Genuine value creation through AI - an introduction for company operators
Tldr;
There is so much noise in the AI market right now that it makes it difficult for Operating Partners and CEOs to confidently launch AI initiatives that create value for their organization
You should ground all of your AI initiative planning around the three enterprise value levers: Increase revenue, reduce costs, increase valuation multiple
To identify the ways that AI can support revenue growth, start by considering more traditional revenue growth techniques that you would use to sell new or existing products to new or existing customers. Then brainstorm a series of “what if” questions that would help take each of those tactics to the next level
When assessing cost reduction opportunities for AI to address, think deeply about Level 1 opportunities (solutions that fundamentally improve your organization’s unit economics) before spending too much time on Level 2 opportunities (solutions that reduce resource needs in a specific function)
AI solutions can increase the valuation multiple that the market applies to your business, but only if you retain the IP to the solution that is created
Driving value with AI requires three distinct phases: Phase 0 – identify ideas worth prototyping and build data infrastructure, Phase 1 – select a subset of prototypes to “productionize” and monitor KPIs closely, Phase 2 – listen to your users and metrics and iterate to ensure your solutions continue to deliver value
Want to talk to our team learn more about how to generate value using AI? Get in touch
Unless it is your core responsibility to stay on top of the developments in the AI landscape, and even if it is, the industry is moving at a pace fast enough to make your head spin. Roughly 50% of Y Combinator’s most recent batch (S23) have the phrase “AI” or “LLM” in their name or company tagline, the MBBs of the world are producing thought leadership at a breakneck pace, massive early stage investment rounds are being raised, and people in your network are sharing their “AI expertise” on LinkedIn (whether they are qualified to give that kind of advice or not). You could spend all day reading the literature out there and still not know what to do about it…and we honestly do not blame you.
At the same time, if you are an Operating Partner at a fund or a CEO of a business, you are fielding questions almost daily about how you are going to use AI to “create value”. It is not our place to judge what other people are saying about the technology or the way that the industry is forming. To be frank, the fact that we are writing these long-form Perspectives is adding the ocean of literature out there. But we hope that if you came across this Perspective, it helps you develop your own mental model around how AI can create value for your company and we hope to do that by tying it all the way back to some fundamental business concepts that everyone reading this is likely familiar with.
Aligning on the definition of value
If ChatGPT, Bing Chat, and Pi were to walk into to a bar and explain to each other what value means, they would all say something slightly different. We did just that and here is what they told us:
“In a business context, generating value refers to the activities and strategies employed by a company to create worth, satisfaction, and competitive advantage for its customers, stakeholders, and itself. It involves maximizing the benefits delivered through products, services, or processes while optimizing resources, fostering innovation, and meeting the needs and expectations of customers to drive profitability, growth, and long-term success.” – ChatGPT
“Creating value is the primary aim of any business entity and it refers to giving out something of worth to receive something of higher worth. Value creation means producing something valuable (a service, product or employee benefit) in order to receive something more valuable in return (money, loyalty or customer satisfaction).” – Bing Chat
“Creating value is the process by which a business transforms inputs (such as labor, capital, and materials) into outputs (such as products and services) that are more valuable than the sum of the inputs.” – Pi
Like all responsible users of LLMs should, we took this information and did some of our own primary research about what value is. One of the best discussions that we came across was from a podcast that a friend in our network published this week. During their discussion they defined value using the following attributes:
It is the difference between what it takes to provide a product / service to the world and how much it is worth
It must be applicable to one or more people
It is not always monetary
It must be measurable against some kind of objective against which people assign worth
It’s a consultant’s favourite word
We like this as a general definition, but for the scope of this Perspective, we are going to talk about the specific value measure that most Operating Partners and CEOs optimize for; enterprise value (valuation). For the remainder of this piece, when we talk about value creation, we are talking about the steps that you take as an operator to increase the valuation of your company, and more specifically, we are going to discuss how to think about how AI can help you accomplish that.
A simple value creation model
Whether you are thinking about value creation in the context of AI or not, the levers that you have to increase your organization’s valuation boil down to:
1. Increase the organization’s revenue
2. Reduce the costs required to provide products or services
3. Increase the valuation multiple by which the market will measure the value of the organization
Increase the organization’s revenue
When we work with company operators who want to focus on their organization’s revenue potential, we start with a very simple mental model (figure below).
An often useful exercise to understand how AI technology can help to increase revenue is by first thinking about the more traditional ways that you might increase your company’s revenue. After you have done that, ask yourself a series of “what if” questions that take each quadrant one step further. You do not need immediate answers to your “what if” questions, but they will act as an incredibly useful research guide as you look into AI technologies that can help. The table below presents the analysis for a young entrepreneur’s neighbourhood lemonade.
You can then take your “What If” questions and do some basic research, read papers, and listen podcasts to generate ideas about how AI might be used to address each one. Within a couple of hours you will likely come up with a list that looks something like the following:
Use natural language processing on social media data and beverage reviews to identify trending flavors or predict the flavors that are likely to become popular
Use classification algorithms and data about local bakeries to rank potential partnership opportunities
Use recommender systems to suggest personalized products or bundles to existing customers based on their preferences and the preferences of customers like them
Use clustering algorithms to better understand the different buying patterns that your customers have
You might arrive at a similar list of AI use cases if you were to just search for “AI to help generate revenue for lemonade stands” but you would quickly get lost in the ether, not know where to start, and not be confident that it is the right AI initiative for your business. The benefit of taking this approach is that it allows you to inherently map the AI solution to revenue drivers. As an operator, you likely already have some intuition or data to suggest what “quadrant” you need to focus on first. Building up your analysis using the approach above ensures that you prioritize your AI efforts on the most impactful revenue drivers.
Reduce the costs required to provide products or services
When it comes to creating value using AI, one of the most widely discussed topics revolves around how this technology can save costs for organizations. The AI landscape is constantly evolving, with an increasing number of vendors providing function or industry-specific automation software. From AI-powered lease review to generating B2B cold outreach emails, to producing product marketing copy, the opportunities seem boundless. However, Operating Partners and CEOs should avoid hastily asking themselves, "What function should I apply AI to first?". Instead, a more strategic approach involves considering what we call "Level 1" and "Level 2" efficiencies.
Level 1 efficiencies refer to implementing AI technologies that fundamentally change the underlying economics of a business. These transformative ideas enable scaling with less incremental cost or increasing the gross margin on products/services in perpetuity. A prime example of this was seen in one of our previous Perspectives about an industrial business that leveraged AI to automate part inspection. By creating a system that allowed them to scale operations 30% more profitably than their legacy manual process, they showcased the potential of Level 1 efficiencies. Investing in such transformative AI applications can lead to sustainable cost reductions, increased profitability, and most importantly more profitable growth.
On the other hand, Level 2 efficiencies involve implementing function-specific AI solutions that reduce the resources required to perform specific tasks. For instance, automated contract review is a valuable use case in this category. When evaluating Level 2 opportunities, it is essential to identify areas where the potential for cost-saving is high and the AI technology, whether custom or off-the-shelf, has been proven and well-tested. It is particularly beneficial to target functions that may be under-staffed or time consuming, as automating these processes can yield significant gains in efficiency and cost reduction.
As an operator, it is crucial to prioritize exploring potential Level 1 efficiency opportunities before getting lost in the endless list of Level 2 efficiency options. Level 1 efficiencies hold the potential for substantial, long-term impacts on the company's unit economics. However, that doesn't mean you should ignore Level 2 efficiencies altogether. They can offer quick and valuable wins, especially in functions with high costs or resource-intensive tasks.
Increase the valuation multiple by which the market will measure the value of the organization
The final value creation lever to explore is the valuation multiple that the market applies to a business when valuing it. Valuation multiples, such as EV/Revenue or EV/EBITDA, are influenced by a multitude of factors, some of which a company can control, while others are influenced by micro and macroeconomic factors. In this section, we'll highlight two powerful ways in which AI can be leveraged to create value by improving a company's multiple.
Creating a Sustained Competitive Advantage with Custom AI
Custom AI solutions trained on proprietary data assets can be a tremendous sources of sustained competitive advantage. Let's revisit our lemonade stand example to illustrate this concept.
One of the critical capabilities for lemonade stands is procurement and inventory management – buying the right number of lemons. Keeping an accurate inventory is vital to avoid waste and ballooning costs that eat into profits and to ensure that you can serve all the customers that want to buy from you. Imagine an ambitious young entrepreneur who diligently maintains a detailed log of every transaction that includes time of year, outside temperature, weather, customer traffic, ingredients, sales, time of day, and waste levels over several years. With this rich database, they develop a custom AI model to predict demand and optimize inventory. Compared to their competitors who rely on guesswork or “off-the-shelf lemonade stand calculators” that factor in a limited number of variables, this entrepreneur gains a significant edge. Their custom AI-driven inventory management system consistently enables a more efficient operation, establishing a sustained competitive advantage. Owning and using proprietary data to build tailor-made AI solutions can significantly enhance operational efficiency, increase profitability, and ultimately positively influence the valuation multiple applied to the business.
Becoming a Licensor of Your AI Technology
Another approach to increasing valuation multiples is by becoming a licensor of your custom AI technology and generating recurring revenue from it. In our lemonade stand example, we fast forward a few years. The young entrepreneur has amassed an enormous amount of data and built an exceptional AI tool to estimate lemon purchasing needs. Instead of setting up the same lemonade stand year after year and battling the summer hear waves, they decide to pivot their business model. They become a software provider, selling their AI software to other entrepreneurs running lemonade stands in their neighborhoods.
By distributing their AI technology to the rest of the industry, they become a software provider and enjoy the higher multiples associated with being a software business. The advantages include higher gross margins, nearly negligible distribution costs, and a recurring revenue stream. The AI software provides a more consistent, predictable, and scalable way to operate the business, leveraging the domain expertise developed over the years. This transformation frees the entrepreneur from day-to-day operational challenges and elevates them to a software provider with an intellectual property-driven asset, ultimately leading to a higher valuation multiple for their business.
While these scenarios are simplified through the high-tech lemonade stand, the underlying principles apply to many industries and organizations who have significant data assets and expertise. The key to successfully implementing these techniques lies in owning the intellectual property of the AI solution developed and maintaining control of the data used to it. These strategic approaches to AI can create significant value and enhance the market perception of a business, ultimately leading to higher valuation multiples and a stronger competitive position in the marketplace.
Avoiding Pitfalls: Common Mistakes in AI Implementation
Before delving into the suggested approach for getting started with AI initiatives, it's essential to be aware of the signs that may indicate your AI endeavors won't yield the expected value. By avoiding these common mistakes, companies can set themselves on a path towards successful AI implementation.
Relying Solely on Best-of-Breed AI SaaS Solutions
While best-of-breed AI-powered SaaS solutions can offer valuable functionalities, relying solely on them for all AI needs will not lead to optimal results. Each business has its unique challenges and objectives, and generic AI solutions may not align perfectly with their specific requirements. It is crucial for organizations to assess whether off-the-shelf solutions fully address their business problems or if custom AI development and a more tailored approach is likely to create more value.
Delegating and Distributing AI Adoption to Middle Management
Pushing the selection and adoption of AI technologies down to middle managers can be a recipe for an uncoordinated disaster. Successful AI integration requires top-level buy-in, support, and strategic alignment with the organization's overarching goals. Leaders should actively engage in the AI implementation process, fostering a culture of innovation and continuous learning throughout the company to ensure efforts are coordinated and deliberate.
Exposing Proprietary Data to Multi-Tenant Software Providers
Handing over proprietary data to multi-tenant software providers without proper control over how the data is leveraged can pose significant risks. Data is a valuable asset, and protecting its confidentiality and security is crucial. Companies should carefully evaluate data-sharing agreements and ensure they maintain ownership and control over sensitive information while collaborating with trusted partners. Additionally, organizations should seriously consider how proprietary their datasets are and err on the side of caution when dealing with datasets that their competitors and suppliers are unlikely to have.
Keeping Data in Silos
Data fragmentation, where different parts of the business keep their data in silos, can hinder the potential of AI initiatives. AI thrives on large and diverse datasets, and keeping data isolated prevents companies from unlocking valuable insights and patterns that span across various departments. Integrating data from various sources can provide a comprehensive view of the business, leading to more robust and informed AI-driven decision-making.
Initiating Your AI Journey with Large Company-Wide System Implementation
Attempting to start the AI journey with a company-wide system implementation can overwhelm both the organization and its employees. Instead, taking a gradual and iterative approach can be more effective. Beginning with small, focused AI projects allows teams to gain experience, identify challenges, and fine-tune strategies before scaling up to more extensive initiatives.
By being mindful of these potential pitfalls, companies can lay the groundwork for successful AI integration that maximizes value and drives real impact. Building a thoughtful and strategic approach to AI implementation, centered around the organization's specific needs and objectives, sets the stage for leveraging AI's potential to its fullest extent.
An Approach to Build Value with AI
To guide operators on the path to creating value through AI, we typically assist clients through three distinct phases, each tailored to maximize the potential of AI solutions for their companies.
Phase 0: Prototyping & Infrastructure
In this initial phase, the focus lies in prioritizing AI use cases based on their potential value to the organization. Utilizing the techniques discussed earlier can help determine the most impactful areas to start implementing AI. Additionally, establishing the right data infrastructure is crucial. The availability and readiness of relevant data significantly influence the success and time-to-value of AI initiatives. Centralizing and ensuring data is in a usable format are paramount steps in this process. During Phase 0, prototyping AI solutions, defining KPIs, and launching them to a small subset of users becomes essential. This approach allows for real-world testing and validation, gathering valuable user feedback.
Phase 1: Productionizing
This phase involves taking the successful prototypes that demonstrated value and scaling them up for a larger audience. This process is known as productionizing. As the AI solutions are rolled out to more teams, gathering feedback through automated and manual mechanisms helps iterate and improve the solutions over time. Continuous improvement based on user input is crucial to refine the AI applications and ensure they align with evolving business needs.
In Phase 1, tracking a small number of KPIs (defined in Phase 0) becomes essential to measure the AI solutions' impact on the company-level value levers. These KPIs serve as critical performance metrics and guide the organization's progress toward its overarching objectives.
Phase 2: Iteration / Retraining
The final phase emphasizes the importance of embracing user feedback and closely monitoring KPIs to inform the next steps. Users' insights provide direction for enhancing the AI solutions and making them more valuable to the organization. Additionally, the data collected throughout the process enables re-training machine learning models to improve their performance continuously. Ultimately, this phase leverages user feedback, and KPI trends to ensure that the applications remain relevant, effective, and aligned with the company's value-driving objectives. By adopting this systematic approach, organizations can build AI solutions that create tangible value through increased revenue, reduced cost, and higher valuation multiples.