VectorShift

During YC, people laughed at VectorShift. But here’s an unpopular belief: if you want to build a successful startup, work on DUMB ideas. VectorShift is a no-code platform to help anyone build AI workflows. Essentially, a “horizontal” solution to build custom AI tools. However, in our batch, “vertical” AI solutions were all the rage. Our batchmates (and a lot of VCs) told us: “Enterprises are already using vertical AI agents to replace entire teams” “Every unicorn SaaS company will have a vertical AI unicorn equivalent” In fact, YC just made a video titled: Vertical AI Agents Could be 10x bigger than SaaS. However, Harj, our YC group partner defended our approach in the video (clip below): “Enterprises are a little unsure about exactly what agents they need and one approach I've seen is letting Enterprises spin up custom agents. It's something I've seen from one of my companies called VectorShift that I funded about a year ago.” We couldn’t agree more (and thanks for the shoutout Harj!) When I look back at recent startups that have succeeded, I’m reminded of the story of Scale AI. When Scale was at YC S16, their batch mates laughed at their product entitled “Human API.” People said that human-labeled training data wasn’t a “technical” solution. That it was just “Filipinos as a service.” However, those people didn’t realize that the buyers of training data don’t care that your solution is not “technical.” They just want the data. Today, Scale is worth $14B and growing. Similarly, what people don’t understand is that most human workflows in a company are custom and buyers of LLM software just want them automated. Custom workflows need custom solutions. We think that most vertical SaaS solutions will be replaced by custom AI workflows that you can build on your own, because it is faster, cheaper, and more performant.

A YC ed-tech company used to manually input data from 15k+ student forms 20+ full time staff used to spend >50% of their day doing this We built a workflow in <2 hours that automated the entire process: - Staff now uploads the forms to VectorShift - GPT4 mini and advanced OCR extracts student information (even from messy handwriting) - VectorShift adds data directly into their ERP system of record

Enterprise AI solutions are 99% showmanship – cool internal demos without real ROI. That is the cold hard truth. <10% of enterprises have GenAI solutions in production. Most enterprises have: 1. Appointed a head of AI 2. Identified use cases across business units 3. Built proof of concepts So why are enterprises stuck before moving to production? - Uncertainty around AI investments: Companies are concerned about the implications of investing in AI solutions that may not deliver the expected results (or worse, result in negative reputational risks). - Too many choices: Firms are flooded with options and are not sure what to choose. Should I use an out of the box solution? Should I build the solution in-house? Should I outsource to a vendor? - Complex decision making structures: I have seen many enterprises form multiple committees that complicate decision-making processes, leading to delays and indecision. We built VectorShift to help enterprises move from proof of concept to production with confidence. Feel free to reach out - happy to share best practices: [email protected]

Last year, we got into YC and raised $3m. But when we launched VectorShift, we only got 3 signups. No one cared. People often think that raising money and getting into a prestigious accelerator are the hallmarks of success as a founder. It is far from it. There are millions of tactical steps in between. This past year: I did sales I did customer support I did marketing I did interviewing I did product QA I did client deployments I did legal and compliance I did bookkeeping And more Many people think being a founder is about raising money, setting the high-level strategy, and giving the team instructions from an ivory tower. The reality is doing whatever it takes for the customer to have a better experience and to solve their pain points. 10+ launches later (and many more failed ones), we now have thousands of paying customers and tens of thousands of users. There are still millions of more steps left, but we are excited to continue on this journey with our customers.

Since YC, we receive 10k+ job applications every week I used to spend 5 hours every Friday night reviewing We automated the process with VectorShift Took <30 minutes to build - I upload thousands of resumes to a form - Claude 3.5 Sonnet analyzes each one against requirements - Adds analysis to a Google Sheet

When I was 13, I coded an e-commerce site that made >$5m per year. The moment I published the site, the app started to take off. That’s when I saw the impact technology can have. A high school intern with basic coding skills could 10x a business’s revenue overnight. A business that was taking phone orders immediately had customers from all over the world. Within days, we received our first customer from Germany, and soon we had customers from 100+ countries. I remember sitting in disbelief and dreaming of creating technologies that could impact a billion people. A decade later, we were accepted in Y Combinator and we founded VectorShift (YC S23) - we built a no-code platform to help anyone build AI workflows. My vision for VectorShift is to democratize AI technology. We hope all our users can experience the magical moment I felt when I was 13. I hope anyone can use VectorShift to create an app that touches the world. If you are looking to leverage AI in your business, my favorite part of the day is the customer support - we only exist because of our customers and I am here to serve you every day.

A YC company was quoted $68k for a custom AI invoice reader Their intern built it in 5 minutes With VectorShift, employees now submit invoices through a form and GPT4o mini: - Extracts vendor name + amount from the invoices - Categorizes them - Adds data to a Google Sheet

At McKinsey, I spent 2-3 hours every day extracting data from reports Took <1 hr to build a financial data extraction pipeline for a bank using VectorShift GPT4o instantly: - Extracts and formats key financial figures - Identifies trends in market data, risks, and mangement commentary - Compares performance against historicals and comparable companies

I have talked to 1,000+ companies this year. <10% of large enterprises have GenAI solutions in production. However, almost every enterprise has identified use cases. There is a big chasm where genAI ideas die when moving from PPT slides to production solutions. These are the top 3 causes: 1. Data is scattered in an enterprise. To make AI solutions useful, siloed data has to be unified and optimized for AI workflows. 2. Enterprises lack the skills and expertise across both developers and front-line employees to build and use AI solutions effectively. 3. AI solutions need to be rigourously evaluated and monitored for performance to limit hallucinations. We built VectorShift to help enterprises cross this chasm. For example, an F500 tech company uses us to deploy 3+ solutions across their sales and engineering org. They use VectorShift to: 1. Centralize and live-sync with their data across all their tools. 2. Optimize the data for AI workflows (e.g., tabular data, data in images, improve retrieval accuracy, etc). 3. Deploy solutions to end users in days instead of months. We have now helped thousands of customers on their AI journeys. Feel free to reach out - we're happy to share more about AI use cases, moving AI solutions from pilot to production, and best practices: [email protected]

Since YC, I've had 5-10 product demos every day I used to spend 3 hours the night before to prepare We automated the process with VectorShift in <30 minutes to: - Research the client: background, size, and person I’m speaking to - Pull their intent: platform usage, errors, and stripe history - Score the lead: identify key challenges, regulatory headwinds, and potential solutions