‍In 2025 I watched as our engineering team gained superpowers. Coding agents like Devin, Claude Code, Cursor seemed to move them to a whole new level. By end of year, 40% of the work was done by agents and productivity is up 2.5x

I was jealous. It’s been 3 years since ChatGPT was released and I was still left wondering: where were my AI superpowers?

This is a letter to people like you and me, 2026 is our year. AI Agents will finally work for the rest of us.

We’re introducing Vellum. All you do is chat and let Vellum build reliable Agents for you.

- Akash Sharma, CEO of Vellum

‍

‍

It’s Finally Our Turn

The AI hype cycle just keeps spinning. Record-breaking funding rounds, daily launch posts, every product suddenly “AI-powered” (🤢). The original promise was exciting: smarter systems, less busywork, and a fundamentally better way to work.

And yet, if you look at how we work, not much has really changed. We still have an overflowing Slack and email inbox, in fact it’s probably worse now. We still update our CRM and review customer calls manually. We carry around a mental checklist of things we meant to complete, but didn’t.

For most of us, all that “AI progress” has culminated in… ChatGPT. We feel left out in all the progress.

We’re left wondering: So when does AI start working for me?

So Close, Yet So Far

Today the models are good enough to transform how we do Operations, Finance, Sales and Marketing work.

But reality sets in. If you want your own agents, you either learn a complicated Agent Building product, stitch together one-off ChatGPT prompts, use ill-suited products for each task or ask your engineering team to build it.

But the asks get added to the engineering backlog which compete with product priorities. The off-the-shelf “AI-native” software is usually just another over-hyped AI product that didn’t quite fit the needs. So we’re stuck with my one-off ChatGPT prompts.

Our ideas stall. Not because they aren’t valuable, but because the path from idea to execution is slow, expensive, or both.

It’s ironic: people closest to the work, those of us with the most context, those spending the time doing it manually, are the least empowered to use AI to do something about it.

This is the gap Vellum is here to close, and work will never be the same again.

Introducing Vellum - your platform to build and manage agents

You just describe your goal in plain language and Vellum turns it into a working agent (think Lovable, for agents). Never mind worrying about model selection, prompt engineering, integrations, tool selection, connecting APIs and pointing LLMs to the right documentation. Vellum asks follow up questions to understand your requirements, confirms the plan and builds a working first draft of the agent. With native integrations to 100+ systems, you can read from and write to data in the places you work.

You also get a UI to understand exactly how the agent works and what it’s doing in each run. When reality changes or unexpected edge cases appear, you can adapt the agent yourself instead of waiting for someone else to fix it. Again, all you have to do is ask Vellum to make the changes.

For each agent, you can use an out-of-box UI from Vellum, build your own with Lovable, or get code snippets to integrate with your application. All your agents are stored in your workspace, so you can easily see which ones are live and which agents are live and which are in draft.

Powering it all is our open source SDK which reliably generates your task specific-agent. For the technically savvy, you can customize your agent by exporting and running the code locally with Claude Code or Devin. Unlike similar products, you can also re-upload for visual debugging (we often prefer this over reading logs).

Once you use Vellum to build your first agent, something big changes. You stop organizing your day around your task list, and start realizing how crazy it is that you’re still doing it yourself manually. These agents run in the background and free you up to work on more creative tasks.

Building agents to automate the most boring parts of your own work becomes addictive.

How teams are adopting Agents at scale

Let’s start with how we use Vellum internally

Our non-technical teams build and run their own agents to automate all sorts of tedious work. Here are some examples:

1. An agent that scans customer Slack messages and surface bugs or feature requests, runs on a daily trigger.
2. An agent to flag upcoming renewals and accounts at risk based on product usage, runs on a weekly trigger
3. An agent that prepares meeting briefs automatically every morning.
4. And, in fact, this post was mostly written by an agent made in Vellum.

And it’s not just us! I’m seeing the best teams in tech are building their own custom agents to scale themselves.

Brex, Ramp, Gusto all have open job descriptions for “AI Automation Engineers.” Vercel recently did a webinar on how they deployed AI Agents throughout the sales function.

The common thread here is there’s no one master Agent to automate it all, but rather, dozens of purpose-built task-specific Agents tailored to each individual business process created by those closest to the work.

They might not all be flashy, but together, they remove hours of friction every week. More importantly, they eliminate that constant feeling that important work is slipping through the cracks.

Build Your First Agent

Think about one task that quietly drains your energy every week.

Not the most important one. Just the most annoying.

Build an agent for that. Let it run for a few days. Notice what it feels like when the work simply happens without reminders, tabs, or mental overhead.

There's something deeply satisfying about describing a tedious task and watching it disappear for good. Build one, and you'll immediately want to build ten more.

What are you waiting for?

Vellum has a generous free tier with free credits to build and run your own agents. No credit card required.

Sign up, connect to your systems, and build your first agent in minutes. Start small, watch it work, and try not to get addicted.

👉 Get started at vellum.ai