Amplitude shipped AI Assistant yesterday. It's a well-designed product, a smart piece of positioning, and the logical endpoint of a bet they placed eighteen months ago.
When Amplitude acquired Command AI in October 2024, this product was the predictable destination. Command AI had the in-product guidance surface. Amplitude had the behavioral data substrate. AI Assistant is what happens when you fuse those two things and give the result a thoughtful product team.
It's also the clearest validation yet of the category we've been building Deway inside of.
What Amplitude Got Right
Read their launch post carefully. Three moves stand out.
First, they grounded the agent in behavioral context. The assistant doesn't just search a knowledge base. It knows what the user did before they asked, where they are in the product right now, and whether they completed the task after the conversation. That is a substantive improvement over every support bot that preceded it.
Second, they claimed the execution layer. Most support tools stop at the answer. Amplitude's agent can trigger an in-product guide, walk the user through the workflow, and for some tasks, complete the action on the user's behalf. That is the shift from chatbot to co-browsing agent. It is the right framing.
Third, they closed the loop with Session Replay. Support bots usually measure themselves by thumbs-up votes and deflection rate. Amplitude can now measure whether the user actually succeeded after the conversation ended. "True resolution, not just chat resolution." One phrase does more positioning work than most product pages.
Credit where it's due. This is a real evolution beyond legacy digital adoption platforms. Teams on Amplitude should try it.
Why This Was Inevitable
Legacy DAPs were built for a world where products changed slowly enough that a human could manually curate every tour, tooltip, and checklist. That world is gone.
I spent years between product and customer success, including working with some of the most strategic accounts at Mixpanel. At every company I worked with, the same pattern held: product shipped faster than users could absorb, adoption tooling lagged behind, and the gap showed up as silent churn in a renewal conversation six months later.
That pattern is accelerating. Every engineering team I talk to is shipping faster than they were a year ago. Claude Code, Codex, Cursor, and the wave of AI coding assistants have compressed release cycles from months to days. A PR that used to take a week lands before lunch.
But adoption infrastructure hasn't kept up. Tour builders designed in 2015 are still the default way most companies try to close the gap between what they ship and what their users understand. Every release breaks the tours. Every tour breaks the adoption team. Every week, the gap grows.
Amplitude saw this. So did we. Their response was to pair AI with their behavioral data platform and automate the parts of adoption the old architecture couldn't handle. It's the right instinct.
The Architectural Unlock
The reason this moment feels significant is that we're watching a category change its substrate.
A modern adoption agent needs access to five things at once: support documentation, knowledge base content, helpdesk history, real-time product data, and behavioral context. Then it needs a semantic layer where an agent can reason across all of that and actually act inside the product.
Get that substrate right and a lot becomes possible that was impossible before. Guidance becomes proactive instead of reactive. Personalization happens at the level of individual intent instead of segment rules. Resolution means the user finished the task, not that they closed the chat window. The agent moves from "here's a tip you can try" to "I've done it for you."
That is not a feature upgrade bolted onto a tour builder. That is a different kind of product. Amplitude recognized this and committed to the rebuild. That is why AI Assistant works as well as it does.
It is also where we think things are headed next, and where we at Deway are placing our own bet.
Where Deway Leans Further
Read Amplitude's launch post one more time and notice a detail they don't dwell on. When AI Assistant needs to walk a user through a task, it triggers an in-product guide. A guide that someone at the customer built. That guide lives in Amplitude's Guides and Surveys product, and it required a human to author it.
Amplitude's AI is smart at choosing which guide to launch and measuring whether it worked. The guides themselves still have to be created and maintained the old way.
Our architectural bet at Deway is to remove the authoring dependency entirely.
We learn your product by observing how it actually works and how users interact with it, build a living model of your product enriched with your docs and support tickets, validate flows with automated testing, and then deliver contextual guidance to users in real time. When something changes in staging, the model updates itself. When a user does something we've never seen before, the model learns from it.
The human in the loop is critical to allow you to see that it was mapped correctly, and with your input the models get better.
The practical version: you install Deway with a single line of code, it's live inside 72 hours, and from there the model keeps itself current as your product evolves.
This is what we mean when we say autonomous adoption layer. The word "autonomous" is doing specific work. It means the product team does not spend their week writing walkthroughs. It means adoption content evolves at the speed of your releases, not the speed of your content pipeline. It means the guidance compounds in intelligence with every user interaction, whether anyone is watching or not.
Amplitude took the category three steps forward from legacy DAPs. We think the next three steps are about removing the human authoring dependency altogether. That is where we're pointed.
What This Means for You
If you're already running on Amplitude, try AI Assistant. The behavioral context, the Session Replay integration, and the execution layer are real, and the team that shipped it clearly understands the problem. Pilot it.
If you're not in the Amplitude ecosystem, the questions to ask yourself are different. Are you shipping fast enough with AI coding tools that your tour library is always half-broken? Do you want adoption infrastructure that learns your product instead of waiting for you to document it? Do you want an adoption layer that doesn't require you to standardize your entire analytics stack on one vendor?
Your product deserves a better support agent. It also deserves an adoption layer that builds and evolves itself.
If any of those sound like you, come talk to us. We'll build you a personalized demo of how it would look on your product.
Request a personalized demo at deway.ai/see-in-action
Alon Binman is Co-Founder and CEO of Deway, the autonomous adoption layer for modern SaaS products. Before Deway, Alon spent 15+ years working across product and customer success, including as a Senior Solution Architect at Mixpanel.