It’s not just big business—large enterprises with access to lots of data and compute—experimenting with artificial intelligence (AI). Small businesses and even entrepreneurs who haven’t yet launched are adopting AI tools and banking AI into their business models in new and exciting ways.

This year, Gusto released a New Business Formation report that found that more than 20% of new businesses are using generative AI (gen AI) technologies.¹ Generative AI is helping entrepreneurs extend their reach prelaunch, a phase that’s typically resource restricted, imposing time and cost limitations on what entrepreneurs can accomplish.

Entrepreneur Sean Ammirati knows first-hand about these limitations. In addition to being a founder, he’s a professor of entrepreneurship who teaches MBA courses at Carnegie Mellon University. Each year, his classes guide students through the process of developing a startup.

This year was different. His students had new technology in their toolbelt: generative AI. While some educators have balked at the idea of students seeking assistance from these apps, Ammirati is encouraging it, recognizing that generative AI has powerful applications for entrepreneurs.

“What’s interesting is the speed at which the students were able to make progress and get a number of customers using fairly robust products during a semester-long effort, relative to what I’ve experienced across 13 years of doing these classes,” said Ammirati. “They’re also doing things with their products that you just could not have done before this moment.”

Some of these startups play in the AI space, but they’re not just AI startups, they’re "AI-native" startups that used AI to get their business off the ground.

In 2023, Ammirati participated in Carnegie Mellon’s GenAI Fellows program, an initiative from the school’s Center for Intelligent Business that aims to build a body of research to enhance the understanding of generative AI business applications.

Ammirati authored a paper, “Applications of GenAI for Entrepreneurs” (link resides outside IBM.com), which details how generative tools can automate routine tasks and help entrepreneurs focus on higher-level strategic thinking. He likens the advent of gen AI to the emergence of cloud and mobile in terms of how those technologies redefined entrepreneurship.

In the paper, Ammirati argues that this technology can be more than an AI assistant, but a “co-founder,” which can provide guidance and information. Never the final decision-maker, but a helpful partner that can assist across ideation, idea validation and scaling.

Over the last decade, small business owners have been integrating low-cost AI tools to streamline workflows and perform higher-level functions that might’ve otherwise been out of reach. The revolution ignited by the release of ChatGPT has hastened this trend, giving even early-stage entrepreneurs who might not yet have a single employee the ability to take on a wider range of business functions and accelerate the pace at which they complete them.

An obvious use case for gen AI is right there in the name: content generation. This is perhaps the simplest way for entrepreneurs to extend their reach before they’ve hired employees to perform such tasks.

This value is no more apparent than in marketing campaigns. AI can use natural language processing to help draft engaging social media posts, SEO-optimized blog articles and e-commerce website copy, which help small businesses maintain a consistent and professional online presence. In email marketing campaigns, AI can craft compelling subject lines, engaging body content and effective calls-to-action to increase open rates and customer interaction.

It can assist in drafting scripts for videos, podcasts or webinars, making multimedia marketing more accessible to founders who probably don’t have such expertise. Amazon product descriptions, ad copy and other time-consuming tasks can be performed in a fraction of the time they would’ve otherwise taken.

It’s not just marketing efforts either. Business plans, investor presentations, competitive analysis, LinkedIn job descriptions, policy documents, training and onboarding materials, customer relationship management (CRM) and FAQs all benefit from gen AI solutions. AI-powered chatbots can also give a startup a customer experience edge, which can represent significant cost savings in early stages.

Streamlining the preceding tasks allows more time to focus on higher-level cognitive tasks, such as research and strategic decision-making.

Of course, gen AI can help with research, too.

One of the primary difficulties of entrepreneurship is not knowing what you don’t know. A large language model (LLM) can be an excellent resource for asking difficult, exploratory questions; questions one might even be embarrassed to ask. For example:

“Will an upcoming EU regulation impact the viability of my product offering?”

“What is the current global market volume for alternatives to my product?”

An important caveat is: entrepreneurs should remember that LLMs are limited in many ways, and hallucinations are common. What’s worse, LLMs traditionally are unable to explain how they arrived at their conclusions.

Fortunately, due to recent advancements in machine learning (ML) algorithms, LLMs that have been augmented with retrieval augmented generation (RAG) can grab information in real time from the internet, and users can check sources to verify it.

It might be helpful to think of AI-powered tools as great sources for thought starters and initial exploratory information rather than definitive answers or legally sound recommendations. Even if such tools aren’t 100% accurate, they can help founders think about a problem in a new way, or send them down a critical research path that they might not have encountered otherwise.

Founders can ask complex questions like what was mentioned earlier, and LLMs can provide context and equip founders to ask better questions of human specialists, such as lawyers, compliance experts or supply chain engineers.

“You don’t want to just ask an LLM to give you legal advice and then, without thinking about it, move forward,” said Ammirati. “But if AI could take a 10-hour project and make it a one-hour project—that’s meaningful value to the entrepreneur.”

For Ammirati, the most exciting opportunity right now is not necessarily the foundation models themselves, but applications being built on top of LLMs. Some might dismissively refer to such apps as “LLM wrappers,” seeing them as repackaging the innovation of an LLM without adding substantial new capabilities.

But apps can be specialized for specific enterprise tasks with prebuilt prompts or templates that guide the model to produce more relevant outputs for specific needs. They often involve the application of domain knowledge.

These tools can offer simplified application programming interfaces (APIs) and intuitive interfaces. This makes it simpler for developers to incorporate LLM functions into existing applications without having to manage complex setups, such as handling API calls, managing tokens or configuring settings.

Look no further than venture firm Sequoia Capital’s report, published in October 2024, for confirmation of the value of LLM-based apps:

“Two years ago, many application layer companies were derided as ‘just a wrapper on top of GPT-3.’ Today, those wrappers turn out to be one of the only sound methods to build enduring value. What began as “wrappers” have evolved into ‘cognitive architectures.’”²

Ammirati co-founded an early-stage startup that aims to be one such tool for entrepreneurs and other innovators.

“I was doing a lot of work with corporates as part of the Corporate Startup Lab that I used to run at CMU, and part of what I realized is a lot of these R&D groups are trying to also translate their inventions into commercially valuable things.”

Growth Signals (link resides outside IBM.com) is a tool for executives and researchers that want to determine how to best apply research and development (R&D) resources. It uses AI to analyze the competitive landscape, writes technology summaries, guides brainstorming sessions, and can even use agents to crawl breaking news and newly published research.

“It helps you translate market and technology signals into concepts that are worth exploring, and helps you manage and refine those concepts and perform early validation.”

Innovation isn’t just about coming up with new ideas, it’s about coming up with them first. And if a tool can help innovators get there faster, perhaps resources invested in a so-called “LLM wrapper” might be money well spent.

Ammirati cited two other startups: Cove and Glean (links reside outside IBM.com) playing in this sandbox. Both aim to take the user experience beyond the chatbot we’re accustomed to when we interface with LLMs. Instead of bots, they use AI to offer a multidimensional visual workspace that’s tailor-made for common enterprise functions.

It’s an exciting time to be running a small business. As entrepreneurs and innovators seek the best AI tools to optimize workflows, automate repetitive tasks, assist with research and handle project management and other business operations, we can expect an influx of new “picks and shovels” products. These tools will give entrepreneurs the momentum to reach the market sooner and achieve profitability faster.