From Segments to Systems: 1:1 Personalization in Action

Julia Nimchinski:
And now, please welcome Mikhail Dorji, Director of AI Innovation and Digital Marketing at Optimizely. One-on-one personalization as an AI system. Finally, it’s happening! Welcome to the show, Mikael, how are you doing?

Michiel Dorjee:
Thank you for pronouncing my name correctly. I know I have the hardest name up there, but yeah, thanks. I’m excited to come talk about ABM as one of the use cases of our platform, for sure.

Julia Nimchinski:
Super excited to dive into this.

Michiel Dorjee:
For sure. So, let me just get started, then. Yep. Yep, let me share my screen, and you can actually see it. So, before I actually share my screen, maybe it’s good to set the scene a little bit. Optimizely, as a business is a B2B business, right?
I work for Optimizely, and I manage our own AI implementation as a business, so, like, I’m customer zero, as we like to say. I always have a joke that’s… We’re our best and worst-owned customer, because we use everything that we sell, but we pay for nothing.
So… Optimizely has everything from a CMS, so you can host your website and build it with us, to an experimentation platform, that’s what we’re most well known for, A-B testing and experimentation on your sites. And now we also have an orchestration product as well for AI. And it kind of sits across those products, but also entirely separate.
So, you can use it, we see lots of businesses adopting it now for entirely separate use cases, where they want to do competitive analysis, and, like, ABM is one of those use cases. And what better way to show off what it can do than just… cases that we have done ourselves. So… I built this little search engine to make it easier for me to make.
I’m gonna show you the output of it first, and then we’re gonna back ourselves way into how we actually built it, and how you can do it with the platform. So, I’ll pull up one of the pages, you can see it here.
We as Optimizely like to position ourselves against competitors, so I built this competitor page, and we actually scanned the Amazon website and said, hey, just for this example, because we would never sell our CMS or something like that for Amazon. we’ve created this beautiful landing page with, like, all this account intelligence built in.
So, it actually built this entire UI, we didn’t, like, actually build the UI itself. And every one of the accounts that we’ve done this for, we have thousands of accounts that we’ve targeted this for. That list is inside Salesforce.
We automatically detect if a new company gets added to that list, and then immediately do all the research and spin up that company. So what does it use, for instance? It uses information about their company that you can find publicly, it uses their information from LinkedIn, the recent news, job profiles.
Actually, interesting insight is that a lot of companies will actually just put what they care about as a company directly in their job specifications, because that’s what they’re actually investing in as a business, right? If you’re hiring people to do that in their own language, so it’s very helpful.
And then we even look at their website and their sitemap, their brands, their accent colors, strategic investments, and then a lot of the communication that we’ve had as a company with a prospect, like Amazon in this case, would be inside of our Salesforce, right? So the system also goes through that to try and evaluate.
Oh, yeah, I actually… Deleted the image earlier, that’s okay. It goes out to their site, takes a screenshot, and then positions our products against a competitor. So in this particular case, it’s detected that they are on Adobe Experience Manager. Boo, terrible product.
And it will actually use approved language from us as a business, and actually go through our own site and our content, through our CMS, to use approved languages and content to build out this page. And then entirely… oh, someone was actually editing this to show another demo.
You can actually see that all of this content is directly cited and sourced. Everything is clickable, it’s directly accessible through our system. And then you can see here it built this nice little pitch. It’s part of the pitch. I even made this little x-ray mode to show you where all the data comes from.
So to come up with this, like, 90-day plan, for instance, you can see it used our own Opal AI. The OPL system is the AI that we used, and you can see here it even named specific account executives, and it got this data from Salesforce, and tries to pull in their profile pictures if they have any, all that stuff. Super neat.
And then at the end, and this is critical, right? Us as marketers, so even in the best-case scenario, you have a beautiful landing page on your website, you optimize it and A-B test it, you might get 4, maybe 5% conversion rate on your site. That’s not enough, right?
Still, if people have to enter a form, I know who these people are that I’m now targeting this page at, so I don’t even need to show them a form. So if they click, I immediately track that action, and then… Tag the right people in our systems to immediately tell them to follow up with this account. So, how did we build this?
If I go back for a second, you can see here, this is directly inside of our CMS, right? So it’s actually created real production systems.
What we hear from a lot of our customers and prospects is that they love that they can vibe code things with Claude and make things with ChatGPT, but the problem is actually making that real and scaling it and putting it back into systems and creating consistency.
We’ve run this over 7,000 times for ourselves, and we’ve only seen 30 different pages where there were issues, and those were mostly related to the companies were just so small to target that it could not find information on the company, like their logo, or there were some data connection issues that we quickly got to fix. Pretty helpful.
Optimizely has been in business for, like, 30 years at this point. We work with the largest companies in the world. Like, we’ve worked with Amazon in the past, but we also work with LEGO, Nike, Nintendo, IKEA, you name it, you interact with it, they likely do some form of our product suite, or all of it.
And we have used all of that knowledge to not only help you get the job done faster, but to get the job done well, right? Because in the age of AI, if you imagine a little triangle. There’s 3 parts of any digital project. It’s getting things done fast, it’s getting things done cheap, and it’s getting things done well.
And traditionally, it’s very hard to actually do things on all three parts of that scale, right? I can do things cheap and fast, but not good. I can pull to good, but then it’s gonna cost me more or take more time. And AI has essentially rendered cost and time zero, right?
It’s so fast that it can do stuff, but everybody’s now generating hundreds of blog articles on their site. Are they any good? I would argue not. So, the way that we solved that is we actually sat with our internal teams and said, hey, you’re an account-based marketer.
If you got to spend 100 hours with every single prospect, what would you tell them? How would you frame our story? And we worked our way backwards from there and said, okay, how do we actually build these into our systems?
So this is our CMS, you can see if I wanted to, I could still go in and make all the edits, I can tag the right people so that they can actually review all the content before it goes out and gets sent to someone, and we even use the data that we then collect on the edits that get made in that phase to then make the next agent run even better.
If I quickly hop over to our agent, you can actually see what the process is here in the background, and I’ll show you in a second how you can actually use the AI tool, but because we have a very AI-forward audience here, I’m jumping straight into it.
So this is our workflow builder, where it can actually set up how this actually uses the data and runs. So in this particular case, I’ve added a trigger that says, hey, I want to run this on a schedule, like daily, for instance.
And I’m going to get the list from Sitecore Takeout List, because that’s a competitor that we might want to go against from Salesforce, and track which companies are on there. And then for every company that I’m gonna find on that list, I created a little loop here, I’m gonna go do that research and spin up that page.
That’s how simple it is, and after that, we can just run the agent, and it keeps going. So let me actually show you what’s inside of that agent. So this is how simple it is to set up in our platform, right? You give it a name and a description. And then you can basically type out an entire prompt for it to run.
And you can see here, there’s lots of information in here, and this looks like a lot of stuff to actually set up. I also use the AI to help me write this prompt, right? Who better to know what tools it has and what it has access to than the AI itself? So if you wanted to, you can even have the AI help you set up some of these things.
It then takes a bunch of things as inputs, so what company I want to create it for, and if I have a… if it’s already in our CRM, you can optionally pass in which Salesforce account this is for. This happens typically, for instance, if we do an event.
So we went to a B2B event, for instance, last week, and I would say out of the 2,000 companies that were there, about 250 of them stopped by our booth and said they are interested in something like this, because we targeted them with this specific landing page, and they got to see exactly what was created for them, and they wanted to kind of dissect what we did.
And you can see here, like, all the different tools that it’s able to use. In this particular case, it very much focuses on Salesforce, but it also browses and searches the web, takes screenshots, creates pages, looks at logos, templates, and then even hosts some of the files and uploads that.
Just to show you, like, what’s configured in our own account, we have over 500 agents that we run ourselves as marketing, it’s pretty easy to orchestrate, and they span across all of these different tools.
So, it can use all of our own products, but it can also use things like Google Analytics, Microsoft Outlook, so I asked it, for instance, hey, color code my calendar. It’s able to do that pretty well. And then things like Ahrefs, SEMrush, we use Profound for, like, geodata, for instance, ourselves, and then things like Bitly.
Everything that you would want to do as a marketer. We are very much focused as a platform on the marketing use case. There’s a lot of generalist platforms out there, like ChatGPT and Claude, but we generally think of those as individual productivity, which is very important.
But the hard part there is if you then want to share it with everyone in the business. It’s generally pretty hard to do that, right? You either need to share a document that says, like, what your top prompts are, or tell people which connector to install.
This is all across the entire organization, so everyone gets the benefit of all these connectors and things set up, and you get to organize and orchestrate these things for other people. So I’d like to say, for instance, that a good agent does three things. It identifies, recommends, and resolves.
So, for instance, if I created an accessibility agent, I might say it should find you know, where the problems are on the page, hey, you’re missing a bunch of alt tags on the page. Then it should recommend, and actually write the alt tag for you, and say, hey, this might be a good suggestion to add. And last, it should resolve.
It can actually use the tools to implement it for you. And if you feel confident, you can go all the way. If you say, okay, I actually want to monitor what it’s doing and review it, you might say, I just want it to stop at recommend. And then flag it to me, and I can take the next step of actually implementing that change.
That’s fully up to the user. And you can use it, like, you can use any other LLM. If I just go to the chat tab, I can ask it something like, what’s my website traffic? Actually, let me ask for the homepage. Specific? And it will go off, and you can actually see it as it’s using it. I can open it here.
It uses retrieval objected generation to actually pull what you have done with the system before. It knows me, right? So it has a memory of what I’ve asked it, what projects I’m currently working on, and then it even has specific instructions around how I like to work, what my team is.
And then even on an organizational level, it has instructions about who our personas are, what our keyword targeting is, all the things that I set up.
And now you can see at the last step, it’s used a tool to go grab this data, and you can see here it actually used one of those instructions, because we, as optimized as a business, want to be careful about what it is that we’re sharing. You’re getting a little sneak peek into our real site data here.
But I’ve added an instruction that says, hey, you should validate this data with the right people, including me, before you actually share it out. And every time someone asks something, even if you’re an email marketer and you just quickly want to know what traffic you get from the email.
Normally, you would have to go to me to ask that question, but with this system, you can just directly ask it, and it will respond in the way that I also like to respond. So, everyone in the business can now unlock that domain knowledge and get access to it.
So this has made a huge impact, obviously, for us, and not just how we speed up the things that we do within our own business, but also being able to do projects that we were never able to do before, right?
We are not able to physically spend enough time with prospects to create individual collateral for everyone in an entire ABM program start to finish. And now, with the system, we can scale all of that stuff up.
So if I think where the future of this goes, and what makes it easier for us to do this going forward, now I actually need systems, right, to manage all of these things. at scale. I physically cannot review 1,700 landing pages overnight.
So how do I actually find a way to surface those things to me directly, so that I can flag only the high priority things, and get a good list of, okay, here’s the improvements I should make to my agent, here’s one I should maybe add some more data sources to it.
And we’ve made that relatively easy in our product, where you can also flag, for instance, different to-dos and other things. So if I just show you that, this is our Chat agent, right? This is the dashboard that most people spend their own time on, right?
I’m not always talking to the AI, I can always pull it up if I wanted to in the sidebar, or approach it separately, but this is where I actually work on the tasks. So, if it needs to flag something to me, like, for instance here, I created a task where I’m working on a new blog article on my site.
This might be one of those ABM pages, right, that we’ve created. I might want it to actually come in and say, okay, I’m going to review it for you, and here’s some of the feedback that I give.
So here on the right side, let me collapse this to make it easier to view, you can actually see that when I created the article, it’s doing things like an e-at check, a link check, making content edits, it can do all those things directly, and instead of pulling you out of this, and then saying, oh, I now need to go to Opal, which is our AI, to actually review this, it just directly comes in and leaves a comment and says exactly what it is that I need to fix.
And if I choose to say, right, I said earlier, identify, recommend, resolve, you’re looking at identify. It found some issues with our brand guidelines in this particular article, and then I could go further and say, now I want it to recommend what to change, and it can even make those changes for me directly. Very helpful.
After all this work is done, right, normally we post a blog article on the site, we publish this ABM landing page, it actually goes out into the real world, so how do I actually get the data back? How do I know that something’s happening with it or needs to be improved?
We have an entire system that can create work requests for me, essentially, to flag certain issues, and it will actually add it to my to-do list to look at. My favorite part of this is when it tells me nothing.
If it runs through the workflow, and it checked 30 days later after the traffic, and it said, well, you’re actually meeting benchmarks, I’m gonna stay silent. So we just created an agent ourselves, which… On our platform, you can make your own agents.
That just says, try to be a sneaky marketer, and try to be very lazy, so it will only flag things to me if it feels it can contribute significant value, and stay silent otherwise. Very helpful. Cool. Are there any questions, by the way? Anything that you would like to see, or any… any initial thoughts?

Julia Nimchinski:
So many. This is phenomenal, Mikael. I’m just curious, in terms of, to your point of personalization, it gets tricky. And sometimes it’s really easy to confuse it with, you know, just generic content at scale. So how do you prevent it? You mentioned that, you reversed the process, but how long does it take to just, you know, the whole setup?
And yeah, if you could just share a little bit more about that.

Michiel Dorjee:
Yeah, and I could actually show you, like, it actually running, just to show you that I’m not lying. Like, I could show you here, it ran… oh, over 8,000 times now, you can actually see it run. And you can see a lot of these are running currently, if I just click open a random one here.
This is the reason why, like, it’s actually good, is you can see it use all of these tools in the background, and then it says things like, okay, I have the existing page, okay, I guess this one already existed. Screenshot taken successfully, just need to update that one. Okay, great. I just didn’t need to do a lot on this one, I guess.
Let’s see, could find a different one. Yeah, it’s actually browsing the web and finding the existing page, and it said here, for instance, oh, I found… I need to change this thing. The reason it’s good is because it knows how to do… the job well, and I can actually show you the background of how we did this.
We go to each individual team, and we say. what does good actually look like? I do not care about the current process, I do not care about how people do this today. I want you just to show me what the ideal version of this looks like. So, for instance, here. You can see our process.
We as Optimizely like to see ourselves in this little… I call it a bracelet. Kind of looks like that, but if I cut it through the middle and lay it flat, these are the thing… all the things that Optimizely can do, right? We can help you plan, create pages, optimize them, do all these things.
If you then go to each individual team and you say, okay, if we want to put on an event as Optimizely, for instance, you tell me exactly what steps you take to do that. They put all these post-its on, like, okay, picking what products to bring, who’s going to go there, which vendors do I use.
Then I want you to actually tell me which things either take too much time, are frustrating to do, or are not at the level of quality that you need. and you’ll recognize these are the same, like, we’re coming to the buckets of the little triangle that I talked about earlier.
So, when people actually plot which things are difficult for them to do, surprisingly, a lot of more interesting things come out, right? If I just directly go to a team and I say, hey, field marketing team, what would you like to use AI for? They’re gonna say, well, we want to create custom swag at every booth. I was like, okay.
That sounds like fun, and I would love for you to spend time on that, but what do you actually spend your time on? Oh, it’s half of your time, you’re creating lead list uploads for Salesforce, and working in spreadsheets.
So if we created agents to actually help them with that work, that’s immediately way more impactful for them than custom swag would be, which is the fun, creative work you actually want to be spending your time on.
So, this is kind of how we got to those use cases, and then how we monitor if we’re succeeding those goals, we can actually see a previous event that we’ve gone to, that we’ve brought products to, and then mirror, okay, does the AI actually do that well now, and pick the same thing that people wanted.
We’ve literally done blind tests with them, where we showed them two versions and said, pick the one that you like best.

Julia Nimchinski:
Love it. And in this new setup, who’s the ultimate user of this? Would it be the CMO, or more technical person?

Michiel Dorjee:
I love that question, yeah, because I always tend to say that, you create an email agent, not for the email marketer, but for the content marketer who’s, like, writing a blog post. All of a sudden, you’re now giving them that knowledge that they normally would not have had before, right?
You’re sharing that within the business, so… I think yes, like, all of a sudden, our CMO does not come to me anymore to ask questions about, like, Google Analytics, for instance. You just type into the chatbot, and then it comes back with the response. So, for sure, it’s everyone else that’s… that I’m focusing on. All the ones that are underserved.

Julia Nimchinski:
This is amazing. So basically, we had this trend where the chief, you know, revenue officers were becoming chief revenue architects, essentially, so we were seeing the similar dynamic we see amongst, basically.

Michiel Dorjee:
That’s right. I think everybody’s, like, super focused on building things for themselves, but, like, if I teach other people to fish, like, imagine how many fish I could catch, like… and that’s kind of the thing, is if we build these tools for other people within the business, we can actually bring everyone along on this AI journey, so…

Julia Nimchinski:
Mikael, we recently had, on our AI Summit, just a couple of months ago, we had the head of innovation, Nero, and they shared their agent development lifecycle. A huge part of that was the governance fees and the improvement.
You showed the decision traces and, like, how it… self-correct and, you know, the whole logic, but I’m just curious, your framework internally. How do you improve the output repeatedly?

Michiel Dorjee:
You can actually, as it’s running here, you’re actually seeing it just literally updated this one, but you can actually set particular runs as examples. If I literally click there, it sets it as an eval example, and you can see in this case, it said… it evaluated its own job, and then said, oh, I did the job 95% well.
So in any of our agents, you can always scroll to the bottom, and then set up any existing examples of previous runs as guidelines for it to adhere to. It doesn’t use that as part of the run, it does it afterwards. So it said, hey, let me go back and actually see what I did. And did I use the same tools? Did I use them in the same order?
Did I come to the same output? Oh, this one said that the page was not created and that there was a system error. Okay, I’m gonna flag that one and under-report it. And then you’re able to notify yourselves, like, oh, this one’s probably wrong, you should go look at it. Which drastically helps with hallucinations and other things as well.
The other part is then, if you would want to use it and say, no, I actually want to give the agent instructions to say. here’s an example for you to work with. You can actually just directly, if I just type in here. I can include any of those samples that I just included as examples, including how it does that exactly.
So even if I wanted to use that as part of the prompt, I can make that part of the agent. And that really creates this, like, great loop of it giving itself feedback on how to do things well, and increases the quality, and I then have a dashboard that literally gives me a little green bubble that says it did a good job at it, or not.
And obviously, we track the performance of it as well, right? We get the… if it works, and people read it and consume it, that means it’s doing a good job.

Julia Nimchinski:
Really cool. Michael, when you deploy this in mid-market enterprise companies, I’m just curious, what’s the process like? Do… would you recommend to… for them to just… a more staged approach, and to deploy it internally, and be the zero customers, essentially? And then, test it in their customers, or how do you… because, obviously, so much risk.
It’s involved here.

Michiel Dorjee:
Yeah, absolutely. So, I always like to say that, you know, in those identify, recommend, resolve buckets, always start with identify. So the first goal with the agent is literally just to find the right data, for instance, in this case.
Like, can it even output exactly what I would say before it even comes up with what to do with that data or create something? Let’s flag that first. And then afterwards, you kind of start building out these other phases of actually recommending and then resolving it.
So when we onboard a company, we actually ask them, hey, manually, literally, if you want to do it in a PDF or a Word document, take an existing piece of content that you have, for instance, and show me how you would personalize it, and what you would want it to look like.
And that’s a pretty resource-intensive exercise, but we only ask them to do that once or twice, just to show us, like, okay, you have a PowerPoint that you want to customize. Show us exactly how you would do this.
And then we, ourselves, kind of go back into our little AI cave, try to replicate that process and build those things out with the same resources, and then we come back to them and say, hey, is this… does this look right to you if I try this for a different company?
And after doing that a couple times, they also grow with their confidence in how the tool can actually do it.
If we find that there’s common issues, for instance, that it has wrong, like, oh, it used the wrong image, or like you saw earlier, mine flagged that it couldn’t load the page, we can just build in tools for it to flag those things as well, right?
By adding additional agents that do the review work, or even just hard-up building tools that say you’re not allowed to do these things. And to only give it access. So, for instance, compliance is a big thing there, right?
We see with a lot of customers, they’re like, okay, great, you’re using Salesforce data, but I don’t want everyone to have Salesforce data. So, we fully adhere to all of the roles and rights that the individual users have. So, if I’m not allowed to access Salesforce, neither is the AI.
and you actually give it your credentials, essentially, to do these things on your behalf. So inside of the system, and that also means that there’s a full audit trail, right? If I’m going to go back and see, oh, someone ran an agent and created thousands of landing pages, I can actually see that they did that. So if I go here, I can show you.
These are the connectors that are kind of individual connectors, and when I hook these up, I can see that, okay, it can only pull my calendar, for instance, I can’t accidentally pull yours, those sorts of things. And that’s how we kind of increase the… The surface area for it to check against.

Julia Nimchinski:
Amazing, and is there a master view where you can see all the logs and all the, you know, activity within the team?

Michiel Dorjee:
For sure, yeah, and you can control who gets access to that and what, but you can see, like, across the business, me, and you can see here Julie on my team, Rafi did something. You can see across all of those, kind of, what people are running, for sure.
And customers also get access to dashboards where they can see exactly which user used which agent the most, and how are they getting most value out of that, from a cost optimization standpoint, but also from a, hey, where should I put my time and attention and do optimizations kind of standpoint? All of that also comes out of the box, by the way.
We have a giant agent directory that a lot of the common things marketers want to do, from geo to brief creation to creating QR codes, everything comes out of the box, and you can actually use these as starting points as well if you want to start building your own agents.

Julia Nimchinski:
This is so amazing. I’m curious, Mikael, what you say is currently, based on all your customer base, is the most innovative marketing use case and case study, that most, you know, mid-market enterprise companies Still enough leverage.

Michiel Dorjee:
what I showed you wasn’t cool enough?

Julia Nimchinski:
to ask, I want to expand.

Michiel Dorjee:
I think some of the most interesting ones that people have done is where they didn’t only focus on, like, the marketer experience, so where they’re starting to build things to optimize their own processes, like, how do I do things faster, but they actually go into the customer experience, right?
So, some examples, for instance, are, you know, how do I put dynamic search, for instance, on the site, and give better content recommendations.
Or, we currently have a project, for instance, to tag and optimize landing pages so that on that page itself, we can show better promotions, for instance, like, which product should I show on which… which promotion should I show on which page? Essentially creating your own, like, ad system.
I think software has become so much cheaper and more accessible to people, that if you kind of know what you’re doing and you’re… a little bit dangerous with the tools, you can create some really cool and nifty stuff, so… And a lot of those find… make their way into the product, right?
You can see these are the ones that we’ve actually used ourselves, and then wanted to share with customers, so…

Julia Nimchinski:
And in terms of integrations, how do you work with, you know, all the traditional stack CRMs, warehouses, I don’t know, sales stack?

Michiel Dorjee:
Yep.
We support a bunch of things out of the box, so we have a bunch of integrations for, like, earlier, so Google Analytics and Salesforce and all these other things, but just in general, we support anything that has a webhook, email, remote MCP, Api integrations, we can work with all of those, and those are user configurable, so you can add those into the platform yourself if you want to.

Julia Nimchinski:
What’s on the roadmap? What are you allowed to share?

Michiel Dorjee:
I think what we want as a business is for people to do more end-to-end marketing directly in the platform, and currently, you spend your time in a bunch of different systems in a bunch of different places. we will build integrations to bring the tools there, right? We hear from some businesses, you know, we can only work with Microsoft Copilot.
We now integrate directly with those tools, and you’ll see more of those kinds of integrations with Slack and other tools as well. Where you can pull us into those situations. in this platform itself, we would also just love to cover all those use cases. So, for instance, multiplayer chat is what I would call it.
is, hey, you can collaborate on a document together, and then actually pull the AI directly into that, into the platform as well, where now I’m mostly just chatting with it directly, right? But adding multiple people in that could create much more rich conversations, so… That’s gonna be fun.

Julia Nimchinski:
And just to bring this home, what’s your favorite customer story?

Michiel Dorjee:
My favorite customer story, so we work… currently work with Bloomberg, for instance, and they’re thinking more about how to automate some parts of their process, for instance, with agents.
So, can we actually create agents that you know, take a particular screenshot and export it from the website and directly save it, so they don’t need to go talk to their legal team anymore. Or can we help them with page structure?
So… one that I’ve worked on, for instance, is, I have a video page, for instance, so when we do a webinar like this one, afterwards, you need to host the video, right? But normally what you should do is go back into the landing page, and then change the copy.
We use Optimizely Translate to, like, 4 different languages as well, and you need to generate captions. I could just stage out all those bits of work kind of end-to-end. Including the video production.
Literally now we can plan the studio to go record something, prioritize the ideas, load the initial version of the transcript on the teleprompter, we film the video, and then when it’s done, it already automatically uploads to OneDrive. We send it off to Descript to do the video editing, and then port it directly back into our platform.
For… to send out for the editor review. So I think just stitching all of those things together is, like, a really neat and cool thing that I’ve seen parts of somewhere, but not really in an enterprise setting where you’re still using all of the same systems, you know, that we’re all used to.

Julia Nimchinski:
Phenomenal. What’s the best next steps? How can I… how can I test it? How can our community test it? Where should we go?

Michiel Dorjee:
Yeah, of course. There’s a great page up on the site, you can actually see, I’ll give you two pages, so there’s Optimizely. dot com… oh, optimizely… Yeah, because I’m… No, come on. Optimizely.com slash AI. There’s, like, a slight lag on the typing. This is a great page to actually look at what the AI does and what things it can do.
That’s one page. And if you just want to see what agents there are, and what actions it can take, there’s a page called Agents as well. You can go look at that, and you can see the entire directory that I also just shared with out-of-the-box agents, if you need inspiration. And you can follow along.

Julia Nimchinski:
logical.

Michiel Dorjee:
Contact us, so…

Julia Nimchinski:
Amazing. Thank you. Phenomenal session.

Michiel Dorjee:
Thank you.