I'm back to writing about my favorite topic - Agents (digital workers). I recently saw this post from Greg Isenberg
the next wave of startups won't launch with marketing teams, they'll launch with AI agents running everything 24/7 - from content to acquisition to analytics.
He even shared a visual that shows all the companies operating in the Agent market. Gives me a headache and I am all in on agents. Wonder how organizations will make decisions on selecting vendors. I guess this is the reason Accenture Q4 topped 1Bn in AI bookings.
Andreesen Horowitz is talking about Vertical SaaS and how GenAI will enable a new breed of niche vertical apps that will tackle workflows for businesses that have traditionally been left out because the work was too manual.
Software replaces labor, making small markets bigger. Imagine a market where a VSaaS company could charge 10k potential customers $1k per month (including fintech revenue). This produces a total market size of ~$120 million. Not very interesting. Now imagine the VSaaS company could win the sales, marketing, customer service, and financial back-office budgets of those 10k customers and charge $10k per month. The market size is now $1.2 billion
What I do see is that everyone is either building agents, building applications with agents or providing platforms to build agents.
There are agent marketplaces popping up like Agent.ai from Hubspot. As companies like Hubspot, Salesforce, AnyQuest etc. create low code / no code tools for business users to build agents, there will be a proliferation of user generated agents.
Then there are going to be applications like Rox.com,Waldo.fyi that have swarms of agents that go do a bunch of tasks that get aggregated into a workflow and an output.
If looking at the AI Agent vendor landscape can give me a headache, the thought of gazillion agents proliferating is hard to conceive. This reminds me of how people would go crazy building Tableau analysis and dashboards and you end up with a gazillions dashboards and then governance becomes a nightmare.
Considering that Agent proliferation is going to be our future - WHERE IS THIS ALL HEADED?
There are 2 parallels this reminds me of -
Mobile Apps - Just like there are a gazillion apps, there will be gazillion agents. You can find 100 calculator apps and 100 flashlight apps and you can find fully functional banking apps. These are accessible via an App Store and installed on specific devices. There are 2 major App Stores that provide some level of control over the chaos of all these apps.
Web Services and APIs - There are a gazillion services that are exposed via REST/JSON that do something specific that can then be called by anything that wants to consume the information the API or web service is providing. This is a much more distributed ecosystem as there is no monopoly or duopoly of who owns these APIs / Services. Common communication protocols / standards were created to provide interoperability across this highly distributed ecosystem.
Realistically it looks like the agent world will end up more like scenario 2. Initially there will be lots of duplicate agents - user built, vendor built, agent products etc. There will be lots of very specific and narrow agents like I've been building on Agent.ai - Executive DISC Profile.
There will be full fledged agent products like Waldo.fyi and Rox.com.
That being said, the true value will be when agents start collaborating with each other to do larger and larger pieces of work. This will mean the need for agent orchestrators (we've seen this with Enterprise Service Bus products in the SOA world, ETL products in the data world), then we will need Governance products and Agent catalogs (we've seen this with Data Catalogs and Service Catalogs) and then we will need Agent Security etc. We will need to come up with some common communication language so Agents built using Hubspot can talk to Agents built using Salesforce and Microsoft. Again we've seen this with REST APIs and JSON which led to broad interoperability and loosely couple architectures.
It is funny how the technology world is so cyclical. While we have these big architectural paradigm shifts, eventually to scale things out we still need to solve for discoverability, governance, security, change management etc.