Using AI to Manage the RFP Process

All right. My name is Janette Roush. I'm the SVP of Innovation and the Chief AI Officer for Brand USA. I'm joined here by my colleague, Skyler Clark. Hi, everyone. Skyler Clark Senior Director, Partner Strategy at Brand USA. We are thrilled that you could join us today, and we're going to be talking about how you can use AI to help you evaluate responses to your RFPs.

And I'm particularly excited about this one because I think it is really broadly useful to all of us. I know going through RFP responses could be very challenging, particularly if you have a large number of responses. And this is a way that people are already starting to use AI. But I think there's systems we can put in place to help us do it better in a regimented, auditable way.

And I'm also excited about this because this particular presentation, it's not in PowerPoint, I've used a Claude Code to build out a HTML file that has all of the content for the webinar. I am seeing a move in AI where it is becoming easier to use it for creating presentations. And I know that is something we all do a lot of. As I show you different things in here, I'll also point out ways that AI helped with actually building this presentation.

To kick us off, this is a quick reminder of the AI strategy at brand USA. Our goal is to lead global destination marketing into the AI era by using intelligent systems to connect the world to the people, stories, and places across the United States. So we're going to make discovery personal, inspiration, effortless, and travel decisions frictionless. And I'm doing that through three areas of work.

So the first is empowering our staff to be more fluent in using AI. The second is taking what we're learning internally and sharing that with the US industry to keep us competitive on the global market. And the third piece is to reimagine discovery. So how are we making the United States more discoverable and more bookable with AI tools?

And today we are focusing strictly on this internal use case with RFPs. So what is the challenge when it comes to evaluating RFPs? We know that those proposals that come in when you issue an RFP can be very long, like they could be 25 pages or more. And you might get five or 10 responses to your RFP, whether you're looking for somebody to create a website for you.

If you are looking for a sponsorship agency, if you are looking for a company to work with you on a new brand campaign, that can bring a ton of responses into your inbox. And that can take a long time to weed through. And then it becomes even more things to weed through. If you have multiple people as part of your RFP committee, which is how we operate RFPs at brand USA.

So multiply 25 pages by 10 responses and five people on a committee. And that is a lot of human staff time going to reviewing these responses. We don't exactly go to RFP school to learn what are all of the things you need to take into account. And that means that people who, you know, came up at different organizations or not in a position to look at an RFP at all, you're creating your own systems for how do you evaluate an RFP.

And so it means if you have a five or a 10 person committee, every single person is taking a different point of view when it comes to looking at the proposals that are coming in. So it's good to help give people a standardized way to approach that process. We also know that you are reviewing the ninth RFP with different eyes than you did the first RFP. And people talk a lot about bias in AI systems as we absolutely should.

But we have to remember there are also biases in the human system. So if you talk about recency bias or primacy bias that you remember or feel have stronger affection or the thing that you read first or the thing that you read last, and all of those RFPs that you read in the middle, maybe that just becomes part of a little bit of brain mush. So having a system in place where AI is assisting you can help to remove some of those biases that come from either recency or primacy. And it helps you also to catch the little mistakes, like the little errors and compliance with the rules you set out in your RFP.

It's going to help apply those consistently across all of the responses that you receive. And then finally, it helps give you, if you have AI evaluate RFPs alongside your humans, that gives you an apples to apples way of evaluating what the responses are. And I've seen this myself because people will provide those responses in all kinds of different formats. But when it comes down to the actual context of what we are judging, you don't want to be thrown off by, oh, this was a prettier presentation.

Or that they had a foreword that the other RFPs didn't include. And that helped set it up for me so I understand it better. Ultimately, like, yes, presentation counts. But you need to be able to cleanly evaluate the actual meat of those proposals without getting trapped in the actual artifice of the process. And so that's another place where AI can be very helpful.

If you're going to bring in AI to be a co-pilot with you in the evaluation process, what we want to do is create a system for this. So we want every RFP to be applied to the same prompt that is evaluating it so that every-- so it creates this equality across all of the responses. We have a system at brand USA that's very transparent. So I have talked to many companies that are using AI for evaluating RFPs, but it involves each person uploading those responses into their AI tool of choice.

And then they come up with a set of prompts or a way to have AI help them. And it might be different for every single person on your evaluation committee. So what we have done is create a system where it's one shared project that everybody has access to. Everybody or the one person will upload each RFP into that project against a specific prompt that AI wrote by looking at the actual RFP that you created.

And then that is used for everybody to read the same responses from the AI. So we're taking that black box piece out of it. And then, of course, it is still human-led. And so when people say you keep the human in the loop, this is exactly what we mean. You can use AI and read those summaries. And it's thought on how would you score the RFP against particular criteria.

But the AI doesn't get a vote in the process. You are the people who are the humans on the committee. They're the ones who get to vote. And so the goal here is to create an even a level playing field for all of the participants in your RFP so that they are all receiving the same level of care when it comes to having their response being evaluated.

For this project, we made a fake RFP. And yes, I used Claude to make this. And I will say it was an absolute godsend because I can't imagine how much time it would take the right fake RFPs and fake responses to the RFPs. So Claude did all of the work for me. And that project is going to be an RFP for Meet Beige County who is looking for an integrated destination marketing campaign.

So their current slogan is Beige County. We are here, too. And they are looking to upgrade their branding and their slogan as part of this RFP response. And then what Claude wrote for us were responses from eight agencies for this RFP. And Claude happened to write in different mistakes and different points of view so that every RFP, every response won't be evaluated the same.

So step one in this process is to create a shared AI workspace. And so here in this screenshot, this is using chat GPT. You can see kind of ghosted out here that it says 5.2 thinking. So when you are doing work like evaluating an RFP, you want to use the very best version of the model that's available. So this isn't where you want to save computing power or save money.

You want to use the most recent model that has been released. And you have the ability when you're in chat GPT. And here I'm actually going to-- you know what? I could do it right here. This is this project inside of chat GPT. So over here on the left, you can see where you have custom GPTs. And then right here, you have projects.

And so to get started, you want to just create a new project. And then if you want to access again, you just click on that beige county RFP. And then up here, what makes this really useful at an organization level is that this is the brand USA chat GPT account. And so we can access shared projects together. And the way that you do that is you hit this share button in the corner.

And then I can type in a name. It will bring up email addresses of other people who are on our shared chat GPT account. And so this is how I was able to invite Skyler to be part of this process with me. And she has the ability to chat and edit in this project. And then I can share this with her by copying the link and sending it over to her.

And then after she accesses it from the link, it will start to show up in Skyler's sidebar. And then what you were able to do next, once you are in this chat environment in this project environment, is to go to sources and you want to upload the sources that shared assets that you want to use in this project. Which here is a copy of the RFP and a copy of the scorecard. So I'm going to show you both of those assets now.

This is something that I had on code right for us. This is the original RFP for visit Beige County. Somebody get information about Beige County. What are their key assets, like neutral coffee grounds, the big school. The scope of work for this project. What their budget is and how it's divided out over different stages of the campaign.

They spell out that the proposals should be shorter than 25 pages. All of the elements that should be part of that proposal. And then what the evaluation criteria would be, along with the submission guidelines. And then you need to provide insurance and talk about your privacy and security. Terms and conditions of the RFP.

And then we've provided some additional materials as background for the agencies. So 11 pages, Claude wrote all of it. And it also wrote the actual evaluation scorecard. So these two pieces are what I have uploaded. And here we're explaining exactly what do we mean by each of these items. So that the human evaluators don't have to guess.

And the AI evaluator doesn't have to guess. And of course, I use Claude code to write all of this because this is all fake. But I'll tell you, RFP writing is the kind of formulaic writing that AI is really good at. So when we talk about using AI is not cheating at work. I mean for writing stuff like this specifically. Now, when it comes to writing content for your website, I really think content you like humans to read should be written by humans.

Honestly, the people on the other end of this RFP, they're using AI to help them read and understand it as well. So to have it be AI created doesn't strike me as a moral crisis. So now back over here in this actual document, the next step of what we need is going to be an actual prompt that we can use to evaluate each of our proposals. And so inside of this project, I asked chat GPT write a prompt that brand USA staff can use to evaluate and score each proposal as it's submitted.

And that we will submit each proposal as a fresh chat inside the project. So that piece is really important because that helps us overcome the recency versus the primacy bias. Just like humans remember what they saw first and last the most. So does AI. So you don't want to do one long chat and you upload every single one of the RFPs into the same chat because they still have memory constraints.

And you're not you're putting yourself at risk of not getting the best possible output if you do that. But if you put in one at a time and give it the same prompt every single time, you're creating more of a level playing field. So here it's confirming, yes, you should start a new chat for each proposal. And then it gave me a prompt that we could copy and paste into each chat.

And again, you want to keep the human in the loop here. So we were doing this for a real RFP. I want to read through this carefully and say, if I were training an intern on what to do with the intern under and this is the smartest intern in the world, but you know, they don't live inside my head. They don't know what it is, you know, brand USA does or a DMO does.

Is this enough information to help that intern evaluate each response fairly? And it's spelling out very specifically, like how the compliance review works. And you have to see, you know, go through point by point and mark if they met or didn't meet the compliance review, go through for red flags. These are the weight that you assign to each criteria.

Then you're going to total everything up. You're going to select one overall recommendation option. If you just qualify them, spell out why. And then say, what are the human next steps here? Because again, AI can help us keep a human in the loop. And this isn't, they're not using this process to say, oh, great. Put all eight responses into AI and then humans only have to read the best too.

The idea is that we still have to read all of them. But we're doing it alongside the scorecard that might show us what to look for. And then it's spelling out some best practices for us. And then also some ideas on how we can use this project to collaborate. Where everybody scores a proposal and then meets to align on what middle, better best, looks like for each section.

We have independent scoring. You then have a committee meeting to talk through everything. And then you come up with an interview plan so that for the next step of the proposals that that includes live interviews, that you are given good questions that you can then take to the interview process. And then here you can see I took that prompt and one at a time I uploaded each of the responses.

You can see that file here. And then this is the prompt that was given to us. And I gave that same prompt to each one of those eight responses independently. And then it's thought you can see here that it went through it for a minute and three seconds. It's evaluating based on compliance. It's looking to see if there are any red flags.

And then it's going through section by section and it's calling out examples from the RFP of why did it. So it's not just the number that you are understanding the thought process of why did it answer the way that it answered. And then you're saying section by section, what was the score. And then if you scroll down to the bottom, you see that it's checking itself to make sure that it included all of the components and what that total score is out of five.

And then it's telling us its score, its recommendation, the justification for that. Some questions that we could ask either in a follow-up email or a follow-up interview. How is this going to tie back in reporting to our stated goals? And so like giving you a really solid platform to move forward on. But this doesn't, as I said before, absolve us from having the human in the loop.

So I'm going to, these essentially, you will get a copy of this as a PDF file at the end of the webinar. But there's a screen shots of everything that I just walked you through. So because everybody's sharing one project, we don't have people doing those eight or 10 different times. It's one person who is doing it. And then they are sharing that project with everybody on the committee so that you are all able to see and read what did this AI prompt say about the responses.

We're going to upload those source documents, creating the master evaluation prompt. And as you can see here, we want one prompt for our proposals. We're going to evaluate each response separately. The AI is going to create that evaluation framework. And then this allows every vendor to get an isolated, unbiased evaluation of their response to your proposal.

And then any committee member at any time can go in and read the response. We see here we have the compliance check and red flags. So you don't get exhausted by the time you're reading the eighth RFP response and forget to check for something. We get the scoring bisection that scores summary with the suggested next steps for the human evaluator and preparation for the human work that needs to happen.

So I would say before we get to this step where we are comparing the responses, we want to bring that actual human evaluator into the loop. And that's why Skyler is here today because she is the human to my robot. So she is going to be the human in the loop showing us how she built an artifact and Claude Code to help the person running the process do all of the measurement and evaluation. That's right.

And I know like being RFP admin for various RFPs, it can be sometimes an added responsibility to your normal day to day. So keeping that in mind to, I think, as we were going through and testing how AI can help streamline workflows, we felt like this is a really good example of exactly that. So I will share a screen to one that uses the exact same example of the greater beige county that Janette has shared of what we did in terms of a scoring setup that we've created for both helpful for the RFP admin, but then also those who are part of the committee. So this is showing screen two was the output, and then I'll go more so that behind the scenes version of how we got here.

But basically, similar to the exercise that Janette shared at the top, what are what's the R FP for? Are we here to do? I'm really making sure the committee is grounded in exactly what we 're asking. So, in the very beginning on the homepage that we've created, almost like a mini site in some ways, the giving instructions of what is the process that we want the committee to actually perform.

So setting up exactly where the sets that we have defined, the category waiting, a little bit more definition behind the scaling scorecards we've created, whether it's very high in terms of a stellar response, very exceptional, where there might be elements of it. Maybe that's not so exceptional. All that's defined here for the committee member to really understand what they're rating. And then after going through a score, seeing what really would advance in terms of moving forward with the response that maybe comes into a next round human element of us all meeting, discussing what we liked and what we didn't like in a proposal, and versus others where maybe there were a lot of things that we weren't so sure of and wanted to discuss a little bit more before advancing.

So that's the beginning homepage. And then going into the actual scoring, we've created a place where it can all be very accessible for the committee member, where everything can be in terms of in one place of who are the individual responses, along with a reminder of the checklist and go a little bit more into the scoring itself, but also keeping a high level view of the proposal. So what's the overview that AI has found of that came across in the proposal? Any key highlights or areas of concern?

And then as a human element, looking through, actually clicking through one of the proposal responses and being able to access it from that same link. So everything's in one place. So you don't have to reference a bunch of different documents or folder structure. It's all here. So I'm going to go through as an RFP committee member, and I will do one form.

We'll do Horizon Digital Partners. So I'm getting a sense of like, what is the overview here? Maybe I have opened exactly what Janette was sharing earlier, the chat GPT function as well and seeing what summaries came up to the top and just kind of going side by side of a screen share of what is really something I wanted to make sure I'm understanding in terms of the proposal. Here we have, again, a reminder of the steps that we had on that first homepage and then we identified what would be the part of the compliance checklist.

So where are things that are must from a compliance standpoint that we need to keep an eye on? Now you'll notice here, this is pretty lengthy. So this is just me doing a first round with Claude to create the compliance checklist. But there are going to be things on here that are maybe more deafness in terms of things that we would usually look out for in terms of timeline, for example, team allocation.

But there might be a few things on here that maybe aren't so relevant for us to have that are kind of look more just taking up screen areas like submitted by the deadline and that PDF format that may not be as necessary. Below that is also red flags. If a budget doesn't add up completely, if there's just noticeable things are missing from the RFP, this can all be checked. And then below actually getting to the scorecard that Janette actually shared through earlier and having that in here.

So it does a lot of the waiting and a lot of the actual scoring taking on one place. So I'll go through and do a couple of tests rounds here. As you can see some of the various sections that might be have the higher weights versus others, like the creative capability being a little bit lower than the proposed approach and mythology to a couple of more in terms of measurement or reporting approach and the references as are a little bit lower on the right seats. So I do that to save evaluations.

And I can actually see my safety evaluation that I just did. So if I click through an edit here, it will take me back to this page. And then I can go through and do the exact same for another response that were received like the copers age, for example. So that would all be stored at the bottom where I can see all of my various scores.

Another area you can also see where the results will be are also up here. So we 've actually created where you can actually start aggregating. Let's say it's myself and Janette and maybe a couple others are part of our committee. All of these will be aggregated for as an RP admin for me to see and understand really which responses and proposals are really rising to the top versus ones that might be more of a discussion and can also export CSV file to and save it.

And then the format that you would prefer for future reference. So going back, that is kind of the everything kind of housed in one place for the evaluation. Now what I did to get here, I started setting up what was called an artifact. So that's on the side screen here. You can see here, artifacts, they're almost second to last option.

And that's where you will start seeing as well, like artifacts that either are more inspiration, that Claude just has set up or ones you can create your own. And so you would create your new artifact and that would create this page essentially of what are you looking to do. So I've done a couple of RFP. So I took instructions I've done for previous RFP.

I was able to pop it in here along attaching files that Janette created, for example, of the RFP itself as a source of truth along with the scorecards. So they were able to see and get us ready for today's webinar and created a blueprint for our use. So knowing that what we wanted to create, I asked it to create almost like a dummy version of what they know from both the RFP instructions we were created along with setting up a blueprint structure that we could use for today. And so they went through and started creating the dummy version that we have.

There are elements of it that weren't working at first. So I wanted to make sure that buttons were working. So sometimes there's a little bit of that back and forth dialogue that you would have with the artifact just to do some of those tweaks. You can do the same exact thing when it comes to like areas where, like I said, that compliance checklist was very long, like we can make that shorter.

So it's more relevant and more top of mind for elements that you want to have the committee really keep a really focused eye on from making an element and then seeing other areas to have like how to hard code it in or I don't really want to hard code it in is another way we can have it work. So there are just some of that back and forth conversation with myself and Claude, when we'll all caps here on accident, by making sure that there's the right files are being added to it. And then you can see here just making sure that they're able to access from a Google Drive standpoint, accessing the right RFP responses that we also created as part of the dummy proposal responses. So this is a similar way of just from a conversation standpoint, just playing with Claude I think is really helpful too.

You'll notice things either maybe not working immediately and that's where you start just kind of calling that out from a conversation standpoint with Claude to make sure it's working the way you need it to. So that's how we got to that RFP scorecard, almost mini sites, if you will, that can be accessible for the other committee members. Yeah. And then you just hit that share button on it in order to share it with other people participating.

Yes, that's correct. I'll actually share screen real quick again. So you can see at the top here, I'm very happy with this last round. So I went ahead and you can share the artifact by clicking that share button and then copy link. And then that link will take you to this view. In the olden days, like a year ago, when you were doing evaluating RFPs as like an administrator on the project, how did you do, how did you co-late all of the scoring?

So it usually would require multiple different Excel sheets, taking it all in and trying to aggregate it with either various Excel formulas or something set up for me manually actually setting up the what is coming up as average scores across all the responses. So if that's multiple proposals, for example, with multiple team members, that can be really time consuming. And this was great because it did it all for me. The other great feature about this that we've heard from previous committees of having this new setup working is the fact that everything is in one place.

So it's easier to access one link versus multiple links of there's a box folder or some type of folder structure that you submit your proposal responses into. And then another box link to submit your scoring, for example, and just it can get a little lost, if you will, if there's various links to go into. So having it all in one place, I think it's been a really nice feature that a lot of the committee members that we've done this with really enjoying. Yeah, I will say as a RFP committee participant in the last few months, the ability to like it's just more fun to fill out a web form than it is to open up 10 different Excel sheet.

And then remember, oh, what was the criteria for this particular item and go back to a separate document that is the scorecard that has that information. So it's like, it's more convenient and a better experience for you is the admin, but also for the participant, which feels like a bit with. Definitely. This is just spelling out that collaborative scoring process that Skyler just walked us through.

So instead of five committee members using five individual Excel sheet, plus the admin manually compiling scores, which then creates a summary spreadsheet that doesn't have any transparency around it. Like this, you can set up transparency. So because the artifact, you are building it fresh every time just using and it 's no programming, right? Like essentially, Skyler's created a website, but she didn't have to be a programmer to do it.

She just had to understand what the goal was and what she wanted that final outcome to be. And then just in plain language says, is this something you can help me achieve ? And the AI puts together piece by piece. And when it's not exactly what she's looking for, she's like, actually not that would you change this piece and do it this way instead.

So that it could include allowing a committee member to see how other committee members scored someone or what comments did they write? Or if you didn't want to share that, you could hide it so that only the admin is seeing that information. So it's like it's completely flexible for the process, not just that your organization, but for that specific RFP and that is specific group of people at your organization so that it's completely customized every time you do the process. And so let's remember again, what do the humans do?

So where the AI can help is reading and summarizing all of those proposals so that even if you read through every single one, I find that I'm not going to remember the ones I read in the middle. So to go back and get that quick. Oh, that was the one with this idea. Having these AI summaries can help you do that. It's a great way to quickly check compliance.

So you're not the one doing it by yourself, catching budget math errors. Now again, we know AI isn't perfect at math either. It depends on whether or not it's writing a Python script to do the math. So still having another entity in the process to help with that is very useful. Seeing can it catch mistakes where it's just giving you a copy and paste response to your RFP that includes old client names in it.

It can help to create those consistent comparisons and build the scoring tools and the dashboards for you. But humans need to look at the cultural fit. We need to look at the level of trust that we have because it's easier than ever these days for something to feel like it 's perfectly aligned on a artificial fake AI sense. But is there any real depth behind the work?

That's what you have to understand as the human in the loop. You have to have that strategic judgment of is this approach right for our moment. You have to have the risk tolerance. So there's going to be that say you're looking at this specific RFP and they have instead of full service agency or they doing this as a one off project, which approach is going to be correct for your needs at this time.

I think again, gut check, I wouldn't trust an AI scorecard all on its own to filter anything for you at this point because also the filter is only going to be good as the instructions that it wrote with you. And you don't always know that those instructions are actually perfect until you actually get in the process of reading through what did the AI say? What does the actual proposal say? And at the end of the day, like it's human beings who are going to be affected by the decision.

So you want human beings making the decision. But a lot of the heavy lifting here, you only have to do once. So Skyler created one time the main template or the artifact that she uses. And then she can just go back and say, Hey, look at that artifact. And we want to tweak it a little bit for this new RFP, which is much faster than if you have an excel sheet system, you have to go back in and paste in all of the unique things for that particular excel sheet every single time you're starting from scratch.

Like it says for a new RFP, you just point your plot account at a new set of document and you're ready to go. So it's one person setting everything up with the entire committee benefiting from that time investment. And I think what's really good to remember here is that we have many processes that could benefit from this same approach. So anytime that you're given multiple options and you have one set of consistent criteria, this is the approach that you can put into place.

So from a hiring standpoint, again, this is something that you want to talk with your legal team before you proceed, but because people definitely will have guardrails around, oh, AI and firing. But think about having, you know, a transcript of an interview that you could compare to a scorecard that you are maintaining that you created with the hiring manager and the HR team to evaluate which criteria does a candidate meet for a particular position. And that's an area where if you rely too much on, you know, vibes or, you know, did we have a connection during the interview and you don't rely enough on the what were the actual things we are looking for to have this outside betting of that the transcript match the needs could be an interesting approach. Vendor reviews.

So not just hiring new vendors, but when you're analyzing, you know, how their production has been over the past 12 months, this could create a repeatable system for you. And I think grant applications, particularly since I know there are DMOs and state tourism boards who've distributed grants, this is a great way to help give you a little more consistency across that process. And so if you wanted to get started today, they'd just decide what is the next RFP you will be doing where you want to align on this approach. Set up the project, you'll assign one person, probably the administrator for that particular RFP to build and upload the RFP, the scorecard, and to create that evaluation prompt.

And then you're going to create the artifact that Skyler walked us through, invite your committee to share in that project. And then everybody is using the same baselines and the same set of roles to evaluate the responses. All right. So we have a question about AI use in grant applications. Some grants discourage AI usage in the applications.

If we use it for assistance, can they tell? Is it best to disclose? How do we navigate this? So I will say, and I'm sure that anything that requires an application anymore is becoming tricky because just like people will say they put out a job posting and immediately get a thousand jobs because candidates can use AI to apply to jobs and to personalize their resumes and the right cover letters.

And I'm sure it's the same thing with grants. Follow the roles of the grant. But also, I would hope that the grant application just doesn't say no AI. Like AI and what piece of the grant application, I don't use AI to write things that I write, but I will use it to help me put structure around what I'm writing. You know, it also comes down to does your AI tool have access to the information that you could use to respond to a grant application.

If there are like some details and facts that live in a folder on your computer and you're using cloud Cowork, which can read folders on your computer, you can say, hey, pull out all of this information for me so that I have it in one place. Or I have to provide a narrative of how will I use this funding to benefit my community? Give me an outline for what would be really meaningful. And then you could go and write the actual narrative.

There's a lot of ways that AI can be stilted or use negative constructions, things that like when I'm reading AI writing, I instantly know it is. But the person who used AI might not know because they only use AI once a month. So if you're not in it all the time, you won't know that you are setting yourself up for failure by using it to write. So I think if you're writing something for a human, we all have little quirks that makes our individual writing interesting and unique to read.

An AI will also gravitate a little bit. It will kind of talk in a grandiose way. And then after you read it through a couple of times, you think, oh, it didn't, it didn't actually say anything. There wasn't really any substance there. And I bet grant applications are a place where that is a big problem. I'm sure there's a big problem with AI submissions and grant writing, using it for the grunt work, and then keeping your quirky individual self as the actual writer of the content, I think is the real wit.

I'll also say that's an area where in RFPs, I don't care if you think I'm a good writer or not, like that in utilitarian work. I'm not here for anybody just to subscribe to my RFPs sub stack. All right. Skyler, did you have any final thoughts that you wanted to share on the RFP process and how using AI has kind of changed it for you?

Yeah, it's definitely been really helpful and all the hours in the beginning and set up, but then also during when we're going to the scoring. And I always love that everything you say at the end is just like using AI for work is not cheating and just a great example of it's not cheating. It's been more of a really helpful resource and tool for our team. That's great.

All right.