Agent assistants are having their App Store moment. Grok, Dots, and hundreds more agents are all getting in on the game, claiming to be able to perform “any task you throw at it.” There was also a recent viral case of a user and an investor’s Instinct negotiating a place with organizers to get her on the Jumbotron screen at the US Open.
https://x.com/oliviaalevine/status/2095705075633000815?embedable=true
Hiring people through TaskRabbit, booking a vacation - well, those are now a piece of cake for assistants.
While all the buzzy conversations are taking place on the consumer side, actual B2B and partner experiences for marketplaces seem to be stuck in 2025.
Where is that agent that can really help partner really scale their business or maybe even simplify their lives?
Why Seller Experiences are ripe for Agents
Being a seller on any online platform (e.g. Amazon, Zalando, eBay) involves a huge number of tasks that are a combination of digital (cataloging, onboarding, inventory management, sales optimization) as well as physical (packaging, delivery, stock).
Some of the criteria that I normally use when deciding if agents are a part of the job include:
- Are there multiple cohorts that use the platform differently?
- Are there a huge number of functions that are routinely performed?
- Do they have multi-step workflows that normally span across 2–3 departments?
- Do users (sellers) create their own playbooks and ways of working that work best for them?
- Is the information largely digital (or can be triggered digitally)?
- Is there a clear way to assess whether the job was done well?
If any business meets 3 or 4 of these conditions, it may be ripe for agents. Seller experience in a marketplace meets all these conditions. That makes it an ideal candidate for agents.
Take Etsy. As a partner, you want to understand what you can do to improve sales. Someone responsible for catalog operations wants to know which improvements could help drive those sales. Someone managing ad campaigns wants to understand which customer cohorts they’re missing and who might want their products.
In this series, we aim to deconstruct the key tenets that I believe, would make for a great agent-first experience for platforms to equip their partners to ensure everyone wins.
What a gold standard seller experience
Business owners are on platforms to do business, and anything we can do to make their lives easier on that front deserves attention.
In that sense, we define a four-level system:
Level 4: Fully Autonomous
Think of it like a seller saying:
“I want $20,000 in sales each month with 20% profitability. Keep returns below 30% and product ratings at 4.7 or above.”
Can the system just take that goal and run with it?
Level 3: Autonomous, but with the Seller in the Loop
The system is largely autonomous but keeps the seller updated on what it’s doing and asks for final go-ahead on key calls that are made (e.g. pricing decisions, policy decisions, and inventory decisions).
Level 2: Seller-Driven, AI Workflows
This is what most firms are aspiring to. The partner asks the agent to perform tasks, which the AI then executes. E.g. “Why is my return rate so high, and what should I do about it?”
AI assists, human in the lead.
Level 1: Seller-Owned End to End
This is where most businesses are today, end to end.
There are some startups like Polsia that are aiming to get to Level 4, but one search for them on the web would show a series of concerns when one moves to full automation directly.
A good place to start for most scaled businesses would be somewhere between Level 2 and Level 3 as processes and knowledge gets more robust.
How this playbook would be structured
- We would start by showcasing what 1-2 variants of a great experience may look like
- Speaking about the basics that still hold : Knowing your cohorts, jobs to be done
- Key philosophies: The principles we want to follow, with each philosophy getting its own chapter.
What we hope you would be able to takeaway at the end of this series
At the end of this series, we hope to give you not just principles, but also specific experience patterns, personas, and practical design choices you can adapt to your platform.
Agentic seller experiences are still emerging. The approaches in this series are hypotheses to test, rather than established best practices. I’ll explain the reasoning and trade-offs behind each one, including where the seller should stay in control and how to measure whether the experience is working.
The aim is to help you choose a useful starting point, design it around your sellers’ goals, and build a path towards greater autonomy as your processes and knowledge mature.
In the next article, we would talk about 1-2 variants of what a level 2 / level 3 experience can potentially look for sellers, and how Agents and AI can (invisibly) add huge value to both - the seller as well as the marketplace in question.
Follow me on Substack here.
Free lightning session on 4 November — register here.
Workshop on the same topic on 5 December — details here.