Automating workflows with Pinkfish is as simple as prompting ChatGPT — but with enterprise predictability at its core
I've spent the last 15 years building AI products — everything from customer service AI to AI for major banks like Barclays to AI for gaming. Nothing has excited me more than what's ahead. The possibilities are massive.
Yet enterprises still struggle with automation. My co-founder CK and I have seen it firsthand — companies hit roadblocks despite investing in tools like RPA and integration platforms. Employees still spend thousands of hours on repetitive work. They hate it. Their bosses hate it. CIOs want to eliminate technical debt, reduce manual labor, and move faster.
The issue isn't just manual work — it's that existing solutions are slow, expensive, and underutilized. Deployments take months, requiring engineers and ongoing maintenance. As a result, frontline workers remain stuck, IT is overwhelmed, and automation never reaches its full potential.
It's time to change that.
By now, we’ve all experienced the magic of generative AI. We give ChatGPT a prompt to write a blog post, rephrase an email, or analyze a CSV and voila, we have an answer! It’s truly delightful to imagine an outcome and have it delivered in seconds.
We imagined a future where you could do the same thing, but for automations. So if you’re that employee spending hours on repetitive work, or their manager trying to make your team more efficient, you could bring automations to life as easily as writing a blog post with ChatGPT.
We imagined Pinkfish — a Generative Automation platform — where you could go from zero to sixty in record time.
Asking for an automation and getting it — that’s pretty darn special. And it works. So if you want to auto-triage IT support tickets based on urgency, or process invoices dropped into a folder, or automatically fire off a series of Jira tickets when new customers sign up, everything is a prompt away.
There’s nothing like this in the previous generation of automation software. It doesn’t matter how low-code or click-friendly a platform may be. Software created over the past ten years can’t match Generative Automation in terms of instant gratification and maximal flexibility. It is, in fact, the low-code movement that hampered the previous generation of software.
In an effort to make building automations easy, these providers built UI node after UI node. If you’ve ever tried to create something complex with UI nodes, you know that it quickly becomes spaghetti. And you know that it takes sitting through hours of “product university” classes to understand the ins and outs of each UI node option.
Generative Automation isn’t like this. It’s fluid. You describe what you want and the automation agent helps you build it. You ask questions and the automation agent responds. You work together to build an automation to your spec.
You don’t have to figure out how to use an interface. You don’t have to understand how each node works or what nodes are available. You just build. And when you’re done, you have a working bespoke automation that can run a million-plus times.
Generative Automation is unlike Generative AI in one significant way. It’s deterministic. ChatGPT may give you a different answer every time you ask. But enterprises need predictability and reliability in their automations. So Generative AI needs to put predictability at its core.
And this is just how Pinkfish works. Pinkfish works with you to create an automation and generate the underlying code. When you’re confident in the outcome, you lock in a version, and that version consists of immutable code. So it’s going to run the same way the first time as the millionth time.
And, of course, it’s built for teams. So you can build an automation and share with your team, or you can build for a client and turn it over to them to manage. Managing and governing automations over time is a massive benefit that Generative Automation delivers.
Generative Automation is also self-healing. So when an automation fails due to a change in its environment (say, an API changes the output it delivers), Pinkfish detects the error and suggests a solution that you can test and accept. The fix-patch is generated using the same method that the automation was generated with. It’s part of the same continuum of Generative Automation.
The promise of AI agents is their ability to handle complex tasks with minimal instruction. The dream is to tell an agent roughly what you want done, and for the agent to have the wherewithal to get the task done to your standards. Pinkfish can generate an AI agent as easily as it can generate an invoice processing pipeline.
Pinkfish also solves the most vexing problem with AI agent design, which is tooling. Every AI agent needs a set of tools to do its work. Building these tools is as challenging as building automations in the pre-Generative Automation era. In fact, many AI agent platforms suggest bringing in your developers to create the tools that your agents will need. But in Pinkfish you generate your tools. A tool is simply an automation that your agent can use.
Enterprises don’t always need AI agents. Often they’ll need a simple deterministic solution. But for workflows that require an agentic approach, where the agent makes its own decisions about how to execute a workflow, Pinkfish is a great fit. It’s got everything you need to get an AI agent off the ground quickly.
Ready to build your first automation with Pinkfish? We created a video to get you up and running with Generative Automation. Watch it here.
Ben Rigby
Co-founder & CPTO, Pinkfish
Expanding possibilities for people at work
