In thinking about applied AI in the business world, we have a lot of different aspects of approach to choose from. We can inspect markets to see where the opportunity lies, or solve a customer’s pain points. We can engineer something that will make stakeholders, or consumers, say “wow.” We can try to predict where things will be in 2 years, in 5 years, in 10 years, which is where thought leadership comes in.
Or we can apply all kinds of jargon. We can look at the application of AI to SaaS B2B situationships, in the junction of procurement and ROI. We can ponder the deployment of 3PL strategy tools to an LLM porting a play on model distillation to a platform.
Ok – that’s probably enough of that. Another alternative is to go along with where I’ve been hearing conversations go now, in terms of ROI for business. One of these is the “vertical” market.
What is a Vertical Market?
So when people talk about a vertical market, what do they mean?
Well, as a student of English, and not of corporate jargon, you’d be forgiven for thinking that a vertical market somehow scales up – like a skyscraper.
Instead, what people mean by a vertical market is a market that serves a particular “vertical” or in other words, a sector of business.
“A vertical market is a specialized business sector focused on a specific niche, where companies tailor products and services to the unique needs of a defined customer group, unlike horizontal markets that serve broader audiences across industries,” writes Julie Young at Investopedia. “Targeting a vertical market can offer deeper expertise and potentially higher margins, such as software built specifically for hospitals or financial firms. However, verticals often have limited market size and higher barriers to entry, creating both focused opportunities and more intense competition.”
I was having trouble reading this, so I asked ChatGPT to simplify. Twice. Here’s what came out.
“A vertical market is when a business focuses on one specific type of customer. For example, making software only for hospitals instead of for every kind of company. This can be good because you know that customer really well, but there are fewer buyers and more challenges getting in.”
Here’s what happens when you ask it to simplify again:
“A vertical market is when a business focuses on one kind of customer. For example, making software just for hospitals. It’s more specialized, but there are fewer customers.”
There. That’s better.
Applying AI to Vertical Software
So presumably, they called it a “vertical market” because it was serving a single defined type of business on multiple levels. Then, “vertical software” would mean software that serves a particular vertical, or customer base, like hospitals.
With that in mind, I wanted to let you into a bit of a panel discussion I saw at Stanford recently entitled “Applied AI: Turning Industries into Innovation Engines,” where a series of experts talked about this. The group included Sri Pangular of Mayfield, Bratin Saha of DigitalOcean, Lisa Dolan of Link Ventures, and Philip Rathle of Neo4J. (disclaimer: I am also involved in Link Ventures).
At the beginning, the panel talked about workflows, and training, and automation of knowledge work, which is happening quickly, in the context of providing software for various verticals or areas of business.
Dolan talked about the old days of training:
“If you remember, when you were training your junior employee, you would not allow them to get to the next level until they aced the low-level work,” she said. “And so what we need to do is actually train, in the enterprise, train our agents similar to how we would train a low-level employee, and then the enterprise will get the trust, and then the enterprise will really be able to roll it out in different verticals.”
“This is really the first time, with generative AI and LLMs, that you can really automate knowledge work,” said Saha, mentioning a list of relevant domains: healthcare, finance, and legal, to name a few.
“The average cost of the employee, or of the knowledge work, is quite high,” he continued. “So if you can automate some of it, the ROI is there.”
Finding the Core Applications
“In terms of just day to day processes, there’s a lot of cost savings opportunity there,” Rathle said. “But where the real money is, and where the real value is, is in core domains, core applications, where the stakes are high. So by definition, when the stakes are high, the quality of a good answer is very valuable, and the impact of a bad answer can be quite detrimental.”
Later, Dolan balanced this kind of thinking, suggesting that on the one hand, those playing around with CRM or other siloed ERP components are not on the cusp, but that it’s crucial to automate centralized processes, while also acknowledging that “we have to walk before we can run.”
The panel also addressed pricing.
“Ultimately, one needs to choose: are you horizontal? Are you vertical?” Rathle said. “The beauty about going vertical is: the higher you are in the stack, the closer you are to the end customer, the more you’re perceived to be generating the bulk of the value, and the more you can get people to buy into new kinds of value-based pricing models.”
Dolan talked about a move from seat-based pricing to throughput, and you can hear her break all of that down in the video. Here’s how she concluded that thought:
“I would say we see most companies that may just be a wrapper and are ultimately giving a lot of their their ‘lunch’ to Claude … looking for, really, those companies that own the customer,” she said, “and that means owning the workflow, and then the learning process, so they’re constantly iterating on their model and self-learning over time, but you need to own that own (sic) customer to be able to do that effectively.”
Rathle agreed: don’t price AI agents like humans.
“You don’t know how many human beings are going to be at the other end of that job in a few years,” he said. “I’ve already seen some of the larger companies charge on a per-agent basis, treating an AI agent as a human seat. And buyers really don’t like that. And you can get into all kinds of definitional ambiguity. So it’s just not a sustainable model in the long term.”
A Teammate, not a Tool
At the end, Pangular, moderating, asked about making AI a “teammate, and not just a tool” for enterprise.
“I think it’s going to be a continuum,” Saha said. “You start with (AI) being a tool, you start with (it) being a really, really good tool, and then slowly, (it becomes) more of a teammate. Because I think, I think there’s a trust factor involved in what kind of tasks you delegate, and what kind of answers you get.”
Also…
“I think it’s a continuum,” he then reiterated. “You start at one point, and then you delegate more and more, and you’re working more and more closely. And you know, ultimately, you get to a point where some real tasks are being handed over, and then it’s a real teammate. And then, in terms of the org structure, you probably end up in a situation where a person is also managing a lot of agents … that’s probably where we end up.”
All of this highlights a few basic, or perhaps more advanced, ideas about the impact of AI on business. People are going to automate, or be automated. Businesses will want to analyze cost savings. Agents will get more sophisticated. And amid all of this, we’ll keep thinking about the right ways to harness a technology that really, at the end of the day, is pretty powerful stuff.
