We’ve built a demo that anyone can try here: https://kenobi.ai/start
We believe that large parts of the web are about to go from being static to dynamic because of how adept LLMs are at transforming rendered HTML. And right now we’re focussing on B2B landing page content (as opposed to application UIs) because there is a lot of commercial opportunity for increasing top-of-funnel inbound conversions.
Our journey to Kenobi today is a long and snaking one. You may notice from the post title that we did YC’s Winter 2022 batch (I know that 4 years is practically ancient in YC-dog-years). Kenobi is a hard pivot from our original idea that we got accepted into YC with — a company called Verdn which did trackable environmental donations via an API. Since the summer, we’ve been hacking on different ideas… We started with personalized UI screenshots for outbound campaigns, but then people told us they wanted transformations to their actual site[0] — so we built an agentic workflow to research a visitor-company and “pre-render” changes to a landing site for them. Ultimately, there was too much friction in getting people to incorporate personalized URLs into their cold outbound campaigns[1]. Besides, people kept asking for us to do this for their inbound traffic, and so our current iteration was born.
Right now with Kenobi you pick a page that you’d like to make customizable, and choose [text] elements that you’d like to make dynamic. You can define custom prompting instructions for these elements, and when someone visits your page, our agentic workflow researches their company, and presents the updated content as quickly as possible, usually within a few seconds.[2] You also get a ping in Slack every time this happens so you know who is using your site.
We’ve been experimenting with features such as generating custom imagery that actually looks good and native to the page design, and pulling in company data sources so that e.g. the right case study can be presented based on a visitor’s industry and ICP profile. Our most requested feature is deanonymizing traffic so that Kenobi’s personalization can happen automatically as visitors land on your page — this is coming very soon, as right now you have to specify where you’re coming from.
It’s surprised us just how much business value we’ve gotten from knowing who (most probably) is on the page and asking for a personalized experience. We’ve seen response rates 3x of what we would normally from following people up from companies we know visited our site.
There are many players in this space already, and everyone seems to have their own angle. We are keen to hear thoughts on what people think the future of the personalized internet looks like!
Cheers from London!
P.S. - there's also a video that Chris recorded showing the end-to-end Kenobi experience right now https://www.loom.com/share/bc0a82a2f2fd40f695315bae80e8f5d8
[0] - Many of them had tried AI “microsite” generators but found the maintenance of managing a separate website(s) just for closing deals to be burdensome and inefficient.
[1] - Despite having a CSV export and Clay integration option for our pre-generated website changes, getting people to weave the URLs into their email sequences (everyone uses different tools) seemed almost insurmountable without building what would ostensibly be our own sequencing software.
[2] - We use light foundation models with grounded search for the research step, and translate these into markup changes via another light LLM pass and our own DSL which is optimized for speed.
I think this is cool and probably the future of B2B websites. My holdup would be, if the buyer enters their company and the copy just changes into what we think they want, are they going to lose trust that the copy is a true representation of our focus? Maybe it's a framing problem, lots of websites have "solutions" sections for different industries. Potentially could be cool to have a "how we can specifically help {your company}" with an exact use case outline.
Great point. That's pertinent to how we've been configuring the research → computing "intent" pipelines. Our focus right now is mainly just to streamline content and show brands "in context" as much as possible without having too much of an "opinion".
Your idea about showing how specifically the company could be helped + a use-case is lovely way of putting some of the more complex layout generation ideas we've been working on!
The core functionality I expect for such a service is for it to automatically detect who I am, I've seen other marketing services do this, there are ways to map IPs to companies and other techniques. Of course, it's rather creepy and not super helpful to the user, but it may have its (shock) value for making certain kinds of products stand out.
And the personalisation itself... Anyone can make a call to an AI + Search service and generate a new version of the HTML with some slightly modified text, which was not all that different, appealing or accurate in the tests I made. I would suggest upgrading to a higher quality model, proper AI can do much better than this if given the right context.
I suppose it's nice that you are making this easy, if you built your site with a visual website builder this wouldn't be completely trivial to replicate. But still, not a very defensible business for now. I suppose that with good marketing and a serious roadmap to beef this up it could be a viable idea.
The number one request we have is to integrate deanonymisation, so you’re right on the money there. That’s coming in the next couple of weeks or so…
Wrt the changes being text-based for now, we do actually have image and complex element and layout generation working, but have kept it as an experiment for pre-rendered pages until we are confident we can get it right in most cases. (Some early beta users used Kenobi to send out in some cases thousands of customised landing pages with imagery)
We‘re also starting our on demand product with text only precisely because we want to hear what people think we should be working on next as we are a super small team of three!
I suppose I was frustrated due to a mismatch of expectations and the fact that I do things like this every day with AI, it feels rather trivial to me.
But I can see how it may appeal to a wider market. I remember coming across a couple websites that were doing this automatically pre-AI, simply detecting who I was and displaying it in some basic ways. And yeah it was a bit weird, but it sure stuck in my mind for a long time, and it was a data-broker type company anyways, it triggered the thought "well if it can do this with me, it must have good data about everyone".
And I can see the more general case for B2B to surface the right use-cases for the user and such. I've interacted with many people with a business profile that would certainly click a "just show me what I care about" button, hell many technical people would love it too just to remove all the whishy-washy hype language and just see what the thing does.
> many people with a business profile that would certainly click a "just show me what I care about" button
You've encapsulated better than I did the kind of visitor segment we're building for right now.
> I do things like this every day with AI, it feels rather trivial to me.
I also agree with this sentiment. It's how I started ("surely it can't be hard to jiggle a page around now with LLMs") and it mostly worked! But the edge cases and heuristics are also proving to be a big chunk of the effort "iceberg" :D
This is a great point, we've heard this elsewhere so I will be adding a section to our site for this. Thanks!
One of the ways we want to enhance Kenobi is to allow the host site to plug in more data sources (e.g. an FAQ, KB, etc). This helps solve the research problem, and it also increases the utility of the tool because more the very generic website can be be customized in more nuanced, specific, and esoteric ways, depending on the visitor.
By users, I mean the people who browse marketing websites. Do they think having their company name / information in your copy is going to be helpful or creepy?
Oh, and did the IP owners give you permission to take Obi-Wan's name in vain?
I tried several different domains and the copy was so generic it gave no indication of being personalised.
We don't know for sure. This is built on the intuition that websites are still a one-size-fits-all approach that makes no sense in an age where we have intelligence at our finger tips, able to process and reformulate information that speaks directly to people.
You could see this idea as being on the opposite side of the same spectrum as agentic browsing (which hasn't really taken off yet).
And thanks for the feedback! There are limitations in the quality of the personalizations in onboarding experience due to latency constraints. These get lifted the moment you create an account, and can start doing some more in-depth context gathering of your website and the types of visitors you're likely to get.
Second, we want to show B2B visitors’ brands in context. I.e. showing you what it would look like if your company was using the service in question with social proof from your industry peers. We don’t have our image tech in the on-demand demo right now, but companies that we have helped pre-render copies of their site with dynamic images (especially e-commerce brands), found higher engagement on their outreach as a result.
Obviously a lot of companies have vastly different personas, and this is a harder problem to solve which we are working on.
Of course internally it might be much more complex than this but this is how I would do this if I had to build it.
It had no highlight styling so if you wanted to highlight a portion of your input it would be impossible to tell what the current selection was.
Looks cool, not trying to hate, I just have a pet peeve around native elements having basic function/accessibility features removed during heavy styling.
Definitely a balance to be struck with going heavy on styles, whilst not breaking native element features and violating users' expectations!
One idea would be to track what recent technologies/products they’ve bought recently (or what products they’re using) using Bloomberry. ie. If they’ve started using Okta, it might mean they’re investing in security tools [1]
[1] bloomberry.com
We're definitely going to be looking into finding more relevant signals on a case-by-case basis, and this kind of idea fits really neatly into that paradigm!
Neat idea to track the technologies they've bought _recently_ though! I think capturing buying signals (and inferring intent that way) would be a neat addition to the pipeline!
Failed requests that are coming in (people are trying some wild inputs!!) we are keeping an eye on and adding to our edge cases pile to fix asap!