Brave New World
5 brave predictions for the future* of the GenAI Enterprise world
1) Extinction of use case companies – as a product manager I was trained to believe that the path to success was to focus on a concrete user problem, understands the user, and builds amazing UI and business logics that address this problem. The underlying assumption however, is that building these solution requires unique resources and expertise. Companies monetize these expertise, and sell their solutions at a premium.
While this will not completely disappear, I expect that for many use cases this premium will shrink to zero. Instead, many of these solutions will be sold for the fixed cost of building (which might also shrink to zero~), as opposed to sustained margin-heavy licensing or subscription fees. The key factors that will contribute to that:
- The marginal cost of building these solutions will go to zero — as a result of continuous commoditization and automation of the entire GenAI stack — chips, models, tools, multi-agent frameworks, RAG and overall connectivity to data and APIs, horizontal no-code and low-code flow building systems, automated coding and app building. Building many of these use caes will be equivalent to today’s use of an excel sheet.
- Fierce competition amongst the hyper-scalers and large players at the lower levels of the stack – GPU and compute providers, model providers, vector DB providers, horizontal framework providers – all will offer higher level solutions at cost / for free, just to attract customers to their platforms and to achieve stickiness
- Entrepreneurial , efficient, and otherwise starved implementors / integrators / contractors will knock on enterprises’ doors, offering solutions at the price of building it, adding further competitive pressure.
We are currently seeing many startups building such vertical use cases and flows on top of LLMs and GenAI. For many of these, I struggle to see the long term defensibility and value creation to sustain a growing company.
2) Deflation of the SaaS Stack– the need for complex SaaS stacks, with hard coded modeling of business data, flows, and UI will significantly reduced. Instead, it will be replaced by data oceans, data infrastructure that is optimized for machine access (as opposed to humans), easy to build / on-the-fly generated UIs and business flows, agents, and marketplaces for task fulfillment (see next item). Overall, the SaaS economy, including the attached professional services industry will shrink into a much more efficient, lean, and mostly automated stack.
I expect that many existing SaaS players will start seeing the early signs of this in the near future (see item #5 below). After failing to spray GenAI on their existing offerings and / or moving at the necessary speed — a FOBO (Fear Of Being Obsolete) fed startups-shopping-spree will take place, offering a nice exit path for a few smart / opportunistic entrepreneurs and investors.
Somewhat related, as I expect that much of the inference will happen on highly optimized models at the edge, the fate of the cloud platforms is also somewhat up in the air (pun intended).
3) The emergence of agents-and-data marketplaces –the idea is simple: buyers advertise their needs in the form of task to complete or data needed, sellers advertise their services and data (in some standard open schema), and intelligent marketplaces that match them together, in a scalable, transparent, and efficient manner. Eliminating unnecessary middleware, middleman, and middle-tiers.
In a bit more detail:
- Buyers – the buyers simply search in the marketplace for the service, data, or task they need. These can be potentially highly complex long running tasks. The buyers can (automatically) publish in multiple market places. Models then translate and breakdown these high level tasks and requirements to the data and services offered in the market. The buyer gets multiple bids and can (automatically choose the preferred provider based on price, quality, time to complete, trust, or any other (automated) preference or optimization. The buyer pays to the marketplace, which then shares the revenue with the data & services owners and sellers.
- Data and service owners – individuals and organizations that are the legal owners of the data, IP, or service offered. A consumer is considered the data owner of their own personal data in any 3p service they consume. So while the owner isn’t necessarily the provider or seller, they are the legal owner and are compensated for that. A data owner can decide and control whether to allow the selling of their data and services, and under what conditions. The data owner can choose how to be compensated. The data sellers can compete on how they compensate the owners. I am especially excited about this, since finally we have an opportunity to monetize data directly and transparently (as opposed to ads targeting that is neither), where the data owner is fairly compensated and is in full control.
- Data and skill seller – these are individuals and organizations that specialize in data collection / curating / cleaning / combining, model (as a service) training, agent building, and overall pricing, packaging, and marketing – they do the actual bidding in the market. In many cases the owner and the seller will be the same party.
- Marketplaces – running on top of the data oceans and connected to all the agents. Can compete on access to data and agents, intelligent routing and bidding of data and services, fees, … The data and services themselves are open standard and independent of the marketplaces. In a way, analogous to the web content (open and shared) vs Search engines and browsers.
- Technology providers – the invisible lubricant underneath it all – running data oceans, data access infrastructure, agents frameworks, model hosting, etc.
4) Humans as agents – one cool twist of the marketplace is that humans can also compete as agents (individuals or teams) — high latency, low availability, high quality. Thus allowing them to monetize their knowledge, skills, and trust-credit. This opens an interesting new venue for the future of freelance work. Turning the tables a bit: AI running the place, and we are the agents… Brave new world!
5) The acceleration by GenAI disruptors – When talking to people about this brave new world, a common response is that it will be slow and gradual due to the speed in which enterprises are (not) moving. I agree with the sentiment , but I disagree that they will be the ones setting the pace. Rather, they will be dragged , kicking and screaming at great speed, by the real disruptors. These disruptors will offer old services for new prices: $10 an hour for an expert lawyer, $10 a project creative agency, $8 an hour software development team, $3 an hour shrink, …). Once trust in these services will be gained, and it will, the world as we know it today — will never be the same again.
I am also confident that as some of these old models will deflate, other new and completely unexpected ones will emerge. Exciting new world!
*How far are we? about 10 years.
