Application of AI at Startups: What Works Today (and What Still Needs Humans)

As technology continues to evolve, startups are increasingly turning to artificial intelligence (AI) to give them a competitive edge. AI is transforming how we operate—helping streamline processes, speed up product development, and discover valuable insights that might otherwise slip through the cracks. 

AI is quickly becoming a key player in the startup toolkit. In this post, we’ll take a closer look at how startups are harnessing AI in various areas, the challenges they face, and the strategic advantages it provides to early-stage companies eager to grow quickly and efficiently. 

In this interview with Salil (Sal) Darji, a mentor at C10 Labs and organizer of AI Tinkerers Boston, he shares some insights into the AI space, highlighting some great examples of AI use and setting expectations regarding human interaction.

Startup Boston (SB): Where do you see the AI space today? 

Sal Darji (SD): Right now, there is a lot of hype surrounding everything in the AI space, to the point where some aspects are overhyped. I believe startups need to provide products that are valuable to people, that show innovation, and that drive change. AI should be a part of it, but in some cases, it may have swung too far. Yes, you should adopt AI, but in the same way everyone should use compute as part of their general business model. AI is computing; AI is the future of computing. 

Startups should consider how they use this compute and its impact. What matters most is the business model and whether it works for the startup, allowing it to grow and scale. 

SB: Are there areas where we need to temper our expectations with AI currently?

SD: AI is extremely powerful and helpful, but there are areas where we need to temper our expectations of its outputs and outcomes. One of these key areas I have seen is in AI’s ability to generate literature or Intellectual Property (IP). No matter how you look at the output of AI writing, it doesn't feel creative; it reads robotic. 

There are certainly areas where this robotic tone and output do not matter, for example, booking a hotel reservation or anything transactional. When booking a hotel, the end user expects straightforward answers, so the AI does not need to be creative in its responses - only accurate and formulaic. Creativity typically doesn't matter in these responses. Compared to publishing a book, some are okay with a formulaic voice, but when I read, I prefer it to sound and be more human.

SB: Which areas have seen the best ROI with AI?

SD: I think photos, specifically those generated by generative AI, have shown and continue to show a lot of promise. At the same time, it may be too far. There are certain times when you want a human-centric photo or digital presence, one that appears more natural, and in these cases, AI is a bit too much.

In project management, there is an iron triangle: Quality, Cost, Quantity. AI has a lot of quantity at relatively low cost, but quality is questionable. In the future, there will be higher quality, but there is also a way to provide creators with these tools to help them stay in the loop. Humans should still generate some things.

SB: Which departments should be using AI today, which should wait a little longer? Is there an appropriate timeline you see utilizing AI?

SD: Every department should be exploring how AI can support their work, but adoption should be phased. Sales and marketing should lead the way, as AI tools can boost efficiency and personalization, though care must be taken to avoid spammy practices. Tools like Salesforce make early adoption easier. Functions like HR and internal communications should move more cautiously, especially when dealing with sensitive topics like layoffs, where the human element remains essential.

SB: Which tasks should lean or leverage AI today, and which should fall for humans -- Why? Are there outliers to these tasks? 

SD: Tasks that should leverage AI today are those that add value or delight, but are often overlooked due to time constraints, such as personalized check-ins or helpful updates from a hotel to a guest. These interactions enhance the experience but aren’t always feasible for busy staff. On the other hand, tasks that require deep empathy, complex judgment, or trust, such as conflict resolution or sensitive conversations, should remain human-led. Outliers exist when AI can genuinely enhance human connection without replacing it.

SB: Which AI specific tools or products do you recommend for building an AI-Stack?

SD: For building an AI stack, I recommend exploring tools like Google Gemini and AI21 Labs. AI21’s JAMBA model is especially strong, offering large context windows ideal for RAG workflows or deep document analysis. Their Maestro platform also stands out for AI planning and orchestration, providing structure beyond just reasoning. It’s also exciting to see smaller players like AI21 innovating quickly. It is definitely worth checking out.

SB: What resources do you leverage to stay up to date with the break-neck pace of the AI world?

SD: One of the best resources I use to stay current in AI is leading the Boston chapter of AI Tinkerers, a global meetup community for hands-on AI enthusiasts. We meet monthly to share, build, and learn from each other. If you're interested in practical AI and staying on the cutting edge, this group is for you.

SB: What is your best AI-pun?

SD: I can’t take credit for it, but one of my favorite AI puns is: “Don’t worry, the self-driving car is still learning—it’s just going through some driver updates.” And if the AI jokes fall flat? Well, they’re probably just a bit too robotic!

To stay in the loop on everything startups and the Greater Boston startup ecosystem, don’t forget to subscribe to the Startup Boston newsletter! We’re excited to send (helpful) weekly tips and news updates directly to your inbox. 

About the author: Joe D'Agostino is a startup founder, developer, advisor, and active security practitioner. Today, he manages the product and application security team at an AI startup in Boston, MA. When he is not securing or building products, he enjoys spending time with his wife, their newborn daughter, and their dog, Josie.