What are the most effective methods of evaluating AI investments?

Evaluating an AI investment involves understanding the startup’s target problem, market size, and competition. It’s crucial to assess how the startup’s unique AI solution stands out and its market viability, focusing on both technological innovation and practical, real-world application.

To begin with, it is essential to recognize the industry environment and the conditions there. Many AI start-ups have their origins in the IT sector and not in the industry of the respective application, which means that fundamental mistakes are made in the business model. To develop an AI application in the healthcare sector, I should first recognize this and the legal requirements.

Grasping the core issue is foundational to AI startup evaluation. It's about discerning the market gap the startup aims to fill and the efficacy of its AI-driven approach. Key to this understanding is exploring the size and potential of the market, identifying the target customer base, and recognizing the competitive landscape. Distinguishing the startup's AI capabilities from existing solutions is vital, as is evidence of its value proposition and product-market fit. These insights determine the startup's relevance and potential for impact in the market.

AI companies key criteria for investment should include the engine to be evolvable with new use cases. what are your perspectives on the list of criteria to see how to assess if its a right investment

In my professional journey, I've observed that comprehending the problem is akin to a physician diagnosing an illness before prescribing a remedy. I remember consulting with an AI start-up struggling to gain market traction. Upon deeper examination, it became evident that they had developed an intricate solution in search of a problem, rather than the other way around. We reoriented their approach to first identify a genuine, unmet need in their target industry, and then tailored their AI solution to meet that specific pain point, which ultimately led to a significant turnaround in their business trajectory.

El tema de "comprender" el problema es la clave. No es cuestion de medir sino de comprender. La cuestion es como probar de manera rapida/agil al menor costo posible para testear el efecto que puede generar en el mercado de demanda. Hoy hay furor por hablar de AI pero todavia esta en una fase que es para el mundo world class y el middle market esta tratando de salir del excel. Lo importante es si hay efectivamente parametros objetivos que permitan evaluar si va a ser valorado por los usuarios y esta es una tarea muy compleja que sugeririra tratarla por metodos cualitativos y por personas que permitan comprender lo No dicho

In my experience, we have developed a matrix similar to Eisenhower matrix but a little more layered which helps us to determine where each AI investment is placed, its importance, urgency and impact, and how to make investments accordingly.

In the current AI moment, it's tempting to back tech-first ventures with "just" great teams and technology. But I believe it's not about the solution in search of a problem—it's about the problem itching for a solution. 🧩 In my experience, the key is to dive deep: Engage with potential users to understand their struggles and former fixes. Does the AI truly address the need, or is it tech for tech's sake? Real value emerges from the user's perspective, not the code's complexity. Let's invest in AI that makes a tangible difference.

This initial step rightly emphasizes the importance of comprehending the problem an AI startup aims to solve. By probing into the pain points, market size, and competitive landscape, investors gain insights into the startup's strategic positioning. Questions about differentiation and validation underscore the need for a clear value proposition and market fit. This approach aligns with the principle that a well-defined problem statement is crucial for assessing the viability and potential impact of an AI investment.

I once mentored a rising AI enterprise where the founders came from impressive academic backgrounds but lacked real-world industry experience. My advice to them—and to investors—was to build a balanced team. They diversified their team by bringing in industry veterans with a record of commercialising technology which complemented their academic acumen. The surrogate mix of visionary technologists and pragmatic industry experts proved to be a recipe for success, validating the idea that a team's capability can pivot an AI startup from potential to prosperity.

The team should be as interdisciplinary as possible, i.e. AI experts, but also people from the industry, marketing, law, etc.

In my experience, we track teams for IT unicorns and unicorns trying to solve a similar problem (using other non AI tools) , use their initial teams as benchmarks, rank the team and use the weights on the ranking for the decision making process.

Recognizing the significance of the team, this step rightly highlights the impact of a skilled workforce on the success of an AI startup. The emphasis on technical skills, domain knowledge, and past achievements aligns with the understanding that a capable team enhances adaptability and execution. Acknowledging the pivotal role of the team is crucial, as it reflects an understanding that the people driving AI innovations are key contributors to a startup's resilience and ability to navigate challenges.

Is there something unique in the case of AI though which requires investors to think deeper before investing in more matured Ai companies especially when PE firms need to acquire

There's a large divergence on where folks focus on the conversation of the "moat" based on what's trendy or easy to understand vs what actually persists and creates real value. Algorithm(s) are rarely if ever a true moat. Algorithms get a lot of hype and buzz but as we're seeing in the LLM world there's a new flavor every month. Data is always a great moat because it can be really hard to get. Especially data that's relevant to critical business outcomes. A company that's built AI solutions on unique data assets is far more attractive that one that has built of generic or publicly available data sets. Finally experiences are critical. Great experiences drive more engagement. More engagement drives more data. Data is king for AI.

AI startups are like apples on a tree. Every day, a new one is growing. Hype is not making it easier to understand which is more valuable. The only way to truly assess is by studying the tech itself. OpenAI had good tech, and when they launched, they skyrocketed. Assess and evaluate the tech of new AI startups.

A key point is how much time and which skills it would take for a competitor to build the same solution. A startup may have spent a lot of time and effort writing code in the past that then becomes available as open source. These freely available frameworks keep maturing. They take more and more care of the heavy lifting of an AI application. To differentiate startups need a deep understanding of the business problem and the ability to translate this into a practical, domain specific solution. It is still hard to create high quality custom trained / fine-tuned AI models and run them in production at scale.

The technology should initially be open so that it is possible to make adjustments once mistakes have been learned from. Therefore, you should not commit to unsupervised learning only to realize later that supervised learning is necessary. The broader and more open, the better.

When evaluating a startup's potential, look beyond the basics: assess the exclusivity and quality of its data, a major competitive edge. Scrutinize the scalability of its AI model, focusing on data management and cost efficiency. Ethical AI practices and bias reduction strategies are important to implement systematically to prepare for future regulation. Consider the ease of integrating the AI solution into existing systems. Additionally, the startup's grasp of regulatory issues in sensitive sectors, market readiness, and potential adoption rate are vital indicators of its success potential.

Drawing from my experiences, the true mettle of an AI startup often lies in its capability to demonstrate a technological edge that's not easily replicable. I recall a case where a fintech startup's AI-driven risk assessment model was initially alluring. However, upon technical due diligence, it became clear that their model lacked proprietary algorithms and relied heavily on open-source libraries. Eyeing technological uniqueness and the potential for patents or trade secrets can signal a more secure investment.

In my opinion, AI model explainability and auditability is a key one to be added without which trust errodes. (e.g. Aleph Alpha)

The focus on technology evaluation is crucial, emphasizing factors like uniqueness, scalability, and alignment with industry trends. Assessing competitive advantages and scrutinizing the intellectual property portfolio reflects a thorough examination of the startup's innovation and protection strategies. This step rightly recognizes that the technological foundation is a core determinant of an AI startup's potential success. It aligns with the understanding that cutting-edge and well-protected technology is essential for sustained competitiveness in the AI space.

The current AI companies are all based on heuristic based algorithms. Hence, its important to understand the data that is being used , what alglorithms are being used to analyze that data to come out with AI outputs. Otherwise the model will not be create robust long term outputs which could actually replace human intervention and thus lose out to other AI engines in the future.

An anecdote that comes to mind involves an AI-driven health-tech company; they boasted rapid user acquisition but suffered from high churn rates. Deeper analytics revealed that while their AI was exceptional in diagnosis, user experience lagged, prompting disengagement. This led to a strategic pivot focusing less on user acquisition and more on retention, spurring a redesign of their user interface to create a more engaging, intuitive experience. The lesson here is clear: traction is multifaceted, and startups need a holistic approach to fostering growth.

Assessing a startup's traction is critical in evaluating its potential for success. By examining revenue trends, user engagement, and retention rates, investors can gauge the viability and scalability of the business. Growth levers such as customer acquisition tactics and network effects offer insight into the startup's ability to expand its market presence. Moreover, indicators like product-market fit and customer satisfaction are telling signs of the company's alignment with market needs and its ability to deliver value. A thorough analysis of these aspects is essential for informed AI investment decisions.

This step underscores the importance of tangible results, focusing on key metrics like revenue, user base, and engagement. By delving into growth drivers, acquisition strategies, and indicators of market validation, it acknowledges that sustained success involves more than just technological prowess. The emphasis on product-market fit and customer satisfaction aligns with the understanding that a thriving user base is a strong indicator of an AI startup's viability. This step rightfully recognizes that, in addition to innovation, real-world impact and user acceptance are critical components of evaluating AI investments.

As with other start-ups, it makes no sense to use classic valuation methods such as DCF or multiples in the seed and first-stage financing phase. Where should the estimates come from? Instead, the business model and, in particular, the financial model should be well thought out.

In my experience, the most contentious discussions often revolve around a startup's valuation. I once advised on an AI startup during a Series A funding round. It had a strong technological proposition but overestimated its valuation based on potential rather than present value. We worked on creating a more balanced valuation, considering not only the AI potential but also current assets, revenue streams, and market conditions, leading to a successful funding round that satisfied both the founders and the investors.

Valuation estimation is a pivotal step, highlighting the multifaceted approach required. The mention of various valuation methods, including discounted cash flow and venture capital models, reflects the need for a comprehensive analysis. Acknowledging the significance of industry context, geographical factors, and market trends demonstrates a nuanced understanding of the AI landscape. By emphasizing the importance of deal terms and expectations, this step recognizes that valuation isn't solely about numbers; it involves aligning investment with strategic goals and market dynamics.

Valuation is a complex but essential part of AI startup assessment, blending art with science. Employing various valuation methods like DCF, multiples, and the Berkus method helps triangulate a fair price range. It's important to contextualize these valuations with market dynamics, including sector growth, regional factors, and investment trends. Additionally, the startup's stage and how it stacks up against peers are critical considerations. The appetite for investment and deal specifics further refine the valuation, ensuring a comprehensive financial picture before committing capital.

I cannot overemphasise the importance of thorough due diligence, having seen investors cut corners to their detriment. One particular case that comes to mind was an AI startup that seemed to have everything going for it—until financial due diligence revealed discrepancies in revenue recognition methods. Further probing justified investor caution as it uncovered a pattern of overpromising and underdelivering, prompting a renegotiation of the investment terms to provide provisional tranches tied to performance milestones.

One of the biggest gaps I see when advising firms on AI investments is the lack of operational expertise when conducting due diligence. This always brings me back to Clark's quote: "Any sufficiently advanced technology is indistinguishable from magic." Diligence based on the premise that the technology is essentially magic results in insufficient interrogation of the underlying hypothesis of company you're looking to put capital into. Having good advice that have brought practical AI solutions into market, at scale, is invaluable. It lets you ask the right questions, kick the tires appropriately but also offer the right guidance to the management team.

AI investments are quite difficult to assess in the current environment. There is a lot of fuzziness in this spaces as business models are evolving and technology is evolving faster than that. I would suggest the focus on team and a clear path to profitability should be very high on the criteria list. Since lot of dry powder is chasing AI Startups, it's quite difficult to negotiate valuations in this space.

Know where in the AI stack the company is in: - Application layer - Middle layer (tools to build applications) - Foundational layer - Infrastructure layer Each layer in the stack have different characteristics such as costs, competition, moats etc. And therefore each layer will have a different investment approach.

Team, finance, technology, or all? Evaluating companies in the AI space requires going back to the basics. Focusing on the team means looking at how it was put together. Not every team will have worked together before, but is it put together thoughtfully. With limited Return on Investment data to use, look at quantitative and qualitative data to draw financial conclusions. Data used for the financial review works together for technology. How has the technology been implemented and what is the user experience? As has been done in the past with other technologies, it goes back to how well the team fits, does the normalized finances stand up to scrutiny, and does the technology solve a problem that users want.

It's crucial not to overlook the softer elements such as company culture and customer feedback. A B2B AI company I consulted with initially struggled despite strong technology and a good team. The issue, unearthed during an extensive customer survey, was the misalignment between the company's product roadmap and customer expectations. By fostering a customer-centric culture and aligning product development with customer insights, the startup was able to pivot effectively, leading to increased satisfaction and sales.

90% of AI and data science solutions do not make an end impact on business outcomes. Either because they aren't brought into end consumer experiences or into end employee experiences. 90%. How the firm is going to operationalize the solution and what its plans are for adoption and incorporation of the problems its solving for are absolutely critical... especially in the early stages. Getting the first few wins and evangelizers are incredibly important for getting the flywheel started, and I find that firms that can think through that often can accelerate more rapidly through earlier growing pains.

A particular insight I found extremely useful to establish a framework when viewing AI investments came from Sarah Tavel, GP at Benchmark. She states that a founder should talk about using AI to sell the "work" and not the productivity increase. This changes the sales cycle and makes its pricing akin to a human performing it. I've tried to look for that amongst the fluff in the newborn tech.

One thing that’s missed when looking at investing in AI companies is whether their advantage is truly sustainable - especially in a world where the “cost” of new models approaches zero over time. As shown in the last year, open source models are moving quickly so you have to assume that any product built on top of a proprietary model will eventually have an open source model to compete against. Does this mean that they’ll fail? No, not at all, but it does mean that they need to have some other advantages beyond a professed “data advantage” or “proprietary model” that will sustain them in the long run!

⛏️🔬 In evaluating AI investments, I always try to dig deep. It's crucial to identify genuine moats—what's truly proprietary? Look for robust proprietary data sets, tailored algorithms, and concrete IP that promise long-term defensibility. 🔮 And it's not just about today's strengths; It's also about about anticipating the future. The power of a self-reinforcing flywheel—feedback loops that amplify a company's unique value proposition—is indispensable. It's these elements that ensure an AI company doesn't just lead but continues to outpace the competition. That's the kind of defensibility I bet on

The article rightly emphasizes the team and problem(CVP). However, it overlooks two critical aspects: profit formula(PF) and competitive analysis. Understanding how a venture will profit amidst intense competition is vital. As incumbents integrate AI solutions, evaluating whether customers will pay a premium for AI-enhanced services is essential. For instance, would they choose AI-powered email, calendars, or tutoring over free options from giants like Outlook, Gmail, or Khan Academy? The democratization of AI technology enables the simplified creation of bespoke solutions like custom ChatGPTs - raising the bar for what constitutes an investable team. These factors are crucial to discerning a viable AI investment in today's market.

Its essential to consider data underlying the AI. Especially in context of Fintech AIs in India. Most of The financial data in India is still held by Government owned Public Sector banks and they don't have this data segmented or collated in a way that can be run through a ML or AI engine. Hence, using existing data may result in Erroneous AI outputs which may work well for small tasks but will fail finally as its data foundation was not the most robust