From Watson to Siri, Alexa to Cortana, virtually every major technology company is testing the bounds of artificial intelligence as a personal digital assistant or other consumer convenience—and with all the attention paid to these tools, ad tech is exploring ways to apply this science to the art of advertising.

A panel comprising representatives from IBM, Pubmatic, Sentient and Ebay answered Jim Cooper’s questions on AI’s role in advertising and what impact it could have for digital ad tech in the future.

IBM Watson content and IoT platform’s Jordan Bitterman felt that, while technology’s role in marketing is sometimes overestimated and underestimated other times, the most important thing is “to move”. eBay’s Beck Kloss went on to say that it is most important to prioritise the capabilities which fit best for a particular campaign.  There is a lot of consolidation, which Pubmatic’s Rajeev Goel sees as a great opportunity to drive down costs. Kloss discussed how his company is always considering how to best marry man and machine at scale using AI capabilities and how to harvest signals for deep learning models to predict outcomes.

With so much data, it’s easy to be enticed to solve a great many problems, but Bitterman encouraged marketers to look at objectives, strategies and tactics to decide what to solve as sometimes there won’t be multiple opportunities to find the right strategy.

Sentient’s Babek Hodjat also discussed how it is important to pose the correct question for AI to solve: “It’s not about big data, it’s about good, relevant data.”

Talking about how to deliver the most relevant ad to a user, Goel described how his company processes tens of billions of adverts per day with huge amounts of data and impressions. However, it is introducing a ‘bid throttling layer’ to decrease DSPs.

Kloss talked about his company’s ‘gradient boosted decision tree’, utilises information about users, context, where the shopping journey begins and what tech to use, to make choices by two-by-two segmentation.

In order to make AI a viable marketing tool, Goel said that IoT was necessary, while machine learning algorithms can be used to predict how much different advertisers will pay for an impression. According to Hodiat, conversion rates go up 40% simply as consumers are more informed on what is available for sale, even on low traffic sites.

Talking about ‘computer vision’, the ability to recognise objects in images, Kloss commented that the “black box” of video could allow native, real time experiences.

Bitterman concluded by talking about the number of times clients have wanted something built from scratch as opposed to an off-the-shelf solution, encouraging clients to change only one or two things each year as opposed to everything at once.

However, the panel agreed that we are in the early stages of AI becoming ubiquitous and marketers must think carefully about their strategies.