We saw the same cycle play out with most of our clients, so we changed our initial approach when we take on fresh projects. Here’s two quotes from separate scoping sessions:

“It's not as fancy as like LLMs or any of this new generative AI stuff. It's just identifying patterns in the data.... If someone had enough time to look at [all the data], they could probably figure it out, but I don't think the team has the hours.”

- SOW Review Session for Client #1’s First Project

“Picking the builds that we can get working, like, as soon as possible to see that value and then taking the time while working to investigate expanding elsewhere. Just makes sense.”

- Discovery Session to Decide Client #2’s Next Project

Our original thesis was that clients would want support for advanced builds that their teams didn’t know how to develop on their own. We were half wrong. It’s true that we help bridge the technical gap for some complex projects. That’s not where anyone wants to start.

Instead, it’s the slow, boring projects that come first. These are those pesky little projects that have always existed but the team just doesn’t have the time to reallocate employees away from their normal roles to fix it.

Solving these problems doesn’t transform a business overnight. (Read our July 22 blog post “Playing with House Money” to understand what actually moves the needle.) However, at the outset it’s the quick, easy wins that seem to win trust and lay the foundation for bigger challenges.

Boring problems typically require boring solutions. As the first quote says, there’s no point in leaping into genAI in round one. We’ve come across a few good ones:

1. Staffing: consolidate employees’ availability and preferences to save managers the hassle of building the schedule.

   Quality Control: nobody likes checking for errors, so a simple automation sweeps data updates in the ERP for errors.

   Proposals: automatic process that writes the first draft for responses to RFPs to decrease the time spent writing similar sentences.

   
   

None of these are easy enough for someone to build in an hour. None of these are entirely difficult to create within a month if it were an employee's full-time job. This dynamic makes this type of work too big to fix immediately and too small to prioritize as an attractive AI project. 

So here’s the game plan: 

• Write out a list of ten AI or automation projects you want to see completed. The first couple will likely be meaty undertakings. That’s fine. It’s the throwaway ones at the end that we want. Forcing yourself to come up with a few small tasks to reach ten total projects often unearths the sleepy projects to start with.

Let the rest of the world chase the flash AI stuff. Ignore the hype and get the ball rolling with your boring micro projects. The rest will come.