How are we using AI in our business?

These are the biggest places we currently use AI:

1. Coding, particularly writing the code that generates visualizations. 

   • Data viz packages, groups of functions within a coding language, each have their own syntax. AI allows us to code much faster and take the best from every visualization tool, rather than becoming experts in one specific area.

2. Code review

   • We’re continuing to build software solutions, both for us internally and for our clients, and reviewing code is a great use of AI. 

3. Keeping up to date on the latest trends in data science/ML etc

   • We have a daily task that runs in Claude Co-work and assembles AI papers and industry updates every morning.

4. Note taking

   • We’ve started using Plaud as our notetaker. It helps us stay in the moment when talking with new people. That being said, we always bring pen and paper too. Taking handwritten notes has been shown to increase retention and understanding, and we’d never ignore good research!

5. Proposal writing

   • We take notes from meetings, as well as a detailed Q&A, to pull together all the information we need for a proposal. This helps us make sure we cover everything discussed in an initial meeting. It also helps with formatting. Good riddance to dragging a text box around a PDF for four hours. 

6. Data cleaning

   • We never expose sensitive client data to an LLM. Instead, we use machine learning tools to clean up data so that we can use it more easily. One example of data cleaning that can be really tricky is ID matching. Imagine you’ve been running a construction company for 20 years. You have projects in a mix of Excel sheets, PDFs, project management software and Quickbooks. However, none of those projects have a consistent ID across each platform. That makes it much harder to know exactly how much money a project made or how far it went out of scope compared with the original proposal. Natural language processing and machine learning models can help match content across each platform and create a consistent identifier. 

Where we don’t use AI:

1. Generating social media content

   • We tried it, we ended up spending more time reviewing what was written than it would have taken to write it ourselves.

2. Client communication

   • This isn’t 100% true. Sometimes we use it to check grammar and spelling. On the whole, it’s faster to get good at writing emails and communicating with clients than to default to an LLM.

3. Logo + graphic design

   • We don’t really have a logo yet, so this isn’t entirely fair. However, we decided long ago that we wanted to work with real artists for any graphic design in the business.