Boring automation is underrated
I’ve spent years working on automation, and one thing I’ve learned is that people often assume the best automations tackle the biggest, most sophisticated problems. I’ve come to believe, though, that some of the most valuable automations are actually the boring ones. The everyday tasks that nobody thinks of as particularly high-impact, but quietly eat up hours of people’s time.
At Ravenna, we’ve become obsessed with these seemingly mundane, repetitive tasks, especially the ones that keep teams from doing bigger things.
The hidden work behind a simple request
Nowhere is this more apparent than in IT. For IT teams, the frustrating part of employee support often isn’t doing the work. It’s all the detective work required before they can do it.
Take something as simple as an employee requesting access to software. Someone in IT has to figure out:
Who the employee reports to
Whether they’re eligible for a license
What level of access they need
Who needs to approve it
How to track this request for compliance
What if they need admin access? Now Security might need to get involved, additional approvals are required, and someone needs to make sure that elevated access gets revoked when it’s no longer needed.
Granting the access may only take seconds, but working through the steps to determine whether, when, and how to grant it can take hours, if not days.
And the next time someone makes a similar request? Much of that work starts all over again. Multiply that across hundreds of requests, and you have IT teams spending their days chasing approvals, checking policies, and moving tickets along.
Automate the detective work, not just the task.
In my conversations with IT leaders, no matter the size of the company, I hear a similar frustration. They have all these big dreams and ambitions for their organizations. The best IT leaders view their departments as being uniquely able to help the business move at speeds it never could before and deliver incredible employee experiences.
But they’re so caught up in the day-to-day work of keeping things running that the bigger improvements they want to make keep getting pushed aside.
The irony is that the people responsible for making everyone else more productive barely have time to improve their own systems.
We’ve modernized how people work, but not how they get support
For all the advances in workplace technology, the way employees get help hasn’t fundamentally changed in years. Traditional ITSM and help desk systems were built around portals, service catalogs, forms, and tickets. While these remain important, the experience feels increasingly out of step with how modern organizations work.
For employees, getting help is unnecessarily complicated: They spend their days in Slack or Teams, but when they need something, they’re sent to a separate portal, asked to find the right form, and required to provide information their organization already knows. And once they’ve submitted a request, they’re often left wondering who’s handling it, what happens next, and when they’ll get what they need.
For IT teams, delivering that help is unnecessarily manual: They spend hours untangling a web of complexities across systems, policies, and people. Requests stall while waiting on approvals, handoffs, and decisions. With dozens of things in flight, keeping track of what’s happening becomes a job in itself, and things inevitably fall through the cracks.
The rules of automation have changed
For a long time, fixing this meant building more workflows, adding more rules, and trying to anticipate every possible scenario. A few years ago, when I was leading AI engineering at Zapier, we started experimenting with how AI could change automation. What if you could simply describe what you wanted to automate, and AI could help build it for you?
We spent a lot of time trying to make single-shot code generation work, but the models weren’t reliable enough to make it practical. Today, off-the-shelf coding models can reliably generate API calls when given the right documentation.
The unlock: LLMs have become far better at understanding natural language, reasoning through complex requests, and taking action across connected applications. They can maintain context across multiple steps and navigate workflows that previously required a lot of manual coordination.
That’s fundamentally changed how we think about automation.
Building an automation for Ramp corporate card requests
Imagine wanting to automate corporate credit card requests. Traditionally, you might be stuck if your platform didn’t have a prebuilt integration with a provider like Ramp. Now, you can simply describe what you need in plain English and have Ravenna build the integration, forms, approval routing, and rules to make it all work.
Building integrations on the fly
Agents have also gotten much better at investigating requests and taking action across systems. For example, if an employee loses their laptop and needs it locked, Ravenna’s agent can gather the necessary context, identify the right device, check relevant policies, and take action through the company’s device management tools.
All of this happens right in Slack - where employees can simply describe what they need in natural language, without navigating portals or figuring out who to contact.
Of course, IT teams should remain in control of which actions can be fully automated and which require human approval. In this new world, that means giving teams control at the individual action level, rather than forcing them to decide whether to automate an entire ticket type.
Requiring human approval for specific actions
Deterministic vs. non-deterministic (AI) automation
Despite all the AI excitement, we’ve always believed practical automation needs a healthy mix of deterministic rules and non-deterministic (AI-driven) workflows.
AI is incredibly useful for making automations easier to build and adapt, understanding intent, gathering information, and delivering a conversational experience.
But not every action should be left to an LLM. Take an engineer who’s on call for the weekend and needs temporarily elevated access. You don’t want AI making a mistake about who gets access, what permissions they receive, and when that access expires. That’s where predictability matters, and deterministic rules make a lot more sense.
We believe IT teams should be able to describe what they need in plain English and have AI help build the automation, while retaining control over how it runs. IT teams need the flexibility to use deterministic rules when predictability and control matter, agentic capabilities when reasoning is needed, or a combination of both within the same workflow.
IT teams understand their business, policies, and priorities better than anyone. Our job is to help them turn that knowledge into automation that’s fast to build, flexible enough to evolve, and reliable enough to trust.
The opportunity is much bigger than IT
One reason Kevin Coleman and I got particularly excited about employee support is the sheer scale of the opportunity.
A new hire needs a laptop. A salesperson needs a corporate card. Someone needs to register a guest visiting the office, get an NDA reviewed, or request access to a new application. The frustrating part has never been any one step. It’s that relatively simple requests can take hours or days to resolve, and the next employee might go through exactly the same process.
AI creates the opportunity to provide employees with reliable, around-the-clock support, regardless of where they’re working, while allowing service teams to scale their impact without scaling headcount at the same rate.
Every minute an employee spends waiting for access, chasing an approval, or figuring out where to get help is time taken away from doing their actual job.
Multiply even small improvements across hundreds or thousands of employees, week after week, and the impact becomes enormous.
I’d love to hear from IT, People, and Ops leaders: what’s one workflow you wish you could automate but haven’t been able to? If you’re curious about how Ravenna could help, let’s talk!
