The opportunity, in one line

You can skip building a whole web app and launch your AI tool where your customers already work. Here is the one rule that keeps you from becoming a tenant who gets evicted.

First, the honest reframe

• It is not just Slack. Chat is becoming the UI layer for business AI.
• And Microsoft Teams is the bigger surface (Teams ~320M users vs Slack ~38-47M, estimates). Third-party
• The real trend: agents are moving into the chat tools teams already live in. Slack is the sharper, startup-friendly version.

What's driving it: distribution

• Agents move "out of isolated browser tabs and into the shared context of where your team works." Official, Slack
• Slackbot rebuilt into an orchestration layer that routes to specialist agents. Official
• The growth engine is push, not pull: from Summer '26, Slack is "provisioned on day one for all new Salesforce orgs." Official

Who is doing it

• Salesforce / Agentforce — native agents + Slackbot
• Anthropic — Claude in Slack, and Claude Tag (Jun 23) Official
• Marketplace — Anthropic, Box (Google is a Drive connector, not a full agent)
• Dev on-ramps — "Add to Slack," Slack Agent Kit, MCP
• Startups — Boring Marketing, the one in today's demo

How fast it's growing (say "Slack says")

• AI-enabled Slack apps up 690% year over year Slack's claim
• 1.7M apps used in Slack weekly; 95% say it makes tools more valuable Slack's data
• Slackbot "on track to become the fastest-adopted feature in Salesforce history" Salesforce
• Salesforce reported one customer saved 125,000 hours in two months Vendor

Who has actually grown this way

• Independent, mid-size: Donut, Polly, Guru, Geekbot reported
• Acquired by the platform: Troops (Salesforce '22), Rimeto (Slack '20) Official
• The pattern: nobody became a giant purely on Slack. You build a nice business, or you get bought.
• Even Boring isn't Slack-native: its real product is the web dashboard.

Who it's a good idea for

• Good fit: chat-native, lightweight, team-used. Standups, polls, nudges, alerts, sales pings.
• Bad fit: rich-UI, dashboard-heavy, data-heavy, consumer, solo.
• SalesUpLevel = good fit (daily practice nudges in chat).
• PageMotor = companion only (a CMS lives in a dashboard, not a thread).

The bear case (the catch)

• API clampdown — Slack's own terms (May 2025) ban training LLMs on Slack data, bulk export, persistent indexes. Official
• Gatekeeper — Slack ships its own agents; yours can be squeezed.
• Lock-in + cost — the platform can hold your memory and your bill.
• Reliability — a remote brain hiccups, and your Slack looks broken.

The tell: Slack locked the data in 2025, then invited the agents in during 2026. On its terms.

Meet Boring Agent

• Built by James at Boring Marketing (@boringmarketer). Credit up front.
• Most AI answers from generic training data. Boring answers from YOUR data: your rankings, competitors, the gaps.
• Add it to Slack, point it at your site, and just ask.

It profiled PageMotor on its own

• Nailed "JSON-native operating system for websites."
• Found the built-in llms.txt + AI-readiness tool.
• Even found the Practical AI show.

On screen: Boring's brand profile of PageMotor, homework it did on a company it had never seen.

10 territories, wide open

• 783 keyword signals, ~15.6M monthly searches.
• Biggest rooms: AI Website Builder (4.2M), Portfolio (4.5M), SEO (3.5M).
• All "Absent." Boring's data

On screen: ~15.6M monthly searches mapped, PageMotor ranks for ~none.

The site is behind the product

• 0 of 10 citations across the AI engines. A real wake-up.
• The cause isn't the product. The public site still describes a 2025 PageMotor.
• AI reads your site to decide who you are. Stale site, wrong story.

On screen: LLM visibility, 0/10 today, because AI is reading an out-of-date front door. Boring's measurement

Currency wins citations

• Competitors with current, crawlable content get cited (builder.io 8/8).
• It's not product quality. It's whose site AI can read and trust today.
• The fix is in your hands: make the site current.

On screen: competitors cited (builder.io 8/8), the gap is site currency, not product.

The "agent SEO" land grab

• ~60,000 searches a month. Nobody owns it.
• It's literally PageMotor's two-websites thesis.
• Plus "best GEO tools" — an uncontested SERP.

On screen: Boring's #1 rec, with a full written brief.

And it works IN Slack

• Reliable lookups live: /boring status · balance · runs.
• It can turn a finding into a brief and a draft.
• But it "never posts or publishes on its own. You decide." Good guardrail. Credit James.

On screen: 62 ranked actions in the queue, not just ideas.

28 fixable issues

• Duplicate content, thin pages, missing meta.
• All marked "agent-executable."

Caveat: it shows "0 pages crawled," so don't lean on the 70/100 score on air.

On screen: site health 70/100, 28 fixable.

Two surfaces, one brain

• The dashboard cockpit hums — that's where the value lives.
• A live @mention round-trip can be slower, because the work runs on their servers. That async step is the hard part for every agent right now, not any one product.
• Frontier, not failure. (And "not approved by Slack" + broad permissions = why I tested on a clean workspace.)

Credit-based, so knowing what to ask pays off

• These agents run on credits, like most AI tools. The deep analysis is the cheap, high-value part.
• The skill that saves money is asking sharp, specific questions, not wandering.
• Lesson for any agent you adopt: learn what it's great at, and lean there.