‘Doing AI’ before AI is done to you

Faster, cheaper…better? While AI offers the strategic opportunity to reinvent the products and services that brands offer and the way in which they market them, its immediate impact is much more likely to be tactical, driven by cost-cutting. In-house teams are under pressure from C-suites with LinkedIn-driven, unrealistic expectations. They are balancing the immediate benefits of short-term, tactical applications while trying to preserve the headcount that might unlock the bigger opportunities that the technology could deliver.

They lack access to data, are hidebound by cumbersome approval processes for introducing new tools and often operate without an overall AI strategy for the businesses that they serve.

BUT, there is an enormous opportunity for in-house teams to own the AI narrative, become centres of AI excellence within their organisations and go beyond their service role to drive growth.

IHALC’s Operational AI breakfast event on 8 July, in partnership with ITG, provided an opportunity to review the discussions from the past six months of our dedicated AI Working Group, as well as hearing expert contributions from ITG’s Lauren McBride and Ian Hudson.

The ensuing round table discussions revealed a picture of in-house agencies balancing the ‘Fast’, approach to AI which results in tactical applications to drive efficiency, and the ‘Slow’, more strategic approach which seeks to embed AI across multiple functions as a partner for innovation.

In-house leaders reported a primary focus from CFOs and others on AI as a means to cut costs. This comes against a background of in-house teams under immense pressure, with asset volumes often tripling year-on-year and expectations for AI to deliver results quickly, despite limited resources. Moreover, senior leaders often have unrealistic expectations, viewing AI as a simple “on” switch, without appreciating the complexities of introducing it successfully. And there’s a clear drive to “do more” with existing resources, with AI seen as the solution to fill gaps when new hires are not possible.

Aside from GenAI, in-house teams report using AI to:

• Streamline briefing processes: condensing briefs, linking to assets, suggesting language, and automatically feeding out follow-up questions for missing information.
• Client feedback analysis: ingesting and analyzing feedback to provide a “constant picture” of client satisfaction and pinpoint areas for improvement.
• Internal knowledge management: developing chatbots that answer questions based on internal guidelines and operational information.
• Asset management: identifying risks, brand guideline adherence, expired permissions, and performance metrics for assets.
• Automating low-tier creative tasks: templated asset creation, format changes, and “data wrangling”.

We’ve reported before about the major barriers to AI adoption in-house, such as lengthy and cumbersome approval procedures for tools, worries over IP and copyright, quality control over output and a lack of training. Additionally, there are major issues around data which are inhibiting AI adoption.

Broadly, this can be divided into two areas: access to and usage of data to power Ai systems and inform decision-making and the data relating to existing assets needed for effective automation.

On the first point, our community reported a widespread problem around access to both first party and unstructured data. Access to first party data ought to be a major competitive advantage for in-house teams versus external agencies. However, a majority of attendees did not have direct access to such data, which typically resides elsewhere in the organisation, or, at least, were not able to use it to aid creative decision-making via AI tools.

When it comes to assets, a significant opportunity is seen in training AI models on a company’s libraries of owned imagery. This allows for the creation of “custom models” that remove external input and allow companies to trace the origin of the imagery, addressing legal and quality concerns. However, a major impediment is the complexity of mapping out permissions and usage rights for existing assets. This includes varying requirements for photographer credits, location credits, specific agency agreements, and limitations to certain campaigns or stylistic uses. One participant noted “so many variants on our permissions or use cases that we’re struggling to even map that all out”. This difficulty in organizing and categorizing existing data significantly impacts automation processes and further limits AI use to “ad hoc” or “tactical” applications.

Nevertheless, there is recognition of the opportunity that AI presents for in-house agencies. For those who have ambitions to become a strategic partner to the business and not just a service provider, AI-empowered creativity can drive new product innovation. We’ve already observed this in workshops, where small groups use market and customer data to get to new product ideas rapidly which can then be tested using simulated consumer research groups. Going from product idea to naming, packaging and campaign ideas can take hours rather than weeks.

With the right data and tools, aligned to a coherent AI strategy, in-house teams have the chance to ‘own’ the narrative around the use of AI in an organisation, seizing the opportunity rather than playing a defensive game of piecemeal cost-cutting.  A year ago, respondents to our AI Survey hoped that AI would allow them to ‘automate the mundane’ while freeing teams up for higher value work. The early signs are that this is not happening. Rather, tactical uses of AI are resulting in a loss of headcount in the pursuit of short-term efficiency. It may well be that the people in your team currently don’t have the skills for your future AI strategy, but if you lose the headcount now, chances are, you’ll never get it back in order to bring in the new skills you need.

We keep hearing of calls from the C-suite to ‘do AI’. For in-house teams, it’s a case of getting ahead of what that could mean, turning a potential threat into an opportunity, ‘doing’ AI before Ai is ‘done’ to you.