Dear Reader,

Change is hard.

Here are some of the early lessons learned from our AI transformation, integrating AI tooling into our workflows and reshaping our teams into AI-natives.

First, identify a team of people who will be your early AI natives. These are people who can ask the right questions to bots and know what AI can do. The critical skill that sets them apart is their ability to know whether to leverage AI and which AI tools to leverage, meaning they understand the nuances and limitations of the AI models and tools. These are usually the people in your organization who on their own have already delved deep into AI. We found some of the following traits in our successful AI-native teams. They are: (i) curious and ask the right questions, (ii) go getters and open to new ideas, (iii) empathetic team players, (iv) great listeners, (v) honest and own their work, and (vi) do not exhibit any toxicity.

Second, build an AI tool-testing team as fast as you can, and document everything. We selected typical Odetta projects and ran them through GPT 3.5, GPT 4.0, GPT 4.0 with browser, GPT Plug-ins, Bing, Bard, etc. We ran A/B tests to determine which tool worked best and how much quality control was required. We built an AI tool repository that our team adds to every week to overcome limitations in current models.

Third, begin by automating some internal departments, such as sales automation and marketing content generation. Sales automation can really save you a lot of time, and yet there are so many steps that outperform with human judgment. We believe the future of sales automation will be sales-savvy Odettians leveraging an end-to-end sales automation tool. Marketing is fully enhanced by generative AI tools, especially on the image and content generation side, and specifically for non-native speakers who can now generate native content. Most of the AI tools still write generic content, unless you instruct properly, so make sure your team uses pre-processing to teach the AI to write in a certain style.

Fourth, the best way to get your teams to use AI is applied learning. Have your sales and marketing lead run a workflow using GPT 4.0 with a browser, and let them compare the results with the same project done manually. Your lead will quickly see a 30-70% lift in productivity which will delight them. Then have them compare their results with a more seasoned member of your team who has already run that same project. This is how average AI Users become Super Users.

Fifth, it’s okay not to have answers right away. With so much happening and AI evolving rapidly, no one knows everything thoroughly. Let the teams know that questions don't have to be answered right away but can be food for thought and something from which to learn and grow.

Katharine

The Hardest Lesson of Them All

In the past several months, we have faced two headwinds—a tech recession and AI work substitution. The result has been a significant drop in our client hours, meaning we could no longer support all of our Odettians with enough work. We made the challenging decision to invest in the skills of the future and to prioritize our top freelancers who could get us there. We had to let go of some Odettians who had truly dedicated their hearts and souls to us, and it was heart wrenching. This was one of the most punishing and challenging moments for me as a leader. While this was required for our next phase of growth, this conviction did not make it any easier to do. 

AI Use Case: Data Labeling with ChatGPT

Our client, a bootstrapped startup in the education space, asked for our help to leverage AI to categorize data for their education resources database. Our client was hoping for more advanced “filtering” capabilities in their database, so users could filter by school grade, and to do this, they needed more structured data on grades. This effort was incredibly labor intensive but more approachable with AI. We used ChatGPT to label each high school program with specific grades using eligibility criteria directly associated with the program. We accomplished this by creating a prompt that included a list of high school programs along with their client-provided eligibility criteria and client-provided restrictions.

Quote of the Month