By PLUS Labs
In this edition of “AI Innovators,” we are featuring Manager Jamie Forte and his strategies for deploying AI in his client work. Jamie works on several accounts in the tech, finance, real estate and media industries. He uses AI daily in his workflows to improve the efficiency and effectiveness of our team.
“At PLUS, we are committed to harnessing the power of AI at every level of the business, from our partners down to our junior members,” said Senior Managing Director Josh Silberberg. “While Jamie is just starting out his career, he is one of the true innovators at our firm. His use of AI is a model for our entire our team, and we’re grateful for his energy and willingness to develop innovative solutions that make us faster, smarter and better.”
“I am deeply impressed with the ways Jamie is using AI,” said Senior Managing Director Tim Donahue. “He understands that the best use of AI is driven by professionals who use it to sharpen and broaden their thinking. If you are lazy with AI and expect it to do your work, you will get subpar results that revert to a mediocre mean. That’s not our culture. Jamie’s use of AI embodies a mindset that is helping PLUS deliver cutting edge results for our clients.”
Read this Q&A with Jamie Forte.
AI allows us to expand our capabilities beyond what was previously possible. For example, I’ve used AI to analyze and categorize data sets which would have taken dozens of hours by hand, without any loss in accuracy, allowing us to quickly deliver time-sensitive research projects on the political and competitive landscapes for our clients. In another instance, I used AI to analyze 12,000 comment letters on a proposed regulation for sentiment and sender type. These uses are invaluable time savers. But what I’m most excited about is how AI allows us to build our own custom tools from scratch. There are numerous areas where a simple tool can make a huge difference, and we already know what that tool should look like. AI allows us to build those tools without knowing how to code – creating opportunities to save time and do more that are actionable.
In Public Affairs, we often need to transcribe online streams – press conferences, Congressional hearings, etc. – and move them quickly to our clients. I wasn’t satisfied with the existing transcription offerings for our specific use-cases, so I built my own – StreamScribe – which can transcribe any online stream and produce a client-ready transcript in seconds. I’ve also had success using AI to build time-saving browser extensions for routine tasks such as clips. Even something as simple as a preformatted article copier can save a lot of time in the aggregate.
The ability to create anything we can envision is inspiring. As the AI tools become increasingly intuitive, more people will be able to use them – even without technical know-how. Eventually, I anticipate that AI will get to a place where it can quite literally anticipate what want from it. In the future, anyone with a good idea will be able to do something with that idea – even if they lack the specific tech skills to implement it on their own.
But I am confident that AI will never fully replace professional involvement and human intuition. The best use of AI is driven by thoughtful people. If you are strategic with AI, you can develop niche tools that will greatly enhance your productivity and that are customized to your work responsibilities. With the existing technology, we can build exactly what tools we need ourselves. The true value of AI extends beyond doing things faster or more cheaply. It’s about creating new possibilities where they wouldn’t have existed before.
AI cannot do your thinking for you. While LLMs are getting more accurate and sophisticated by the day, it’s critical to bear in mind that these are still predictive models. They do not understand things the way that humans do. If we end up relying on AI too heavily, we are at risk of having our critical thinking skills atrophy from disuse, especially for younger professionals who grow up using AI regularly. It’s important to bear this distinction in mind, lest we grow accustomed to outsourcing too much of our decision making to AI when it is not up to the task. Be the captain of your favorite LLM or open-source model, not a passenger.
I would like to do more work on improving content generation models based on past examples. In my personal view, AI’s writing capabilities still lag its other uses – such as coding and video/image generation – and I believe there is a lot of room for improvement here. With more robust retrieval augmented generation, I’m hoping to develop models that are better at drafting the kinds of content we regularly need – especially at a basic level where quantity and turnaround time are paramount.
I check in periodically with open-source AI development communities. Frontier labs will make announcements when they release a new model or have a breakthrough, but dozens of small improvements happen all the time in community development forums – especially around local AI models. These are exciting because you can download and run the new models yourself, and as local AI improves, it becomes increasingly viable compared to existing large models that are closed source and only available online. This is also a great way to monitor the status of new research developments, especially whether they are actionable in the short term. Someone in the community will usually put that new finding to work not long after it is published.
The better you can direct AI, the better results you’ll get from it. Rather than just asking AI questions and letting it decide for you, it’s best to come prepared with a clear sense of what you want – and how the AI should get there. When I develop programs, I tell the AI exactly how I want it to go about completing my request, which frameworks to use, how it should approach the design, what the primary considerations and constraints are, etc. If you don’t give AI a clear enough picture of what you want, you may end up wasting time going in circles. Likewise, it’s helpful to be able to recognize when the AI is stuck because it is missing information. Many models are trained to solve problems on their own, but this can be a waste of time if you already have the answer. If you think about how the AI will try to answer your request, you can often identify key details that are helpful to include in the prompt up front, so that it doesn’t waste time going down unnecessary avenues of inquiry.
"*" indicates required fields
