Shawn Spooner Global CTO at billups recently built an AI model called audrai to assist billups, the out of home advertising agency. Billboard Insider talked with Spooner about what he learned.

Shawn, what is audrai and how is it benefiting billups?

The first version of audrai, built over two weeks of evenings and weekends, focused on two key workflows: matching client briefs to available audiences and markets, and quickly generating custom maps from audience data. When Shawn first built audrai it was saving 1 day a week of an analyst’s time. It now does the equivalent work of 3 full-time data analysts per month.

How do you maintain privacy and keep billups data from being open to all the world?

audrai does not use any publicly available versions of large language models (LLMs).  It runs either Vertex-based Gemini entirely within the Google environment or on Bedrock, which is an Amazon prompt entirely within the Amazon firewall. This approach ensures that the company is not using public services such as Claude or ChatGPT.

Where is AI strong and where does it have limits? 

AI is highly effective when dealing with well-known boundaries and tasks requiring one or two quick steps.  AI’s performance tends to degrade as the number of required steps increases, particularly if those steps demand incrementally more judgment (five, six, seven, or more steps). To manage data sourcing and hallucination, audrai is designed to cite the exact sources used for every part of an answer upon request, and every release is tested against a full suite of scenarios with known output criteria.

What challenges have you had from users?

Two main phases of pushback have been identified during the adoption of AI: initial resistance due to fear of mistakes or general apprehension; and subsequent lack of trust even once the tool is being used.  Users who lack trust often spend time double or triple-checking the AI’s output, consuming most of the benefit gained from using the tool, which is a harder challenge to overcome than improving the models themselves.

How can AI change out-of-home buying?

AI will increasingly integrate into workflow automation and shift the role of media planners. Instead of coordinating tasks manually, media planners will transition into roles resembling air traffic controllers, overseeing systems and confirming actions before moving forward, which allows them to focus more on client connection and communication. audrai will tie together measurement tools so users can ask about campaign performance, client expenditure, and future strategy.

What advice do you have for an independent out-of-home company that knows nothing about AI?

Start using it. Use tools like Claude or ChatGPT. Conduct deep-dive Google searches with AI.  Then start to treat AI not just as a search tool but as a conversation partner.  Push its boundaries and understand its reliability.  Don’t use AI to replace strategy. Use AI to execute the strategy, vision, and goals.

To learn more about how billups is using AI contact Shawn Spooner, CTO, billups at ai@billups.com.

To receive a free morning newsletter with each day’s Billboard insider articles email info@billboardinsider.com with the word “Subscribe” in the title.  Our newsletter is free and we don’t sell our subscriber list.

Paid Advertisement