What changed in seller automation over the last two years

Two years is a long time in this industry, long enough that describing where things stood in 2024 already sounds like describing a different era. 

Here’s what actually happened between then and now, not the vague “AI is changing everything” version that gets repeated so often it’s stopped meaning much.

Where things stood two years ago

In fall 2024, repricing was still rule-based by default: a min, a max, and a tool that matched or beat competitors inside that band. Hybrid AI-assisted repricing existed, but it wasn’t what most sellers were running day to day.

AI-assisted listing writing was real but far from standard. Roughly a third of sellers were using AI to create or optimize listings, with a smaller share, around 14%, using it for marketing or social content. A meaningful chunk of the market, but nowhere near universal.

Tooling was fragmented by function. A repricer, a PPC platform, a listing optimizer, and a forecasting tool were typically separate subscriptions from separate vendors, each working in isolation from the others, with no shared view of the account across them.

And Amazon’s own Seller Assistant existed, but it worked as a reactive chatbot. A seller asked a question and got an answer. It didn’t monitor an account or take action on its own.

The turning points that actually moved things

Rufus and the shift to intent-based search

Amazon’s generative shopping assistant launched in 2024 and scaled hard through 2025 and 2026, with usage reported up 115% year over year by Q1 2026. 

Rufus runs alongside Amazon’s underlying semantic search system, and together they reframed what “optimizing a listing” even means. Listings built for 2024-style keyword density now underperform ones built for intent-based, conversational content, a shift that happened underneath most sellers without much fanfare.

Amazon’s own assistant went from answering to acting

This is the clearest before-and-after moment in the whole window. Seller Assistant’s agentic upgrade turned a chatbot that answered questions into what Amazon calls an “always-on agentic AI partner.” 

It started monitoring accounts continuously, identifying issues on its own, and executing plans once a seller granted permission. The distinction is worth naming plainly: a chatbot responds to what you ask. An agent acts without being asked, once it has permission to.

Amazon opened its systems to outside AI agents

Amazon introduced a native Ads Agent in November 2025, then opened its Ads MCP Server to public beta on February 2, 2026. That let external AI tools, not just Amazon’s own, connect to a seller’s ad account through one standardized integration instead of a custom-built connection for every vendor. 

It’s a meaningful shift because it’s Amazon actively choosing to make its systems reachable by outside agents on defined terms.

A compliance layer arrived alongside the capability

Dynamic Canvas launched March 3, 2026. Amazon’s formal Agent Policy took effect the very next day. For the first time, which AI tool a seller chose became a compliance question, not just a capability one. Any agent touching a seller’s account now has to self-identify as automated and operate through Amazon’s official SP-API rather than simulating a browser.

What this did to the third-party tool market

The category built around PPC, repricing, and listing management shifted from rule builders sellers configured by hand to agents that read the account and execute with approval. Older tools required a seller to set bid rules and thresholds manually. Newer ones read the account, set a strategy against a target, and execute changes with the seller approving rather than operating every lever by hand.

That shift came with real price movement at the high end. One example that shows the scale of it: a leading PPC automation platform’s entry tier went from roughly $250 a month a few years ago to $695 a month by mid-2026. That kind of jump prices some smaller sellers out entirely, pushing them toward cheaper or bundled alternatives instead of the leading standalone platforms.

Repricing followed a similar arc, moving from simple min and max rules toward hybrid tools that layer AI or game-theory-driven pricing on top of a rule-based guardrail. And compliance became a real axis of tool selection for the first time in this window. 

Choosing a repricer or a PPC platform isn’t just a question of what it can do anymore; it’s also a question of whether it operates on Amazon’s approved SP-API channel, since a non-compliant tool now risks losing account access without warning.

Where it’s headed next

None of this happened because any single vendor pushed a breakthrough. It happened because the ground shifted underneath the whole category at once, Amazon’s own tools, the protocols connecting outside agents to them, and the rules governing how any of it is allowed to operate, all moving in the same direction within a matter of months.

AI Operator is a current expression of that same shift already underway, not a departure from it: an agent working across a connected stack of Seller 365 systems rather than inside a single one, built to operate on approved channels rather than around them. 

If you’re weighing where your own stack fits into this shift, the AIO launch overview is the place to start, and what Amazon’s own AI tools do and don’t cover is worth reading alongside it, since Amazon’s own native AI is a real part of this same story, not a separate one. 

AIO is included with Seller 365 and runs across Tactical Arbitrage, InventoryLab, SmartRepricer, and FeedbackWhiz. Take it for a spin today or start a free trial to see where it fits into your own stack.