Is Your AI Idea Any Good? A 5‑question Stress Test From a Serial Founder

In the gold rush of AI startups, how do you separate the valuable ideas from the interesting ones? In her latest piece for Every, Stella Garber offers a five-part framework for evaluating business opportunities before you expend time and capital. Drawing from personal experience across five startups and 32 angel investments, she provides a practical roadmap for founders navigating the AI landscape. I love the mix of strategic wisdom (focus on execution over ideas) and tactical guidance (invite users into a Slack channel to share feedback)—read on to learn more.—Kate Lee

Was this newsletter forwarded to you? Sign up to get it in your inbox.

With startups, it’s almost never about the idea—it’s about the execution.

As a founder, I would know: I’ve had bad ideas that turned out well, and some ideas I thought were great that turned out to be…not so much. Now that it’s easier than ever to make your product idea a reality with AI, how do you tell if your idea is fatally flawed before you waste time, energy, and seed money?

Before my startup Hoop had its current product direction—an AI executive assistant for top busy founders, leaders, and execs—we were looking to solve decision-making for teams. We began making an app where teams could lay out decisions asynchronously and create team-wide decision logs. There weren’t many participants in the market, so we figured it was a big opportunity. This turned out to be a bad omen: Decision-making is convoluted, people don’t want to add another collaboration tool to their workflows, and AI doesn’t solve the complexity of human communication. So we pivoted into an area where my team had a lot of experience: work management—and conversion to paid users went from zero to 15 percent.

Now we’re building an AI-native solution to an age-old problem: How do I make sure I don’t drop the ball on everything I have to do? AI gives us an opportunity to make something that was prohibitively expensive, an executive assistant, more widely available to more people.

With every hype cycle, there are a lot of founders who want to get in on the action. And the temptation to seize on an idea and charge ahead as fast as possible is real. But if you want to build something lasting, you need to make sure you’re not falling into the trap of doing what’s popular versus doing what’s right for you (and being the right founder to solve the problem).

I’m currently on my fifth startup, with a range of outcomes from not-so-great to life-changing success. As an angel investor, I’ve written 32 checks into early-stage companies. Over time, I've developed a set of critical questions that help me distinguish between AI ideas that are merely interesting and those that are genuinely valuable. Here is my five-part framework for thinking through whether there's a market for your startup, and how to think critically before diving into a specific idea.

1. What existing workflow are you fundamentally changing?

AI isn't magic. It's a set of tools that needs to dramatically improve an existing process. The most compelling AI startups don't just add a feature: They reimagine entire workflows, remove the need for them, or figure out ways to automate them.

Innovation makes people’s lives easier by helping them accomplish more in less time, or freeing up their time for creative or leisurely pursuits.

Take task management. Most of my teammates at Hoop worked together previously at Trello, where we worked on the product, design, engineering,and marketing teams. We brought this experience into the next generation of task management by thinking through how AI can help people in ways software previously couldn’t.

The breakthrough wasn't just, “AI can prioritize my existing tasks.” It was, “I don’t need to create tasks or remember to do things because AI should know what I’ve committed to by being omnipresent in my workflow.” We didn’t want to make task management incrementally better. We wanted to eliminate the need to write down and organize tasks entirely.

Startups have an edge on incumbents in AI because many incumbents tack on AI features to existing legacy codebases. The real innovation is thinking through first principles on how to fundamentally solve a problem.

2. Where is the customer's deepest frustration?

Great AI products emerge from acute pain points. Not theoretical problems, but specific moments of genuine customer frustration.

The best founders can articulate stories that intuit existing problems in their customers’ lives. The stories come from a deep understanding of the problem their product is solving. Tell me about the specific moment a professional wastes time, feels overwhelmed, or gets stuck. Tell me about dollars wasted or resources burned at an enterprise level. That's where transformative AI solutions live.

In my world of work management, busy professionals don't just want their work summarized. They want to eliminate the soul-crushing hours of reading through irrelevant information—searching for that nugget they swear they saw somewhere but can’t remember where—or follow through on commitments they’ve made. That’s the problem Hoop is solving. When users see a task in Hoop they would’ve forgotten or missed otherwise, they understand the promise.

Another example: I’m an angel investor in a company called Wander, which provides a tech-enabled, quality controlled-alternative to Airbnb. All Wander properties meet a certain bar of excellence and are designed for people who travel for work and fun. The founder’s insight came from having inconsistent experiences with Airbnb and a desire for a better solution for the growing market of digital nomads. We both believed that the future of work would be more flexible, and a certain class of traveler would be willing to pay a premium for an experience that enabled them to mix work and personal travel without sacrificing quality. Now, with AI, they’re further able to customize experiences by providing faster and more personalized concierge services to customers.

In that example, AI is helping a company operate more efficiently and provide services they weren’t able to previously as a result. Sometimes that’s enough of an edge if the solution is truly innovative in a market that’s not serving a set of customers’ needs.

3. Can your AI unlock a dependency or superpower someone’s capabilities?

Here's a litmus test: If a human with enough time and skill could accomplish a task, how much time is your AI tool saving? What can that person do better or faster as a result of your tool?

The most exciting AI startups unlock capabilities previously improbable for someone to accomplish themselves. You used to rely on an engineer to take an idea from Figma to a coded prototype. Now you can prompt a prototype to life in Lovable. All of a sudden, designers, product managers, and tinkerers are freed from waiting on, paying for, and finding someone to code their project. Here are some other potential unlocks:

1. Does your product process incomprehensible volumes of information? Legal AI startups like Harvey combine both legal expertise and the ability to analyze large amounts of data to accomplish something that used to take a team of specialized humans.
2. Does your product recognize patterns invisible to human perception? Customer insight startups like Kraftful ingest a ton of customer data and recognize patterns that help inform product roadmaps.
3. Does your product get someone to 80 percent done on something that used to take time, craft, and specialized skill? Creating pitchdecks like Gamma help knowledge workers solve the blank page problem and get most of the way there with a task that once took hours.

4. What's your unique data advantage?

In AI, data isn't just fuel—it's the entire engine. The most promising startups are transforming data in interesting and useful ways that were previously either hard or impossible to do. Plus, ingesting more data entails higher switching costs for customers: The more data and insights a tool has into your workflow, the more time consuming it will be to replace it, thereby making you more likely to stick with what you have.

Startups that can train on specialized data sets are a big opportunity—especially in the enterprise. The more data your business generates, the more opportunity there is for analysis—and resulting insights. What do customers want? What do employees want? Until now, we’ve been resource-constrained by the capacity of human analysts. Now, we’re all human analysts, and with AI, we can extract more insights from existing business data more quickly.

Companies are already porting their data into secure tools like Glean in order to make search across documents and tools easier at an enterprise level, reducing multiple tool-specific queries into a single-answer search. Extrapolate into data-heavy industries like finance and healthcare, and it’s mind-boggling to consider all the ways AI can be applied to make existing workflows better, faster, and easier using existing data.

Another early example is the explosion of sales enablement tools with new capabilities in the go-to-market space. The sales and marketing teams and tech stacks of even three years ago have been completely transformed. This transformation is about to happen in every data-heavy industry.

5. How will you solve the trust problem?

AI brings unprecedented capabilities and healthy skepticism, so your startup must explicitly solve for trust. You can’t build the perfect product right away, but the initial product has to be robust enough so that customers can see where you’re going and trust that you will get there… and feel like it’s worth being along for the journey. Companies must operate with the following principles in mind:

1. Radical transparency about AI capabilities and limitations: At Hoop, we take building in public seriously as a way to build trust. Besides publishing regular product updates, we invite customers into Slack channels so they can directly share feedback with the team.
2. Clear human oversight mechanisms: LLMs hallucinate. So humans need ways to safeguard against any weird data being presented on their behalf.
3. Predictable, consistent performance: As AI gets more baked into workflows, startups need to build their evaluation muscle to make sure results meet a quality bar.
4. Ethical considerations baked into the product architecture: Users of AI do not want models to train on their data without their consent. AI companies have to be upfront about what is done with their data and how it is protected.

Companies have been burned with sketchy practices like training on customer data without explicitly saying so. The new AI consumer is wary, and many companies already have AI policies that limit which new tools employees can adopt. At the same time, companies are making a huge push on individuals and teams to incorporate AI into their workflows. The more rapidly you can be AI-native, the more you can safeguard your career for the future.

The intangible factor: Founder conviction

Beyond these tactical assessments, there's something harder to measure: founder conviction.

The best AI founders don't just understand the technology. They've lived the problem. They can describe, with visceral detail, why the current approach is broken and how AI represents a paradigm shift. They are unrelenting in their optimism and enthusiasm for solving a customer need. This is essential for investment in any company, in any industry, at any moment in history: You always bet on the jockey, not the horse. Especially at a time when change is happening at an unprecedented rate, you want founders who are nimble, focused, but also flexible.

Mariam Hakobyan, the founder of web development tool Softr, is one of them. Her vision was to make it easy for businesses to build custom software, and I invested in her Series A three years ago. Initially, she did this through developing a no-code app builder. As AI capabilities for prompting in natural language became a reality, she made it even easier for Softr’s customers to build custom apps without writing code by using AI.

A final word of caution

Not every problem needs an AI solution. The most mature founders I work with are as comfortable walking away from an idea as they are pursuing it.

AI is a powerful lens, not a cure-all. Your job isn't to apply AI everywhere, but only where it creates transformative value.

Trust the problem more than the solution. Listen to customers more than your own excitement. And always, always prioritize human impact over technological complexity.

Stella Garber is the co-founder and CEO of Hoop, an AI assistant for busy professionals. Previously, she led marketing at Trello, building the team as its initial marketing hire to Atlassian's acquisition and beyond.

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We also build AI tools for readers like you. Automate repeat writing with Spiral. Organize files automatically with Sparkle. Write something great with Lex. Deliver yourself from email with Cora.

Get paid for sharing Every with your friends. Join our referral program.