Hurry Up And Wait - The Challenge of AI Workflows

One reason AI adoption is slow

Happy Sunday and welcome to Investing in AI! I’m Rob May, CEO at Brandguard. We teach machines to understand brands and branding and our product ties in a bit to today’s topic. I also run the AI Innovators Community in Boston and NYC, but hopefully it will come soon to your city too. And check out our podcast. We are always looking for new guests.

This week I want to talk about one reason AI adoption is slow. It’s the “hurry up and wait” challenge of AI workflows.

You may wonder why, with all the breakthroughs in AI - the cutting edge research papers, the LLMs, the cool new tech - why does it still seem like most of our lives are still not permeated by AI tools?

From where I sit as an AI entrepreneur and investor, the problem is that the advancements only affect one or two parts of most workflows. If it takes 7 steps to get a task done, and 2 of those steps are now rapidly better because of AI, all that may have done is create a bottleneck on the other side of the workflow. It becomes a “hurry up and wait” situation where you accelerate part of a workflow but then, just sit waiting on the rest of it.

This insight is actually what led us to the invention of Brandguard. If you’ve followed along with the whole story, we started off building a generative AI ad platform, but we ran into the issue of “hurry up and wait.” In many large companies, there are layers of approval and sign off, sometimes even legal gets involved, and so, if you use generative AI to speed up content creation, that content still just sits there for 3 days waiting on approval. That doesn’t really help you respond faster to market trends, news, or customer feedback.

On top of that, CMO after CMO told us that even thought it was cool we could create 5,000 flavors of an ad to make them hypertargeted and better performing, they didn’t have time to review 5,000 ads, and they wouldn’t send them out without a review.

That led us to realize that generative AI could only be adopted at scale we could teach machines to be the reviewers as well as the creators. And while some creation platforms are building in similar functionality, it just makes more sense to have a centralized and independent review process, so we dropped the generation functionality in favor of a focus on machine approvals for all customer facing content. When machines do the approval, the whole workflow can now move at AI speed.

As an entrepreneur, how do you play this space to avoid the “hurry up and wait” problem? You really have 3 options:

1. Wait on the rest of the market to provide solutions that can be matched together to accelerate the full workflow and all the steps. This is difficult because it makes the timing out of your control. But, if you focus on one of the final enabling steps when the rest of the market pieces are there, it could be a rocketship.

2. Do the full stack yourself. This gets around the hurry up and wait problem but, requires you to raise more capital and compete early across multiple product categories. It’s a great move if you can pull it off but, you also risk getting picked apart over the long term by best of breed point solutions.

3. Focus on the slowest step in a workflow. If a workflow has 7 steps and you can make the slowest step much faster with AI, there’s value there and you speed up the workflow even if the step after you in the workflow is still human-based.

Over the next decade, AI is going to transform most all workflows. That’s obvious. But investing in AI requires some insights into how those transformations will happen. How will the world move from the way something is done today, to the way it will be done in a decade? There will be lots of great product ideas that die along that path, because of the hurry up and wait problem. So be thoughtful about what you build and invest in, and try to minimize your chances of getting stuck in that hurry up and wait dilemma.

Thanks for reading.