Ai as a Service

Streamlining Businesses for an Automated (and better?) Future

Photo by Nathan Wong

The beginning of a new era

The first era of mass market software was the desktop era. It worked with limited internet connectivity and relied on on-premise software for people to exploit the power of computing for their work and everyday lives.

When internet connectivity became powerful and ubiquitous enough, that on-premise software moved to the cloud and continuous updates and improvements on software gave birth to a reliable subscription model. Constant improvement of the software, with new features and new data services delivered value on a recurring basis and allowed for a subscription business model to be built on top. This became Software as a Service, or SaaS. The idea being you pay for the service of getting updated software all the time.

In the last two decades, software started to eat the world: more infrastructure that made it easy to build software was created and, as a result, more and more workflows started to be digitized into software.

In the same way better internet enabled SaaS to exist (ie. to avoid having to install the software on premise, and instead run it on servers and “stream” the software live via a website), AI may unlock a new set of possibilities to enable higher productivity.

Agentic AI, or the creation of software that can evaluate tasks, generate a plan and execute on those tasks is a giant leap from traditional software making. Up until now, software had to be mapped to a step by step process a person takes to get a task completed. You need the person to stick to that process, and you need the person for the software to deliver a result.

Agentic AI removes both requirements:

• No need for a predefined process.

• No need for a person to actually perform tasks in the process.

This means that while software enabled people to be efficient and effective and performing tasks and delivering, ultimately, a service, Agentic AI can jump straight into delivering the service (without the person).

Can this happen today? Well, not quite. Yet.

Enter AI as a Service

We believe AI technology is good enough to enable an intermediate set of companies: AI as a Service companies.

There currently are three realities about AI technologies:

1. The technology is evolving very rapidly and is, still, hard to use effectively by end users. The tooling required for companies to implement and use it is incipient at best.

2. It is not very well understood how the technology will behave in terms of output, so learning is quickly being acquired in the traditional trial and error way. People all over the world are playing with it, and sharing their creations.

3. LLMs still hallucinate a great deal. As a silly example, I recently uploaded a dozen statements from our utilities providers and asked to parse the amounts and dates and it made up transactions that had never happened and were not in the statements. Oversight is still very much a must.

In essence, it is still early to fully deliver a service, unsupervised. But it is a really powerful tool to increase the operation leverage of a service business, and it may be the way in which you end up having ai workflows that ultimately deliver services without human intervention.

In a recent interview, Jensen Huang was describing how companies are likely to infuse AI into the their operations and become more efficient, and enable employees to explore bigger/different challenges. But more importantly, Jensen talks about the true question he asks himself (and i believe everyone should be asking): how long can I be relevant? Your ability to stay relevant in the face of AI, requires you to use AI to continue to learn and perform to the level that ensures staying relevant.

Everyone will be a CEO

Jensen mentions that managing a direct report is not very different from prompting the AI. And as AI Agents start to permeate our lives, it is likely that we will all grow to become managers. Great delegators will likely become the top performing individuals, and promotions will likely become more meritocratic. Our own personal output will become AI as a Service, both on a personal and professional level.

In the near future, it is very reasonable to assume that you will interact with AI several times a day, and will morph how we behave: how we consume information, how we analyze it, how we research, how we learn, and how we make decisions. We will become bionic beings: not because we embedded machinery into our bodies, but into our minds.

Similarly, companies will blend people (biologic thinking agents) and AI agents (technological thinking agents), and improve productivity at the company level.

The rise of previously under-appreciated companies

Services companies have been broadly rejected by technology investors. With a few exceptions during the ZIRP era, in which some crazy tech-enabled services businesses were treated as software businesses, services businesses have not been seen as technology businesses due to a margin structure that isn't attractive to investors.

Agentic AI has the potential to change that. With trillions of dollars in the service economy, AI is unlikely to replace that but, rather, to improve the economics of those service businesses. Will that open up the appetite from VCs to these type of investments?