GenAI is Changing How Agencies Think About Automation

There’s been a lot said about what AI will do to agencies over the last couple of years, as tools like ChatGPT, Claude and Gemini have massively increased in popularity and usage. AI evangelists like to tell you on social media that very soon agencies will be a thing of the past. A lot of what is currently being promised by “AI” was already available under the guise of “automation”, it was just never fully embraced.

I've spent most of my career working on automation projects within digital advertising (for agencies, advertisers and tech companies) and it's been very interesting to see the mindset shift in the post-GenAI world.

ChatGPT and the like have broken into the mainstream and captured the imagination in a way that classic automation never did. Maybe because of how visible their outputs are compared to machines completing tasks quietly behind the scenes. Or maybe because GenAI has created a genuine paradigm shift in how things can be done using computing power.

Either way - over the last couple of years, we’ve moved from just forward-thinking advertisers and agencies wanting to automate and improve processes, to the entire industry needing to be seen as “using AI”.

Before this mindset shift, I heard and saw many instances of automation not being a top priority, perhaps other than to lighten the load on overworked and unhappy campaign managers. In some cases this even stretched to not wanting to share some of the work being done, in order to avoid awkward questions from clients paying based on headcount.

Now agencies and other companies are suddenly jumping to implement AI solutions (there’s no shortage of people selling them), and a lot of the time by doing this, they’re using a flamethrower to light a candle. A lot of what is being done nominally “with AI” are actually tasks that could have been automated at any time in the last 10+ years with scripting or RPA (as the forward-thinking agencies were already doing). This shows a fundamental misunderstanding of what AI (especially Generative AI) actually is and what it is, and isn’t, capable of.

Previously when scripts and automations were doing logical, deterministic work behind the scenes, there was an inherent mistrust of what they were doing because there was no visual component to what was happening. Now, with LLMs being watched “writing” reams of text on screens, even with them being inherently much less deterministic and more varied in their outputs, counterintuitively people seem to put more faith into the results. That’s the biggest mistake I see people make (see the Deloitte vs Australian government case) - LLMs spit out a huge amount of content, and some of it’s great. But a lot isn’t, especially based on simple prompts with little to no relevant context, and often people don’t bother to interrogate the output thoroughly.

At this point, you may be thinking that I don’t believe AI will have as big an impact as a lot of people are predicting, but that’s entirely untrue. I truly believe that the recent (and coming) advancements in technology will fundamentally change how we work, but only for the positive if they’re properly understood and correctly applied. The companies I work with understand that simply throwing a load of data at ChatGPT and expecting it to produce a good media plan/optimisation suggestion won’t return anything worth the amount of electricity used to produce it. The AI tools that actually work are those that are built with the knowledge and experience of those that have been there and done it as an intrinsic part of the whole system. They operate within a specific context and align with existing processes - with human oversight wherever it is necessary.

With these changes there are two universal truths agencies (and everyone else) have to accept in order to survive:

• Everything that can be automated, should be
• Not everything can be automated

Andy Oakes (New Digital Age) wrote in a recent email: AI isn't going to take away the need for external resources. It's going to change the type of help businesses need and agencies offer, shifting away from tasks in-house people didn't want to do, to becoming more trusted advisors and partners. I thoroughly agree with this.

Agencies can still have a place as those trusted advisors. They can advise on what should be done and how it should be done for each specific case/customer, based on their cumulative experience/expertise. What they won't be able to any more is charge for doing the actual hands on keys work, that world will go away. There’s still hope for agencies, they just have to get the technology right, starting yesterday.