RFE Online
Applied Intelligence
On 25 June 2026, Notion announced it was killing its Skiff-influenced email app. Users had already switched to AI agents. That is the Notion product story. The tech budget story is different — and it has line items with your name on them.
TL;DR
Notion killed a product because agents replaced the workflow. Execution-layer SaaS — email triage, scheduling, simple workflow automation — is being cannibalized right now. Your SaaS stack needs a vulnerability audit before your next renewal cycle, not after.
Definition
SaaS cannibalization — The process by which AI agents absorb the execution tasks a SaaS tool was purchased to perform, rendering the tool redundant without a formal replacement decision being made. The licence continues to be paid; the tool stops being used. Notion/Skiff is the first major vendor to acknowledge this publicly and respond by discontinuing the product.
Key questions answered
On 25 June 2026, Ars Technica reported that Notion was sunsetting its Skiff-influenced email application. Notion’s stated reason was unusual in its directness: most users had already switched to AI agents for the tasks the email app handled. The product was not technically broken. The market for email had not vanished. The specific tool Notion built had been made redundant by a category of technology that did not exist when it was first conceived.
Applied-intelligence lane score for the Notion/Skiff signal (peak: 28 June 2026). Named in Vanessa’s 16:00 roll-up as a strategic pivot signal distinct from the trust-collapse cluster. Current issue alignment: 65/100. Category fit: 83/100. Macro narrative: 68/100. Source: data/research/signals-2026-06-30.json; ideas-db entry notion-killing-skiff-influenced-email-ap-4a821c; provenance rss/ars-technica/technology-lab, published 25 June 2026.
What Notion has done is rare: it made the cannibalization event visible and public. Most of the time, this process is invisible. Licences renew. Usage dashboards go unread. A tool that employees quietly stopped using months ago continues to generate monthly charges until someone runs a stack audit and asks why this line item exists. Notion just ran that audit in public, and the answer it found is the same answer that is sitting inside your next renewal cycle.
Notion did not shut down because email is dead. It shut down because the specific execution task its product was built around — email triage and composition — is now being handled by agents. That distinction matters enormously for every vendor whose product is built around an execution task.
The Notion/Skiff signal identifies the pattern clearly enough to apply it across your existing stack. The relevant question for each tool is the same: is this product primarily an execution layer, or is it primarily a system of record? Execution-layer tools are vulnerable. System-of-record tools are defensible.
| SaaS category | Primary function | Cannibalization risk | Why |
|---|---|---|---|
| Email triage and prioritisation tools | Execution layer | High | Agents already handle triage, drafting, and scheduling at the task level. Notion just confirmed this at scale. |
| Scheduling and calendar assistants | Execution layer | High | Booking coordination and availability negotiation are routine execution tasks agents perform natively. |
| Standalone note-taking and knowledge capture | Execution layer | High | Agent-generated summaries, meeting notes, and brief captures have replaced the manual capture habit for many knowledge workers. |
| Single-purpose workflow automation (lightweight Zapier-tier) | Execution layer | High | Agent orchestration is absorbing rule-based trigger/action automation at the simple end of the market. |
| CRM activity logging and follow-up prompting | Execution layer | Medium | Logging and prompting tasks are being absorbed, but the underlying CRM as a record of deals and relationships retains value. |
| Project management (task creation and status updates) | Mixed | Medium | Task creation and update entry are at risk. The project plan and stakeholder record are more defensible. |
| ERP, HRIS, compliance databases | System of record | Low | These tools hold the canonical data that agents query. The data stickiness protects them. Agents read from these — they do not replace them. |
| Finance and payroll systems | System of record | Low | Regulatory and audit requirements anchor the record. Agent automation runs alongside these systems, not instead of them. |
The pattern across the high-risk category is consistent: these tools were built to help users do a task. Agents now do the task. The tool is no longer the interface to the task — it is just an optional layer that most users have already bypassed, whether their organisation knows it or not.
The Notion/Skiff case was unusually transparent. Notion could see the usage data centrally. It could measure the adoption collapse. Most organisations cannot, because the usage data for individual-tier or team-tier SaaS tools sits inside the vendor’s dashboard — not in a consolidated view that a CFO or IT lead reviews at renewal time.
The result is predictable: organisations continue paying for execution-layer tools that agents have already made redundant, while simultaneously paying for the agents. The double spend is invisible until someone looks for it. And looking for it is uncomfortable, because it surfaces a question that no one wants to answer out loud in a renewal meeting: has this tool been useful in the past twelve months, or have we been paying the vendor to observe our agents doing the work instead?
A SaaS tool that is licensed for 40 users but actively used by 8 — while agents handle the workflows the other 32 have abandoned — is not a usage problem. It is a cannibalization event. The vendor will not surface this at renewal. Your finance team will not see it in the aggregate spend number. Only a workflow-level audit of where agent activity has replaced app activity will find it.
SaaS vendors built stickiness through habit formation — daily check-ins, notification loops, inbox-zero rituals. When agents absorb those rituals, the habit breaks. But breaking a habit is not the same as cancelling a subscription. The tool stays on the invoice long after the behaviour that justified it has shifted to an agent workflow.
The switching cost that kept execution-layer SaaS sticky was not the data — it was the workflow habit. If the habit has already been broken by agent adoption, the moat has already eroded. You are paying for a moat that no longer exists. At the next renewal, the vendor has no retention lever except price, because your team is not using the product for the thing it was designed to do.
The conventional technology procurement model asks: what is the right tool for each job category? That model produces a stack of per-seat licences organised by function — email tool, calendar tool, CRM, project management, automation layer. Each tool addresses a workflow. The IT budget is a map of workflows.
That model is breaking. Not because the workflows have disappeared — users still need email triaged, meetings scheduled, and follow-ups logged. It is breaking because the execution of those workflows has moved to a layer that does not have a per-seat licence. The agent does not have a line item on your SaaS invoice. The governance of the agent does.
The strategic pivot from SaaS buyer to agentic operator involves three reorientation moves:
The organisations that will have a SaaS budget problem in 2027 are the ones that keep renewing execution-layer tools while also paying for the agents that have replaced them. The organisations that will not have that problem are the ones that ran the workflow audit now and redirected the spend.
Notion made its cannibalization event public on 25 June 2026. Your vendors will not do the same thing — at least not until they have run out of ways to retain the licence. The signal is visible in usage dashboards if you know what to look for: active user counts that do not reflect your licensed seat count; workflows that were completed monthly that are now completed rarely or not at all; agent completion logs for tasks that used to require the tool.
Three actions worth completing before your next significant renewal date:
RFE Online’s Agentic Services practice is built for operators who are running agents in production and need the governance layer to match. If your stack audit surfaces agent-cannibalized workflows — or if you want to design the agentic replacement before the next renewal cycle — the strategy call is where that conversation starts.
Book a Discovery CallAgentic Services
When agents replace your SaaS tools, the governance layer is what you actually needed to buy.
Agentic Services is RFE Online’s practice for designing, governing, and operating AI agents in production: scoped authority, workflow integration, and the operating layer that makes agent-driven execution safe and auditable for client-facing work.
Explore Agentic Services →Audit which parts of your SaaS stack have already been cannibalized — and design the agentic layer that replaces them safely.
Share this insight
Written by
Founder, RFE Online — Fractional Strategic Technology Mentor
Andrew Russell founded RFE Online to close the gap between what the modern world demands and what people and organisations are equipped to handle. His writing spans AI systems design, financial independence, career architecture, mindfulness, and the questions that cut across all of them.
Was this insight useful?
Agentic Services
The Notion/Skiff signal is the first vendor to say it out loud. Agents are replacing the execution layer that SaaS was built to deliver. Agentic Services is how you design the governed replacement — before your next renewal cycle forces the conversation.