About Viooh
Viooh is a leading SSP in programmatic OOH adveritisng. In plainer terms, it provides the platform to Out-Of-Home media owners—Like JCDecaux, our parent company—to book and execute programmatic campaigns. After years with a product centered around inventory management and campaign ops, in 2026 we set ourselves to creat our first agentic offering.
The challenge
Viooh is an Programmatic Out-Of-Home advertising SSP. An AdOps specialist will use it to filter down their inventory to create the campaigns that better respond to the briefs.x=
When rolling out AI capabilities to our customers, a natural language interface can no longer be considered a clevers solution but a starting point.
The language of our users
Up until today, the sales people using our product would have to decode the briefing they are getting from their customer into an selection on their inventory. Should they focus on big cities or smaller ones? Shall we focus on bus stops or will they do better on Billboards? What about Birmingham? Shall we include it or exclude it from our campaign. With the traditional UI users have to speak the language of our database to apply those filters until they reach a proposal that can make the client happy.
Bolting a chat box to replace this interaction may be a good first step but it is not enough anymore
Chat is a good on-ramp into AI for an audience that is not the most tech savvy. Also replacing all the clicking in our filters with some typing specifying the selections adds some efficiency gains. However, if this is all we do we are leaving a lot of intelligence on the table.
The architecture is the UX
A trully intelligent sollution should be able to speak the language of our users instead. What if the user could type:
I want to target 500,000 NBA fans in Portugal during Q3 2026
Very easy to imagine, not a lot of screens to mock up, and a difficult but achievable technical challenge. This is how we did it:
- We partnered with Displayce, a Burdeaux based sister company specialised in providing the data layer that will translate data to locations.
- Built an MCP to expose this data through an LLM.
- The seller agent, our front end product would manage user requests routing them through an LLM in our server to recover the data from displayce.
Doing this, we are not only enabling our users to do more natural queries but they also get to choose their own tool. Whoever prefers to connect their LLM to the Seller Agent’s MCP can directly do so.
Interesting on its own terms but even more relevant regarding what it enables for the future.
A vision beyond the chat
At Viooh we understand the future of AI professional tools is headless. UIs and frontends will become less and less relevant the more we can use LLMs to interface with these services. An LLM-centric world also seems to have short legs. Internally we’ve experienced going from the excitement new AI tools can bring to being almost burned out by the amount of typing they required.
We are still in the Terminal era, but we can think beyond.
With this product’s architecture we can rethink what an AI harness is. Evolving from just the interface where the user types their commands to the LLM, to a system designed to harness natural language wherever it happens.
Hi team, please set up a campaign of 200,000 impressions across Lisbon, Porto and Coimbra, running 15–30 September.
When speaking to sales people like the ones that would use this product, most of them iron out the details of the brief in an email conversation. It feels redundant going to an LLM to type everything again when that conversation could just be forwarded to the platform and have the proposal there waiting for you when you open it.
The result is a harness that meets natural language where it happens rather than the other way around.