Eoin Hinchy: Eliminating “muck work” through workflow automation
Tines CEO explains how AI will free us from repetitive tasks
What prompted you to create Tines?
In 2017, I was running cybersecurity at DocuSign. I had an amazing team, but everyone was spending more and more of their time doing the same repetitive tasks over and over — running down phishing emails, responding to security alerts, making sure compliance was applied to infrastructure. This undifferentiated work was leading to burnout, human error, and unsatisfied and disengaged team members. We started to look around for tools that we could buy that would allow the teams to automate these repetitive parts of their job without having to write code.
Over the course of six months, we surveyed the market and evaluated about a dozen different products. We hated everything we looked at. They were too hard to use, couldn't handle the complexity and mission-criticality of my team's workflows, or they were designed for very siloed use cases. Out of frustration, I said, ‘You know what? I think I can go build this product myself.’ In 2018, I founded Tines to scratch that itch and build the product I wished I had as a practitioner.
How does AI change the game? Where do you see it adding value for you and for your customers?
In late 2022, when large language models (LLMs) began to emerge, we realized that this was going to be a tectonic change for how all software operated. But it took us nearly a year to figure out how to include AI in the product in a way that actually added value to customers and wasn't just some marketing checkbox. We started by asking our customers why they weren't using AI. Our customers consistently surfaced two problems. First, LLMs have no access to real-time, proprietary information that makes businesses run. Businesses could move all that data from scattered, disparate systems into a single location and then fine-tune it, but this requires a lot of skills and resources that companies often don't have access to or aren't willing to invest in.
The second problem is that companies don’t trust these models. People can't predict why they perform in certain ways, so they don’t want them taking action on their behalf. We realized that deterministic workflows — those whose outcomes are predictable and repeatable — are the perfect answer to both these problems. We’ve spent seven years building connectors to all these tools so our customers have access to real-time and proprietary data regardless of where it exists. The big realization for us was, if we combine what we do really well with what LLMs do really well — understanding large amounts of data, providing next steps, and summarizing — then you have a very novel approach to scaling the automation of undifferentiated work.