The Future is AI: How to Build Value in an AI Software Company Nov 19, 2023
This essay is based on the work of Val Bercovoci. It was inspired by a 10 bullet note and call to thinking harder about creating valuable AI companies. I wrapped some more examples and strategies around ideas using various AI tools and models. This is a work in progress, that should evolve into a valuation framework for AI focused software and services companies. RT
The world of artificial intelligence is evolving at a breakneck pace. In 2023 the AI startup Anthropic made waves by unveiling its conversational chatbot Claude.
This comes on the heels of OpenAI debuting ChatGPT, showing the massive potential of generative AI. However, the dust is far from settled, as evidenced by the recent ousting of OpenAI CEO Sam Altman. As AI capabilities grow exponentially, how can software companies position themselves to capture value in this burgeoning market?
The core of any successful AI software business lies in data. Without abundant, high-quality data to train algorithms on, even the most advanced AI techniques falter. This reliance on data acts as both a barrier to entry and a competitive moat. Companies like Google and Meta that have access to oceans of data have inherent advantages in developing AI systems. For startups and smaller firms, obtaining sufficient data remains an obstacle.
Creative data sourcing is thus essential. Strategic partnerships with data-rich companies can be hugely beneficial. For example, in 2016 DeepMind, acquired by Google in 2014, signed an agreement with Moorfields Eye Hospital for access to de-identified ophthalmology data. This allowed DeepMind Health to develop AI-based products for spotting eye disease. Such win-win data sharing agreements are critical to fueling next-gen AI.
Of course, data alone is not enough. Talent is the other indispensable ingredient for AI success. The top minds developing novel machine learning methods are scarce and highly sought after. Attracting and retaining elite AI researchers requires compelling projects and compensation packages comparative to Big Tech salaries. Startups must get creative, offering equity and the lure of cutting-edge R&D to recruit top talent.
Once data and talent are secured, delivering value via AI software involves crucial product development decisions. Will the software be industry-specific or general purpose? Will it be sold as a product or operate as an AI-as-a-service? There are merits to both approaches. For instance, large tech firms like IBM and Microsoft have successfully marketed their deep learning software tools and APIs to enterprise customers across diverse sectors. Meanwhile, emerging players like Perplexity.AI and H2O.ai focus on making AI more accessible to non-technical business users.
Use cases that demonstrate concrete value are imperative when bringing AI software to market. For example, the healthcare startup Zebra Medical Vision has built its reputation on AI algorithms that analyze medical scans to detect disease. Its focus on solving a real-world problem provided a beachhead for expansion into new clinical areas. AI software companies should ground their offerings in tangible improvements whether it be increased efficiency, cost savings or enhanced experience.
Monetization of AI software also warrants careful consideration. Licensing and subscription models are common, though freemium tiers can help onboard users. Transaction fees based on usage are another option, employed by algorithmic stock trading platforms like QuantConnect. Creative thinking is useful here, as novel pricing approaches can help strike the optimal balance between market penetration and profitability.
Of course, there are open questions regarding ethics and regulation that AI software companies must also grapple with. AI systems have demonstrated troubling and damaging unlawful biases that can perpetuate discrimination. Until algorithms consistently make fair and equitable decisions, public trust will remain elusive. The companies in the industry got a layer of self-regulation to ensure that they follow the rule of law.
To uphold principles of accountability and transparency, AI developers should take proactive steps like supporting AI licensing, transparent red team results, external audits and documentation of model logic and data set inputs. Following ethical AI best practices is not just altruistic—it makes business sense by boosting consumer confidence and trust.
Regulatory scrutiny of AI is heightening as well, though policy lags behind tech advancement. The European Union’s AI Act lays out requirements around data and algorithmic transparency. While the legal landscape remains nebulous, AI software companies would be prudent to get ahead of regulatory risk. Initiatives like responsible AI training and thoughtful, lawful data collection demonstrate corporate social responsibility. Creators of content must be respected. Licensing and royalty frameworks must exist and revenue must be shared.
This also implies the taxation frameworks must change for these types of companies that use data to create and transform other data.
For individuals, the ascendancy of AI also raises important considerations around data rights and ownership. After all, it is the collective data people generate that fuels AI progress. Yet most users surrender their personal data freely to private companies like Google and Meta that monetize it to power their systems. Calls are growing to rebalance these dynamics in favor of greater user control.
Initiatives like data unions and data cooperatives are emerging to consolidate individuals’ data value. The idea is to pool user data which can then be selectively licensed to organizations the group approves of. This helps ensure ethical uses of data while providing potential revenue sharing back to the data contributors. Such user-centric approaches to data aim to make AI work for everyone’s benefit.
The meteoric evolution of AI guarantees more upheaval ahead. Incumbents will be displaced by scrappy startups which themselves face threats from future innovations. What endured competitive advantages still apply in the age of AI? For companies, proprietary datasets and reserves of talent seem unassailable. For individuals, collective control over personal data looks increasingly vital. The AI landscape is shifting rapidly but those realities appear foundational.
By mastering the fundamentals—data, talent and use cases—AI software companies can thrive amid turbulence. Partnering strategically, making ethical products and watching the regulatory horizon will also help navigate uncertainty. For open and user-driven AI to prevail, ordinary people may need to stake claims to their data.
The era of AI has dawned and those that align interests to harness its power ethically will define the future.
This article is based on the thinking and work by Val Bercovoci on LinkedIn Nov 18, 2023
Founder & CEO @CLICK. Building the user-owned Al...
Major takeaways from the OpenAl Red Wedding:
- Al models and companies will come and go. Evolution playing out at unprecedented corporate speed
The science of Al remains limited by data, even after decades of research trying to break that bottleneck
- The engineering of valueable Al is completely dependent upon data
- For a business, safely harnessing your data to leverage a growing Al ecosystem will be existential
- For individuals (collectively we generate the most data) establishing ownership with optional licensing is equally existential.
Rob Tyrie is the Founder, CEO and Principle Advisor of Ironstone Advisory and a Founder of The Grey Swan Guild, a virtual Think Tank. 🧲🦢
