're researching is AI you've never shipped before, you don't have a clean schema to survey against yet — you need to discover the schema. That's a conversation, not a checkbox.

There's a specific irony here. Slack's entire category thesis is that conversation beats forms and email — that work flows better when people talk in threads instead of filling out static artifacts. Then, to learn how those same people use the product, Slack reaches for the static artifact. The conversation company researches with the form.

The conversation company that researches with checkboxes#

Slack's product feedback challenge is structural, not accidental: at 1.5 billion messages a day, the temptation to summarize users with metrics is overwhelming, but metrics tell you what changed, not why. A team can watch huddle-notes adoption dip after a release and know something is wrong. Knowing what is wrong — the UI confused power users, the summaries hallucinated, the feature buried a workflow people relied on — requires asking, listening, and following up. At Slack's scale, asking and following up by hand doesn't work; there aren't enough researchers in the world to interview a meaningful slice of 47.2 million daily users.

This is the exact gap conversational AI research is built to close, and it's why the discipline of voice-of-customer programs in 2026 is shifting from survey aggregation to conversation at scale. A few precise contrasts:

The first three rows describe how most large software companies, Slack included, hear from users today. The fourth row is what good qualitative research has always delivered and never scaled. The fifth row — AI interviewers that ask open questions, probe vague answers, and run hundreds or thousands of conversations at once — is the row that didn't exist as a product category until recently. Perspective AI's AI interviewer agent sits in that fifth row, and its concierge agent replaces the intake form itself with a conversation, so the very first touch captures context instead of dropdowns.

What conversational AI research unlocks for a product like Slack#

Conversational AI research unlocks the missing "why" at the scale Slack actually operates — turning a 0–10 NPS score into a transcript that explains what would make it a 9, and turning a feature-request count into an understanding of the job behind the request. For an AI-first product roadmap, that shift is decisive, because you cannot ship the right agent if you only know that users are unhappy with the current one.

Concretely, here's how a Slack-scale product team could use it:

1. After every major AI release, trigger an AI interview to a sample of active users instead of an NPS prompt. The interviewer asks what they tried, what worked, where they got confused, and what they expected the AI to do that it didn't — following up on each answer the way a good researcher would.
2. For feature requests, route submitters through a conversational intake that asks "walk me through the last time you needed this" rather than a text box. You learn the job-to-be-done, not just the feature label.
3. For churn and downgrade signals, run exit and de-adoption interviews that capture the real reason a team stopped using a feature — the input that no dashboard surfaces.
4. For continuous discovery, maintain an always-on study so the roadmap is informed by a steady stream of conversations, not a quarterly survey blast.

This is the same move other category-defining software companies are making as they navigate AI bets. The same conversation-at-scale problem shows up in how Notion decides what to build, how Asana researches its roadmap, and how Zendesk listens to support teams. It's also visible across the broader work-management field — monday.com's voice-of-customer approach and ClickUp's research practice wrestle with the same depth-versus-scale tradeoff. Slack is simply the highest-profile example, because its product is conversation and its research, historically, is not.

For teams evaluating where this fits in their toolset, the modern stack is mapped in the customer research tools modern product and CX teams actually use, and the buyer's lens for AI-era listening platforms is laid out in the AI customer engagement software buyer's framework. Slack's own product teams and CX teams are the exact personas these workflows serve.

The 2026 context: why "Slack AI" raises the stakes on listening#

The 2026 context makes customer research more urgent for Slack, not less, because the company is now betting its product identity on AI that has to feel native to how its users already work — and the only way to know whether it does is to ask them in depth. When Slack was a messaging app, a missed feature was an inconvenience. When Slack is the agentic operating system where 750,000 organizations expect their AI to live, a misread of user intent compounds across every agent, every summary, and every automated action.

Two industry signals raise the stakes further. First, the talent and competitive picture is volatile: Slack's CEO Denise Dresser left for OpenAI's revenue leadership in December 2025, as CNBC reported, underscoring how fiercely the enterprise-AI front is being contested. Second, the entire customer-research layer is being rebuilt around conversation rather than the survey, a shift Perspective AI tracks in its analysis of what's replacing the survey layer in 2026 and in the case for an AI survey alternative. A company whose whole thesis is that conversation beats forms is the most exposed of all to the gap between what it preaches and how it listens.

The lesson generalizes well beyond Slack. The same pattern — massive AI investment paired with form-based listening — shows up across the largest software and platform companies, from HubSpot's CRM research to Stripe's payments research to Shopify's merchant research. The winners in 2026 won't be the companies that ship the most AI features. They'll be the ones that learn fastest from the people using them.

Frequently Asked Questions#

What is Slack AI and when did it launch?#

Slack AI is the suite of generative-AI features built into Slack, including channel and thread summaries, natural-language search that answers questions, daily recaps, and AI huddle notes. It launched for enterprise customers in February 2024 and later rolled out to Pro and Business+ plans, initially as a $10-per-user-per-month add-on before Slack revised pricing and packaging in 2025. By Dreamforce 2025, Slack had expanded it into a full "agentic operating system."

Does Salesforce own Slack?#

Yes, Salesforce owns Slack. Salesforce acquired Slack for $27.7 billion in 2021, and Slack now functions as the conversational interface for Salesforce's broader AI strategy, including Agentforce. By fiscal 2026, Salesforce reported 2.4 billion "Agentic Work Units" delivered across Agentforce and Slack combined, reflecting how tightly the two products are integrated around AI agents.

How does Slack collect product feedback from its users?#

Slack collects product feedback primarily through in-app surveys, NPS prompts, feature-request channels, beta programs, and usage analytics, supplemented by feedback-tracker templates and integrations with roadmap tools. These methods capture ratings, votes, and behavioral signals at scale but tend to flatten the reasoning behind a request into structured fields. Conversational AI interviews are an emerging alternative that captures the "why" behind feedback at comparable scale.

Why is conversational research a better fit for a company like Slack?#

Conversational research fits Slack because the company's entire category thesis is that conversation beats forms and email, yet its traditional listening tools — surveys and NPS — rely on the static formats it argues against. AI customer interviews let users explain intent, workarounds, and context in their own words, then follow up automatically, which matches how Slack users already communicate. It closes the gap between what Slack champions for work and how it researches its own product.

How many people use Slack in 2026?#

Slack reaches an estimated 47.2 million daily active users and 79 million monthly active users, across more than 750,000 organizations, with 77% of the Fortune 100 using the platform. More than 1.5 billion messages are sent on Slack each day, and the platform generates an estimated $6.98 billion in annual revenue. That scale is what makes both Slack's AI bets and its customer-research challenge so high-stakes.

Conclusion#

Slack's AI strategy is one of the most ambitious in enterprise software: take the messaging platform 47.2 million people open every day and turn it into the operating system where AI agents live, get context, and do work. The bet is sound. The risk is that a company built on the belief that conversation beats forms still learns about its own users through surveys, NPS scores, and feature-request boards — instruments that capture signal but discard the reasoning that actually moves a roadmap. The faster Slack ships AI, the more that gap costs.

Closing it doesn't require choosing between depth and scale anymore. Conversational AI interviews deliver the "why" of a manual interview at the volume of a survey — exactly the capability a product team building AI for millions of users needs. That's what Perspective AI is built to do: run hundreds or thousands of customer interviews at once, follow up on every vague answer, and surface the intent behind the metric. If your AI roadmap depends on understanding people, not just counting them, start a research study or see how Perspective AI works — and stop letting the form flatten the conversation your product depends on.