“In your interview, ask the engineer how they use GPT in their work. If they don’t, you probably shouldn’t hire them.”
A founder I recently spoke with shared this hot take on hiring. I’d argue that this applies to nearly every role within a startup today.
As an early-stage investor, one of the traits I search for in founders is velocity of learning: Finding tools and processes that can help small teams experiment, learn and iterate rapidly in their search for product-market fit and getting to liftoff. Figuring out what works — and what doesn’t — in when it comes to building and scaling your efforts.
What ChatGPT and other “generative” AI tools today enable is a turbo boost to that learning. Two important tenets to keep in mind: 1) The output is only as good as what you put in, in terms of scaffolding (prompt engineering, as well as targeted and thoughtful follow-up prompting); and 2) you’ve got to check its work and while you can expect it to get you 80% of the way there, it won’t actually do the work for you.
In this blog post, we’ll highlight eight ways every startup should consider using AI, based on our observation with founders and also within our own work.
1. Supercharging education content development
As education investors, we’re particularly interested in how generative AI can reduce costs and accelerate the development of course content and curriculum. We’ve backed companies like Koalluh that lets kids shape the stories they read, and Ello which is reimagining the instructional experience for early readers.
AI is helping to accelerate content development across our portfolio. One tutoring company is using it to generate personalized content and questions based on student and teacher inputs. Another in the workforce training space uses it to create questions and explanations from video transcripts or slides. (“It’s pretty good, with a ~90% accuracy rate!”) Another uses it to produce draft copy for the overview descriptions for each course session. They also feed transcripts of live courses to create key takeaways for students and to feed their content marketing engine.
2. From streamlining to reimagining customer support
Staying close to customers is a critical mandate for every early-stage company. But that personal touch is sometimes lost as one rapidly scales. AI has been deployed for customer support for some time to help field common queries and provide step-by-step guidance on technical issues (akin to a FAQ) to free up time for teams to focus on the most critical issues.
ChatGPT’s AI capabilities can assist with a wider range of task and go further to offer a more personalized customer experience. Colleges and universities have been working with Mainstay for years now to do just that. Mainstay creates conversational AI that supports students to and through college – from FAFSA applications to course registrations – while delighting them along the way.
Today’s AI tools can offer real-time, empathetic support tailored to each customer’s unique use case (e.g. when implementing a new feature) and anticipate and recommend services along their journey. It’s a powerful way to augment your customer support team, assist with a wide range of tasks and reduce response times.
3. ChatGPT as your interim Head of Operations
One charge for early-stage founders is to serve as the head of whatever is most important to the business, learn as quickly as possible, and then hire an expert to fill the role. Most seed stage companies don’t have finance, accounting or HR leaders in-house who can address questions like: What do we need to keep in mind for QSBS eligibility? What’s important to consider in making my first VP of Sales hire?
While blog posts and opinions abound, GenAI can serve as a TL;DR for founders on how to address some of these common early-stage challenges. Maria Barrera, CEO and founder of Clayful, uses ChatGPT for what she describes as “market research but more… internal. We don’t have full-time HR or Finance support so we leverage ChatGPT to learn and verify how to do things instead of reading endless articles (ie what is QSBS and how does it work) and/or for templates [for things like HR].”
4. Surfacing insights from unstructured data
Messy, unstructured data is the nemesis of every company trying to build fast and do things that don’t always scale on the road to product-market fit. ChatGPT is remarkably effective at taking unstructured data (such as open-ended survey responses from user research, a PDF of events, or conference attendee lookbooks) and translating it to structured data and insights.
One K-12 portfolio founder shared: “Every time we go to a conference we get a lead form in a different structure. Even the school/district names are different! We’re using ChatGPT to restructure and normalize data to make it easier for us to upload it into our CRM — it’s much faster than manipulating the spreadsheet.”
We’ve found the key to better outputs is to be very specific about what format you want and provide examples of how you want it to interpret the data. It’s also helpful to break more complex data asks into a series of steps.
5. Turning talk and notes into action
Moving at high velocity means decisions happen every day. There’s plenty of talking and typing in our line of work, and then we spend more time tidying up our notes to share. ChatGPT and other AI-powered writing tools today can be incredibly effective at turning digital chicken scratch into coherent memos. This not only saves your time; your colleagues will also appreciate the succinct digest of whatever you’re trying to convey (especially if writing is not your forte).
We’ve found a couple things that can help: When prompting, you can define the output format (memos, email) and any required action items (deadlines, responsible parties) to help AI understand your intentions and desired output. You can also organize your notes into sections before inputting them into the AI, which can help create a more coherent and well-structured summary. For instance, you can divide your notes into categories such as “Key Points,” “Action Items,” “Decisions Made,” and “Next Steps.” Just a little structure can go a long way.
6. Getting over coder’s block
Just as it has done for writing, generative AI coding can be a way to get over “coder’s block.” It can help developers explore different alternatives for the architecture of a new feature and help supercharge the adoption of a new library. AI code explainers can make it easier for programmers to document their work and also understand what their predecessors have built. Debugging assistants can help identify and fix code errors more quickly.
Tools like Replit Ghostwriter are helping teams bring their ideas to life much quicker, without getting bogged down in syntax, formatting and other nitty-gritty issues that can frustrate even experienced programmers. In fact, GitHub recently conducted some research that found developers who use GitHub Copilot were both able to complete their work 55% faster and found more satisfaction in their work. GenAI is rapidly narrowing the gap between intent and action – and helping developers spend more time on complex, critical thinking and problem solving – and less time on repetitive tasks.
7. Breaking language barriers
Today’s distributed workforce means that teams are spread out across different time zones and languages. ChatGPT and others can help bridge the language divide by translating emails, product descriptions, internal documentation, all to enable more rapid learning and collaboration.
We’ve found tools like Deepl and WordTune to be game changers for breaking through writer’s block and adapting to cultural norms and contexts. (Generative AI tools can more accurately translate the intended meaning of idioms, for instance.) These smooth the way for evaluating pitch decks in Portuguese or Mandarin and even closing a deal in a new geography. In many cases, you’ll still want a professional translator to cross check, but ChatGPT can get you 80% of the way and speed the exchange.
8. As your in-house marketing copywriter
With the plethora of writing assistants today (Jasper, WordTune, Copy.AI), generative AI tools can help accelerate the marketing content development process – everything from sales and marketing collateral to pitch decks. Leverage these tools for idea generation, design iteration, and wordsmithing content for A/B tests. We used ChatGPT extensively to brainstorm transitions, examples, and even draft a few spicy tweets to amplify this post on social channels.
Much has been written about how generative AI can assist with marketing collateral. Maven co-founder and CEO Gagan Biyani recently shared how his colleague has perfected the art of ChatGPT-created blog posts that “looks as good or better than most SEO articles and takes 1/3 the time.”
AI is quickly becoming an integral part of our daily workflow, and every day we see more creative use cases. Technical or not, every founder is discovering new ways to incorporate it to support their work. Winnie CEO Sara Mauskopf recently shared how she upgraded her team to the paid version of ChatGPT to see how it can be integrated across a variety of tasks, from content creation and customer outreach to coding assistance and translations.
How to think about AI if you’re not an AI company? Here's my framework. It's still evolving and I'd love your thoughts. ⏬
🏛 Pillar 1: How can AI make your internal operations more efficient? The key here is to start small. Think about incorporating AI into your existing…
— Sara Mauskopf (@sm) May 2, 2023