Stripe's Knowledge AI Platform

Stripe's Knowledge AI Platform stripe.dev

Stripe's Knowledge AI Platform is our versatile AI agent platform built to handle diverse non-coding knowledge work, from quick queries to complex, multi-day projects. By connecting employees to over 1,000 internal…

189 points•ltononro•8 days ago•118 comments•

#ai #platform #knowledge #stripes #knowledge ai #ai platform #stripes knowledge #stripes knowledge ai

118 comments

quadrifoliate7 days ago

Stripe is the company I usually hold up as the exemplar of polished internal tools. Hopefully this is taken the right way and I don't want to be negative towards the teams working on this, but I see a distinct lack of polish in these tools and presentations. Some examples:

- Unnecessary AI copy throughout the interfaces like "Browse, discover, and manage skils for your agents", "No favorites yet — hover a card and click the star to pin it here", "One execution environment, shared across agents". These instantly read as AI copy and decrease my enthusiasm.

- Inconsistent, AI-sloppy look-and-feel with different typefaces spattered across the interface.

- The session metrics slide looks busy and AI-generated. It repeats 360,014 sessions in one of the cells at the top, but also has a "360K sessions" in the heading.

I don't know if I'm the only one that notices this stuff, or whether others see it too.

reticulates7 days ago

I think there’s been somewhat of a mask-off moment post-AI in which we’ve realized that a lot of the careful and considered output we’ve come to expect of some companies wasn’t out of a respect for the craft or desire to produce “good work”.

Pre-AI the attitude was: if we are going to do something, it is going to use up our precious resources, so we should do it well, because our staff are capable and the marginal cost of doing it well vs. doing it at all is negligible.

Post-AI: we can churn out things quickly, we don’t have to worry about resource allocation, churn churn churn!

Ultimately, it is pragmatic for businesses to behave this way, but it is a shame for those who love the craft. I think we took for granted the beautiful ornate hand carved furniture era of software engineering. We are now in the ikea era.

burningChrome7 days ago

>> Ultimately, it is pragmatic for businesses to behave this way,

This is how I realized we truly are in an AI arms race. Literally every app or website or tool I use now forcibly pushes their "AI features" or "AI assistant" on you instead of just letting you use their products the way it makes sense for you. Stripe just confirmed what OP was saying. Poorly built, poorly thought out software is now the norm not the exception any more.

quadrifoliate7 days ago

> Ultimately, it is pragmatic for businesses to behave this way, but it is a shame for those who love the craft. I think we took for granted the beautiful ornate hand carved furniture era of software engineering. We are now in the ikea era.

I feel this so much. I thought AI would make it easier to get lots of hand-carved ornate furniture; but right now what we are getting looks like the typical Ikea product line -- a bunch of mismatched pieces by a host of different designers with no common underlying design theme or continuity (I'm talking specifically about the looks and design, not the materials).

rkagerer7 days ago

We are now in the ikea era

If Ikea furniture were lopsided, ugly, not-quite-properly functional and traded their minimalist design for a Frankenstein hodgepodge of redundant parts.

hax0ron37 days ago

I don't think the beautiful ornate hand carved furniture era of software engineering has existed for many decades now.

What's happening right now is more like, we're moving from a hyper-optimized era of just-in-time manufacturing (in software, this is called "Agile", "scrum", "sprints", etc.) to getting our manufacturing outsourced to another country. In software, the "other country" is AI.

csomar7 days ago

I find it interesting that China today is moving in the opposite direction and going as far as banning low-quality exports and stopping this “churn” of products having learnt that low-value high-quantity products are a bad deal all around.

chrisvls7 days ago

I am really interested in this topic, but man I had to give up reading the document because there were too many paragraphs that contained lots of words but no additional detail that I could actually use. It could be AI-generated or human-generated with too many buzzwords. Doesn't really matter, still gave up.

quadrifoliate7 days ago

It's not just you. Let's take a small paragraph and break it down.

> Surface-agnostic APIs: Kai ships with an opinionated web application and a Slack integration, but the main primitive is the underlying API that powers them both. The agent is a service, not an application, and surfaces are simply customized views into it.

I...kinda get that they built an API that the agent polls. And then Slack or the webapp both invoke the agent? If so, the above paragraph is obscuring the point. Also, why isn't there a diagram showing it? There is a diagram below showing "Execution Environment" below, but that one doesn't show the API. And Kai and "Stripe Product Agents" are parallel paths on it. Does it not call out to this API anywhere?

I doubt if most people these days actually try to understand any of this, or just go "Agents? Cool! Here's some product that's vaguely similar that I like or have worked on!".

moograms7 days ago

Hi, I'm Drew, Stripe's new engineering blog editor. Thanks for sharing your experience! I didn't edit this one, but this is helpful for future posts.

nullbio7 days ago

It sure feels like companies have lost their way now. It's no longer about high polish and quality, and all about volume of features. I don't think users want to be overwhelmed with features, they want things that work nicely, and consistently.

yojo7 days ago

Working in b2b SaaS, there’s a real gravity toward checkbox-driven development. Any time we lose a deal because a competitor had feature A, the immediate response is “we need to build feature A.”

And from a sales perspective, that’s probably not wrong.

I think a new (or newly critical) engineering challenge is how to build in a way that keeps a codebase that’s churning out features from crumbling under the weight.

enraged_camel7 days ago

>> I don't think users want to be overwhelmed with features, they want things that work nicely, and consistently.

In the case of Stripe, users want one thing, which is to receive correct answers to their questions nearly instantly. And from the sounds of it, that is indeed the case.

ashkankiani7 days ago

That's funny, because I recall in 2015 hitting their endpoint for a large-ish customer, and if you added a boolean to get the total result count, it would 500 every time, presumably because it was doing some kind of "SELECT count(1)" over a postgres table. IIRC, Stripe was ruby internally for a long time, no?

Also their documentation was frequently just straight up incorrect (as in the described json schema for a response was violated. keys missing, different field names, etc.).

But it's been over 10 years, has it improved since then? I'm still in my impression of their stack from back then, although they were decently mature by then as well.

davidee7 days ago

I recently had to implement a stripe payment system for a client (having never used it before) and found their documentation to be accurate and excellent (at least for the Java SDK and relatively basic use case).

No love for Stripe but IMO their documentation feels like a first-class product.

ltononro7 days ago

s happens, first impressions are what they are

anthonypasq7 days ago

> Stripe is the company I usually hold up as the exemplar of polished internal tools.

unless you work there how would you know this?

edlebert7 days ago

I worked there. All the internal tools are amazingly polished. I’ve never seen anything like it.

quadrifoliate7 days ago

They have published a bunch of stuff about their internal tools in the past. Look up https://stripe.com/blog/stripe-home and compare it to this, as an example.

It's always possible that in reality they were always a bit less polished behind the scenes though.

lukebuehler7 days ago

Very cool demonstration of managed agents built for the needs of their own business.

I think this is where a lot of companies are going to go: on-prem platforms that give various teams access to agents that are as powerful as coding agents, but much more managed and governed.

I'm betting on this with my open source project, Lightspeed: https://github.com/smartcomputer-ai/lightspeed

apsurd7 days ago

You wrote your readme yourself. It shows. So refreshing and i was able to finish it too.

aleqs7 days ago

The readme was very clearly AI-written (or at least heavily AI-assisted)... not saying that's a bad thing, and I agree it's fairly well written... but it is very obviously AI

lukebuehler7 days ago

Thank you. Yes, I did. There might be some AI-isms in there, because agents just can’t help themselves to dump their arcanae in there.

hmokiguess7 days ago

Your project seems way better than the stuff they built, nice work.

neilellis7 days ago

I like it, very smart for agents at scale.

sebmellen7 days ago

Wow, where do you work? Can we talk? Looking for smart people!

hek2sch8 days ago

I read a buzz word "Knowledge AI Platform" but I did not see any specific feature helpful for knowledge management like verification or transparency. It is more like any generic Agent builder. Maybe it meant to justify building something internally.

ofjcihen7 days ago

None of these have been especially impressive.

I think that’s because none of them go past “I’ve set up agents to be orchestrated this way” and that’s about as impressive as “look at my cloudformation template”.

bob10297 days ago

> Building a standalone agent product wouldn't work. Instead, it would force users out of their natural workflows and into a new app.

I am finding the opposite to be true with one of my clients. They explicitly want a new channel. The chat-style UI/UX is vastly preferred over their poorly maintained internal tooling.

I suppose if you are Stripe actual, most things would be well designed and more difficult to abandon. Most places aren't like Stripe.

The hard part is bringing everything that matters to the new channel, but it's certainly feasible to do this. I am doing it right now. We will soon be able to delete hundreds of wildly inconsistent cshtml views and related controllers in favor of a single agent tool that applies json patches. This will easily cover 99.9% of use cases. A manual json editor is retained for the rare case where we need some multi-megabyte merge operation.

atonse8 days ago

Ahhh I built an internal set of agents to run our company (finances, all info in the karpathy-style LLMWiki and a database of clients, contracts, billing, time tracking, all managed by MCPs, etc) and it's also called Kai (company name is Kaizen)

Happy to be in such good company!

infecto8 days ago

All hand written or did you have any underlying libraries that you used?

atonse7 days ago

Mostly custom. Full data model and full web and mobile app to help run day to day operations.

The agent has switched a few times, initially NanoClaw, then Hermes, and now Vercel's Eve (maximum customizability).

NextJS/Shadcn web app, postgres db, eve agent layer, MCP tools (over 100 so every single thing can be done by an agent), SwiftUI mobile app, etc.

Integrations with gmail, gcal, gdrive, quickbooks, using mdx for the wiki displays (so we have rich diagrams, 3d models, etc).

We write about it here: https://kznconsulting.com/work/how-we-run-kaizen

Read the full thread on Hacker News →

118 comments

Related stories