AI Agents Take Center Stage: Betaworks’ Bold Investment in ‘Camp’ Program

Betaworks is embracing the current AI trend in a unique way. Instead of creating just another LLM (long short-term memory)software, they are cultivating a clutch of agent-type models to automate everyday tasks that may be difficult to define.

The investor’s latest incubator, called “Camp”, has trained and funded nine startups that specialize in creating AI agents. These agents have the potential to tackle mundane and monotonous tasks, making our lives easier.

As promising as these companies may sound, AI still has a long way to go in terms of fulfilling its promises. Would you trust a shiny new AI to sort your email for you? What about extracting information and organizing it from a webpage? And would you allow an AI to manage your meeting schedule?

There’s an important element of trust that has yet to be established with these AI services. This is a common occurrence with most technological advancements that change the way we carry out our daily tasks. Remember when asking MapQuest for directions felt strange? But now, GPS navigation has become an essential tool in our everyday lives. However, are AI agents at that stage yet? According to John Borthwick, CEO and founder of Betaworks, they are. (Disclosure: Former TechCrunch editor and Disrupt host Jordan Crook left TC to work at the firm.)

“You’re keying into something that we’ve spent a lot of time thinking about,” he told TechCrunch. “While agentic AI is in its nascence – and there are issues at hand around success rates of agents – we’re seeing tremendous strides even since Camp started.”

Although AI technology will continue to improve, Borthwick explains that some customers are already embracing it in its current state.

“Historically, we’ve seen customers take a leap of faith, even with higher-stakes tasks, if a product was ‘good enough’. The original Bill.com, despite doing interesting things with OCR and email scraping, didn’t always get it right, and users still trusted it with thousands of dollars worth of transactions because it made a terrible task less terrible. And over time, through highly communicative interface design, the feedback loops from those customers created an even better, more reliable product.”

Borthwick continues to explain that although most early users of products in Camp are developers, founders, and early adopters in the tech industry, this group is willing to patiently test and provide feedback on these AI products. Eventually, these products will become mainstream.

Betaworks Camp is a three-month accelerator program where selected startups in a specific theme receive hands-on guidance with product development, strategy, and connections. At the end of the program, the startups can present their work and innovations in front of potential investors on demo day, May 7.

Let’s take a look at the three startups that caught our attention.

1. Twin

   Twin aims to automate tasks using an “Action Model”, similar to what other startups like Rabbit have been doing. This model is trained on a large dataset representing software interfaces. It claims to be able to learn how to complete common tasks that are too complex for an API to handle, yet not too complex to be delegated to a “smart intern.” This could potentially allow you to delegate tasks such as organizing resumes, Dropbox folders, and Slack notifications. Their goal is to automate the 20% of tasks that consume 80% of our time, but the real question is can they do so in an affordable way? (Twin declined to provide further information on the nature of their model and training process.)

2. Skej

   Have you ever been in a situation when finding a suitable meeting time for everyone involved is a tedious task? Skej aims to solve this problem by simply being cc’d in an email or Slack conversation. It will then reconcile everyone’s schedule and preferences to find a suitable time. If given access to your schedule, it can check that too. It also considers factors like preferred days and time slots, priority attendees, and more. As someone who dislikes human interaction, I appreciate this solution. It’s well within the capabilities of today’s AI agents, which primarily focus on understanding natural language.

3. Jsonify

   Jsonify takes website scraping to the next level by using improved understanding of visual AI models to better parse and sort data. It can extract information from unstructured contexts, which has been done before, but the engine doing the extraction has never been very smart. This can be frustrating when trying to extract information from visual lists and tabs meant for humans to click around. Jsonify aims to make this process easier by creating a structured list with columns for price, distance from airports, ratings, hidden fees, and more. This could potentially save you time and money, especially for tasks such as searching for vacation rentals. But can LLMs be precise enough for this job? According to founder Ananth Manivannan, they have built a robust guardrail and cross-checking system. They use various models at runtime for understanding the page, providing validation and improving extraction accuracy.

Resolvd AI – automates cloud workflows.

Floode – an AI inbox organizer that reads and responds to emails.

Extensible AI – helps companies test and troubleshoot their AI models.

Opponent – a virtual character for children to interact and play with.

High Dimensional Research – offers an infrastructure for web-based AI agents.

Mbodi – provides generative AI for robotics, specifically for tasks with limited training data.

There’s no doubt that AI agents will play a major role in the automation of software workflows in the near future. However, the extent of their role is yet to be determined. Betaworks seems to be getting in on the ground floor by funding and supporting these startups, even though some of the products may not be ready for the mainstream market just yet.

If you’re interested in seeing these AI agents in action, mark your calendar for May 7 and check out Betaworks Camp Demo Day.