Upcoming Events

SF & Bay Area

NYC

Boston

Miami

This week, we are thrilled to share an insight piece from our fellow community member Alex Young. Alex is leading the generative feature development at Squint AI and dives into how GenAI is revolutionizing industries by either expanding possibilities or reducing inefficiencies. In this article, he explores how these two distinct approaches are enhancing productivity and unlocking enterprise value across various sectors.

Generative AI Strategies: Expanding Possibilities vs Reducing Inefficiencies

Two approaches to Generative AI

While there are many lenses through which to view the sweeping changes brought about by Generative AI since its public debut in 2022, we believe GenAI has enhanced human productivity and demonstrated potentially huge enterprise value in two basic (and nearly opposite!) ways. The first could broadly be described as expanding possibilities—this is the use case most familiar to many of us early adopters where generative tools drastically reduce the learning curve for picking up a new skill, hobby, or even professional expertise. We create new stories, marketing copy, impossible images, and functional code with these tools. We use it to expand our palate of choices and take on challenges we would never have approached before, making near-professional level performance accessible to all. The second way GenAI empowers us is by reducing inefficiencies that occur when tackling existing problems. Examples include the tools built to summarize, identify relevant information, transform data, automate repetitive tasks and eliminate manual effort. The initial wave of AI adoption has been mostly characterized by the former approach. As a recent Writer report shows, content generation in various forms has been the most common use case across most American workplaces. However, the latter has the potential to create incredibly sticky products, and we believe that in the enterprise context, a significant portion of the overall value will be captured here. By tackling difficult problems and using AI to narrow the associated search and action spaces, we expect to see huge improvements in employee efficiency across a diverse set of industries.

Understanding the approaches

It's easiest to grasp this distinction with a few examples. Here are some well-known products best described as using AI to expand possibilities:

  • ChatGPT: As the first generative AI tool on most people's radar, ChatGPT is the most emblematic of this approach. People who haven’t written an essay since high school are now using it to publish blog posts. Tinkerers who never learned to code are creating websites. ChatGPT is so open-ended because it wasn't built to solve a specific problem for a specific set of customers—like the PC, it functions as a destination-less bicycle for the mind, and its proficiency across reasoning tasks is growing exponentially with every new iteration of the core foundation model.

  • CopyAI: Much of marketing is about new ideas and exploring options. CopyAI was an early winner in the latest AI wave because it recognized the value of rapidly creating a large number of imperfect choices and letting humans pick their favorites. Now that anyone can generate content rapidly and professionally, the challenge for creatives lies in curating and refining this content to ensure it aligns with brand voice, engages the target audience, and maintains originality–a higher order problem. 

  • Devin.ai: The latest VC darling realized that treating autonomous agents as black boxes made it hard to imagine trusting them as real coworkers. By creating a programming agent that shows its work, can be instructed to modify outputs, and requests user feedback when necessary, Devin.ai has created a tool that may make entirely new engineering approaches feasible for technology companies. Massive leverage on technical expertise will be unlocked as individual contributors shift from writing code to task delegation for their army of autonomous agents. 

On the flip side, some GenAI Products are best described as reducing inefficiencies in existing problems: 

  • Github Copilot: Programmers extol the virtues of flow state and deep work with almost religious fervor. This coding tool allows them to shift focus from many small choices to fewer, more impactful decisions without disrupting their focus or interrupting their process. As a result, Copilot has become an essential component of many engineering setups and many companies require new employees to already be proficient in these tools. The genius of Copilot was to enhance a familiar IDE pattern, autocomplete, instead of trying to invent a disruptive new user interaction. Tasks like writing boilerplate code, crafting simple logic, and developing unit tests have prevented engineers from focusing on what really matters for decades. Github Copilot provides a powerful solution that is already unlocking outsized developer productivity and happiness! 

  • Squint: The manufacturing industry is often one of the slowest to adopt technology and least digitized, and many of its core challenges around training, safety, record keeping, and day-to-day operations have plagued incumbents for decades. Squint uses computer vision, spatial awareness, and LLMs to digitize work instructions, get trainees up to speed faster, and provide operators with critical information in a timely manner. That said, the key unlock lies in listening to customers who often aren’t accustomed to using generative tools, so we can build a deep feature set that embeds seamlessly into their workflows. In many ways, the goal isn’t that different from Github Copilot, just applied to a different industry–how do we leverage LLMs to increase workplace productivity and satisfaction in manufacturing facilities?

  • Harvey AI: While details about this Series B startup have remained tantalizingly out of reach, the little that we do know indicates a clear "narrowing search spaces" approach. Lawyers spend hours pouring over documents, identifying patterns, and matching existing information to new corporate problems. Harvey is likely helping reduce a massive pile of information into smaller, more digestible pieces, allowing lawyers to focus on the critical thinking required to win cases and tame boardrooms. Customers have noticed the profound impact they’re making on the legal space, and their recent $100M Series C that has catapulted them to unicorn status suggests investors do as well.

How to choose what's right for your product

The decision about which approach is a better fit for your company and industry should emerge from a clear understanding of your customer's problems. At Squint, our core insight is that in the real world, access to the right information at the right time will always be valuable. In occupations that don't take place behind a desk, there is often a shortage of readily accessible information and limited time to find it. Questions like "What do I do now?", "Who do I ask?" and "Where do I go?" are a pervasive part of the operator experience in the manufacturing industry we serve. At Squint, we define this approach as Manufacturing Intelligence. So for us, it was natural to take the reduce inefficiencies approach and attempt to answer these questions for our users with AI. We've built our own Copilot, which answers complicated questions about tools and procedures by searching through hundreds of long documents to glean insights. This can eliminate hours of operator effort. We also use LLMs to help operators digitize old work instructions by transcribing videos and generating procedures in our app, saving them the effort of manually re-entering existing information. Explicitly focusing our efforts on this second GenAI approach has helped our customers develop expectations for new AI features as we continue to launch them.

Ultimately, both approaches to building GenAI products are likely to create generational companies. The mistake is to start with a preference—listen to your customers' problems and see which direction the solutions pull you!

Events Spotlight

The GenAI Collective Garden Gathering 🌺

A huge thank you to everyone who joined us at Stanford for an evening of discussing multimodal AI! Discussion topics included generative art, deep-fakes, leading open-source models, and policy and labor implications of multimodal AI. Shoutout to Sunny Scott for being an incredible host, and to Billy Asel for creating the packed agenda.

The GenAI Collective Office Hours 👨‍🏫

Massive shoutout to our amazing startup pitch teams and esteemed investor partners for making our inaugural GenAI Collective Office Hours so incredible! From go-to-market strategy to sales, business development, and product market fit, our discussions touched all parts of the startup journey and proved invaluable for our participants. This event series is just beginning, so don’t hesitate to reach out if you have what it takes to be in the next cohort! Are you ready to enter the arena?

Value Board

Join the GenAI Collective team! 👷

This is a true labor of love and we are always looking to involve new leaders who share our passion for community building!

Does this sound like you? Learn more and reach out to us here!

About Alex Young

Alex is a Full Stack Engineer at Squint, where he has led the development of several Generative AI features. Based in NYC, he is always eager to discuss the latest in tech and startups. DMs open on X @alex_j_young! 🧑‍💻

Keep Reading