Introducing: RoarOS Beta for Brands

RoarOS is our flagship software product here at Shortlist. It's an AI-native platform where brands can activate, operate, and measure their influencer marketing campaigns, all in one place. We built tools into Roar's in-app agent that specifically address the painpoints we face when running our own campaings, things like creator sourcing, automated reporting, post analysis, creator relationship management, and more. Now, RoarOS is available for brands everywhere in a public beta.

Rather than sell you on the product itself, I'd like to take some time to talk about the history of how we made RoarOS, to enable the brand-side of the creator economy to finally use AI in their daily business workflows. The process of us figuring out what Roar actually is made us arrive at some interesting conclusions for how we think about building product in the AI age. Let's get right to it.

Building the right product requires finding the right problem

The earliest version of RoarOS was actually just called "Shortlist Dashboard". We created it to serve a very specific purpose; we were running a multi-brand creator campaign to promote products across a distributed collection of "Wishlist" posts on TikTok. The idea was simple: brands partner with Shortlist to give their products exposure, and Shortlist partners with creators to seed these products in "wishlist"-style TikTok carousel posts. We thought about every angle of the campaign upfront. How much we were going to spend on influencers, how many posts we wanted, how many brands we could work with, etc. The one thing that we didn't have a good answer for was how are we going to report campaign progress to partnered brands?

We could have built software to handle any of the other above open questions. And we did; just not in the form of a product. We made custom forms for creator registration and product selection that helped us manage the campaign operations. For our partners, we wanted better than a PDF delivered via email months after the campaign ended. What if we could give them a basic login to a basic dashboard that, when they visited the page, would tell them the real-time engagement stats over time, and which posts were actually driving organic traffic?

The problem that we were solving for here wasn't anything new to the world of software. Having scheduled jobs tracking specific post across IG and TikTok wasn't very hard to build. Instead, we were solving for trust. The source of worry for marketers rarely comes from how a campaign is actually doing. The worry stems from the lack of real-time understanding. It became very easy for brands to trust our process after we handed them the keys to a central place to see how our content was performing any time they needed to check-in. Even after we hit some snags in the first week of the campaign with some of our content being "shadowbanned", we found that brands actually didn't mind as much because they could easily verify our claims.

Afterwards, our product philosophy shifted dramatically. Sure, we still wanted to create simple solutions to complex problems. But we began to focus on the right problems for our ICP; themes like trust, activation, and organization. As long as we could keep delivering value that hadn't been realized by marketers yet, we knew we were building in the right direction.

This also gave us a lot of confidence to go after our dream of becoming an independent creator-led growth agency. If we can give every customer access to this platform, it tightens the product-feedback development loop while still making our partners feel secure in their decision to work with us.

The birth of our AI social marketing agent

The next problem that we tinkered on, following trust, was intelligence. We suddenly had this massive volume of post snapshot performance data in each of our customers' workspaces. How could we enable our customers to derive meaningful insights from this data, beyond just "what happened and when"?

Rather than build some neat dashboards, we decided that building an AI agent was the right product decision.

In the original Shortist pitch deck, we argued that creator marketing is entering its "third phase". We still largely believe this to be true. The first phase was brands asking for access to creators that they could contact for partnerships; which spawned successful companies that could do just that by selling access to filterable spreadsheets. The second phase came about when suddenly, everyone had access to talent, but nobody had experience with the operations side ("how do we actually run a campaign?"). This too spawned industry giants like GRIN, Sprout Social, Asipre, and other players.

Now that anyone can run a campaign, the byproduct of all of these campaigns is heaps and heaps of data. Content analytics, creator relationships, hooks, hashtags, all sitting unused. AI agents, which are very good at reading unstructured data and performing workflows, simply weren't ready yet to be deployed within marketing. Today, they're more than capable of processing this data and extracting real answers to hard questions. Why did one post outperform another? What are my competitors doing differently than me? These are all questions that could be answered with sufficient data. And RoarOS, as we know now, began as a data warehouse. Implementing agentic capabilities on top of the data was also, for our team of experienced engineers, not very hard. However, it enables operators to get answers to questions that were previous buried inside of seemingly disjoint data.

After building our agent infrastructure, we could suddenly add tools that extended our own taste and workflows. Our personal creator sourcing methods became skill files and web search tools. Influencer pricing calculations became a routine sequence of tool calls, too. Finally, we could productize the operations that made us so effective, and deliver those capabilities to everyone.

In a world where products are all becoming a chat box connected to a LLM, your edge as an individual, and as a business, lives in the domain that AI struggles with. Claude by itself isn't great at performing the verifiable tasks that marketers do every day. Not because it isn't capable, but because it has insufficient tools, data, skills, and weights. When you begin to imbue your own knowledge within homegrown agents, that's when you start to see AI unlocking real economic value. Our social marketing agent still has a lot of to learn, and our architecture needs time to mature so we can handle the scale of cloud agent rollouts that we anticipate. We're betting big on these agents delivering massive opportunity within creator marketing. To enable anyone on earth to get their brand or product seen by audiences everywhere.

RoarOS, today

We're officially launching RoarOS for Brands. We want to give brands access to the very product / agent infrastructure that enables us at Shortlist to run effective campaigns. It's our hope that opening up access to this tool will provide us with close working relationships with agencies and indie growth hackers who can help us iterate on a better product experience. Above all else, we want everyone to feel like they can safely experiment with their creator marketing strategy.

When sourcing creators takes minutes instead of hours, you can afford to take bigger risks with the types of creators you reach out to. When the effort required to run campaigns begins to collapse, you can also freely experiment with campaign structure and payout incentives. Relentless, free-form experimentation lets artists and creatives take bigger risks, and to take a risk is to start something meaningful. Shortlist and RoarOS can handle the boring parts of your marketing process, so you can focus on the all of the elements that actually matter.

If we enable even one person to take bigger risks with access to better analytics and intelligence, then everything we built was all worth it.