We live in a time where the capabilities and accessibility of artificial intelligence (AI) and machine learning (ML) technologies are increasing exponentially. It's pretty exciting. (For simplicity's sake, we'll use the term AI to refer to this entire spectrum of technologies). While there is a lot of hype on this topic, it is hard to dispute that AI will fundamentally transform our day-to-day lives. Startups often lead the way by introducing and innovating with emerging technologies, so it's no surprise their use of AI is steadily becoming the norm. However, as with any game-changing technology, introducing AI into the mix without understanding it can cause confusion, misuse, and unrealistic expectations.
Hollywood's sensationalized portrayal of AI as omnipotent killer robots is a far cry from today's reality (if you are interested in the ways AI can go awry and how we can prevent this, search for AI safety). The long-term goal of AI is to create machines that exhibit intelligence comparable, or even superior to, human beings, commonly referred to as artificial general intelligence (AGI). While we are far from achieving AGI, machine and deep learning — subsets of AI — are making considerable progress in many practical fields of interest. Historically difficult computational areas like computer vision, speech recognition, and natural language processing have made stunning advances in recent years using AI-related technologies.
When used correctly, these technologies exhibit remarkable predictive power, resulting in new or improved solutions to a variety of real-world problems. Most of this progress has come from using sophisticated machine learning algorithms to create predictive models, which exploit the vast amount of data most organizations now have in their possession.
While today’s AI capabilities are impressive, they have significant real-world limitations. And just as real-life Terminators are not in our immediate future, AI cannot magically provide solutions to most of today's general-purpose, real-world problems. But for the areas in which current capabilities exist, the results can be remarkable.
Against this backdrop, it's easy to see why many technology startups are rushing to adopt AI. Some startups have already made it a core part of their product offerings, while others are only starting to explore the possibilities. Since AI is a hot topic, an understandable amount of FOMO (Fear of Missing Out) is driving the perception that startups need to have AI.
For those startups who already effectively use AI, congratulations! You are positioned to take advantage of the current AI wave: creating value by solving genuine business problems. But, for those who are behind the curve, don’t worry. Take the time to understand the technology, validate its fit for you, and plan to implement it accordingly.
Remember, not everything requires or benefits from AI. Don't seek out problems to force an AI solution. Understand AI's current constraints and stay on top of its developments. While there may not be an AI opportunity available for your product or service today, capabilities may advance and offer new opportunities in the future. Be honest with yourself.
As a startup, you spend significant energy and focus on selling to potential investors, customers, or both. Every now and then, it's a good idea to reflect and consider what you're selling to yourself. Take a moment to review the following use scenarios for AI and see how they align with the current realities of your business.
In this scenario, AI capabilities are part of your current product and serve a genuine business need. You can fluidly discuss the models in use, current training and inference infrastructure, and share backing artifacts such as Jupyter notebooks (or equivalent) used to validate these models. Or you utilize commercial offerings from AWS (Amazon Web Services), Azure Cognitive Services, and similar, demonstrating their integrated use. You clearly understand the applicability and limitations of these models.
Here, AI has a legitimate use in the product — it’s on the product roadmap, or there is a way to get it there — but the claim is currently aspirational. In some cases, early exploration may already be underway to validate the approach and prove the use case. Either way, you may be looking for funding to build this area, and that’s great!
At the minimum viable product (MVP) stage, an accelerator such as LogicBoost Labs can provide resources to transition AI use from aspiration to reality.
In this situation, a business attempts to reframe a particular technology or technique as AI when it isn't. Your average person may buy that, but those knowledgeable of the space will find it disingenuous and in poor form. It's not worth risking opportunities for investment or new customers by pushing AI capabilities that don't exist. Furthermore, whether it's AI or not, be proud of the knowledge, product, and capabilities you do have, and don't oversell yourself. People in the know will appreciate that.
Handwaving AI is when a business doesn't understand what current AI systems can and cannot do. Often, these businesses don't know how to apply AI to their product or struggle to force a connection — commonly referred to as a solution in search of a problem.
Far too often in this scenario, large parts of a proposed solution are ultimately not in touch with any current AI capabilities. Recommended reading: “How to recognize AI snake oil.”
So, which category do you fall under, and why?
For non-tech savvy people, descriptions of or interactions with AI can seem like magic. In contrast, the people behind the curtain building these solutions are well aware of AI's challenges and limitations.
That being the case, if you invoke the use of AI during a pitch, make sure your claims are grounded in reality. You may not know whether your audience is coming to the table with a wealth of knowledge on this subject. So, don't put yourself in a position where your perceived solution is redefining or handwaving AI. Be prepared to expand upon your initial assertions to a level appropriate to the audience and setting.
Access to and successful implementation of AI has traditionally been limited to highly skilled researchers or well-resourced organizations. However, as these technologies mature, their capabilities and ease of use are improving rapidly. Consequently, there has never been a better time for early-stage startups to understand these technologies and how they may provide value to your product offering.
Now, don't fall for the trap of blindly delegating the solution of complex problems to AI or assuming that the exploitation of any data has a ready-made AI solution. AI is not a magic box that can solve arbitrary problems on a whim.
Instead, find a single problem where you can improve or introduce a new capability. Understand which AI capabilities currently exist that are actually helpful to your startup. And while the temptation may be to use the latest, cutting-edge capabilities, start with the most straightforward solution. Prototypes can also be invaluable for validating assumptions and refining the data, models, and other supporting technologies you use.
If your exploration with AI shows promise and is targeted for use in a product, prepare yourself for ongoing maintenance. Unlike other technology assets, which you may be able to set and forget, this may not always be the case for AI solutions. Often, models continuously evolve to reflect a constantly changing data landscape. A model built on today’s data may perform poorly against tomorrow’s reality. New patterns may emerge, requiring tweaks or a complete rethinking of initial assumptions.
As such, introducing AI requires a level of commitment that can sometimes be challenging for startups. It requires sufficiently knowledgeable team members who can spend the necessary time for ongoing evaluation and refinement of your AI solution.
If your startup is using AI today, be prepared to answer questions in the following areas:
● What business problem is it solving?
● Can you demonstrate the stated capabilities and the systems/artifacts that drive them?
● How is the data gathered and prepared?
● What models are you using?
● How is the model trained, validated, deployed, and retrained?
● Are there future scalability issues to be aware of?
● What are the economics around model training and inference? Is this factored into your pricing model?
● Do you have ethical and bias controls in place?
● If used in a regulated industry or mission-critical setting, what measures are in place to ensure compliance and understand failure modes?
● Who on the team is the resident AI expert? Do they have the skills and experience needed to evolve and refine the offering?
The point is, clarify whether AI for your startup is aspirational or in actual use. Describe prototypes and experiments that are on the horizon. Most MVP-level product offerings are rough around the edges, especially around AI management and deployment—sometimes referred to as MLOps — and that's perfectly fine. In fact, it may be one reason you're looking for additional resources. But, no one expects a startup to have a fully automated MLOps workflow on day one. You can incorporate these capabilities appropriately over time.
Given the state of AI technology today, it is imperative that technology startups take the time to understand the current landscape and how it may apply to their product offerings. Honest self-assessment is an essential part of this process. It is incumbent upon founders and key personnel, technical or otherwise, to fill any knowledge gaps and keep updated with current developments. Given the rapid pace of ever-increasing capabilities, this is a daunting task — but the payoff may be significant.
AI technology has never been more capable and accessible than it is today. As such, it presents numerous competitive opportunities for startups. However, startups must ensure that any plan or inclusion of AI is grounded and appropriate. Don’t let the excitement outstrip the pace of your reality. Remembering these points will help keep your goals and execution realistic and ensure that communication with investors and customers strikes the right balance between sharing your excitement and what you can deliver today.