Generative AI: it's all about what you do with it

Embracing generative AI doesn't mean killing your business model

While Amazon has invested a staggering $4Bn investment in Anthropic, a company whose revenues are estimated to be in the low tens of millions, there are over 20,000 open-source large language models available on HuggingFace.

HuggingFace’s leaderboard covers the best-performing models, and they change weekly. Recent leaders have included The Falcon model from the United Arab Emirates Technology Innovation Institute, Orca from China Unicom, and Facebook’s Lllama models.

The HuggingFace leaderboard as of 26th September 2023

These models share the same features: the ability to answer questions based on a large corpus of training data and generate text virtually indistinguishable from what a human might write. With so many open-source - and thus free forever - models already available, the business challenge is less about building a bespoke model than about how to leverage them within your organization.

The poster child for adoption has been computer programmers, who have embraced these models to increase their productivity, with GitHub’s copilot now built into the software development environments used by most companies.

Screenwriters and actors, who probably think they have a better idea of how the story might end, have gone on strike partly because they are concerned that AI will take their jobs away.

While it might be possible for a virtual Tom Cruise to dominate the box office for the next 150 years, there are much simpler and more mundane applications of these models that will positively impact people’s daily routines and increase the efficiency of any business.

No one misses the typing pool or the post-room from the modern workplace, and if we think about it the right way, AI can have the same positive impact on our working lives and competitiveness.

Although much consumer research has moved online, many repetitive, manual tasks are still involved in fielding a survey and extracting insights from the data that comes back. Think about all the little decisions that we could hand off to a large language model.

How should we word a question about someone’s household income? What label should we give that question when we put it in the report? Which other question shows a strong correlation with household income? What brand is the consumer thinking of when they write “Maccy D”?

None of the applications might appear particularly ground-breaking, but if you imagine a workflow where those little decisions are handed off to an AI, then you end up with a process that offers much faster consumer insights in a timeline that allows faster reaction to market changes than most marketers can imagine.

The winners of the generative AI revolution aren’t going to be the people who train the best model. It’s going to be the people who best understand how to apply them in their business. The first agency to realize the cost and speed advantage will win market share from the incumbents. The technology is there, and it’s just about who can use it first.

The race is on.