One of my favorite parts of my role is working closely with innovative customers like Chris Oh, Senior Director of AI Enablement at Headspace. Chris and I recently caught up to talk through how Headspace uses Tines 3B to give teams the freedom to build with AI, without the operational risk.
We covered the problem Headspace set out to solve, why they chose Tines 3B, and some of their early wins with the product.
The problem: wild code and patchwork automation processes
As Chris explained during the webinar, automation has always been central to Headspace's engineering work. But as the organization's user and employee bases grew, so did its stack of independently owned, siloed automations.
As Headspace's leadership pushed for broader AI adoption across marketing, legal, clinical ops, and beyond, what Chris calls "patchwork automation processes" became a real problem.
"People tended to try a bunch of third-party tools on their own to see what they can do, without consulting IT or InfoSec. That naturally led to concerns about data privacy and security."
That's a textbook example of what we at Tines call "wild code" — apps, agents, and automations built with AI outside the oversight of IT and security teams.
Chris didn't want to shut down AI innovation altogether, but he flagged the ownership risk across vendor relationships as significant.
"Understanding who owns what is important,” he says. “With this push on automation across the entire company, we felt we needed a set of tools we're able to centrally govern and manage, while still letting people really get creative and own the automation."
The solution: Tines 3B
Chris's interest in Tines 3B began while exploring Tines Stories: "I started just talking with it, seeing what it could do to actually build automation through natural language."
That curiosity led him to a 3B pilot, where his real test was building something "without any technical knowledge... from planning all the way through to a deployment."
Tines 3B promised what leadership at virtually every organization is asking for right now: an answer to the questions of who's building what, how much it costs the business, and what systems and data are being accessed.
What Headspace has built so far
LLM spend dashboard, built in under an hour
At first, Chris's goal was to build weekly automated reports on LLM spend. "But after interacting with 3B, I thought it might be a good time to build a dashboard that people could check on their own time, instead of just spamming everyone with more emails."
From start to finish, the dashboard took less than an hour to build. So how'd he do it? You can watch the webinar for a full look under the hood, but here's a recap.
"It's based on an automation that takes off every 4 hours. It pulls data in from the cloud, analytics APIs, and Cursor APIs to various LLM providers, as well as some of our internal tools to reconcile with the actual people, do some calculations and trending work, then present it in a nice, pretty format."
Chris was "pretty impressed at how quickly it went from ideation all the way to completion." He kept tinkering, expanding his build to include a subscription option for viewers — scratching that initial itch to send out weekly reports or similar notifications.
He also appreciated the business and leadership context behind 3B's suggestions for workflow upgrades throughout the building process.
It created an automation that wasn't just useful, but also thinking of ways it can run better and use less resources.
Chris Oh, Senior Director, AI Enablement, Headspace
These include 3B's Autofix and Autotune features, which help organizations ensure their teams' builds are running as expected and in the most efficient way possible.
Closing a 6-figure content localization gap by automating translation
Headspace's marketing and content ops teams were quick to see value from their builds with 3B, too. "We have around 10,000 meditation courses," Chris says. "We're great at creating content in English, but expanding language coverage meant a huge amount of work. You have to make subtitles, convert it into the new language, transpose it directly into the content, and reupload it to the library. The challenge was that the burden was on a single person, and it was manual work."
Translation still requires a degree of human input, review, and polish, especially for a company like Headspace: "Our type of content is not typical content," Chris explained. "There's a lot of stuff in there… that we either don't want translated at all, or mistranslated."
Headspace's workflow for automating translation with Tines 3B, Contentful, and LLMs:
Pull all Headspace courses across domains
Check each course to confirm if it has corresponding translations
Identify which translations are missing
Create a list of transcripts to create and translate
Use Claude to extract the English transcripts/captions and translate them into the target language(s)
Send Claude's translation(s) to ChatGPT for a quality assurance check
Generate the video caption (VTT) file and reupload it to the CMS entry in Contentful
Route the result through human review and approval before publishing live
Chris noted the workflow can also run this process automatically for future content, or be triggered manually, including a one-click option to convert Headspace's entire remaining backlog at once.
Rethinking ROI
I appreciated how honest Chris was when it came down to measuring the impact of Tines 3B: "I can't forecast how much it's gonna save per automation. The fact is, most of the automations are being done to offset human time. And this is where it gets tricky: You're saving time spent, but the company is not saving the money itself because everyone's still on payroll."
For us, it's easy to justify ROI, because a lot of people that have been doing the work manually can be redeployed without any challenges.
Chris Oh, Senior Director, AI Enablement, Headspace
For Chris and his team, investment isn't solely financial — "informal" benchmarks matter just as much as dollars saved.
"When I do my ROI calculation, it's always going to be a combination of cost, plus human capital savings, as far as the amount of work that went into supporting the workflows and engaging with other teams. Also, the empowerment aspect of it will naturally create less load on the engineering teams."
That load gets especially heavy for Headspace in Q4 of each year, ahead of the company's peak season, when Chris and his team get hit with "a huge rush of work, right when people want to take holidays and vacations."
Naturally, the topics of burnout and attrition came up.
"My goal for rolling out AI stuff… is not to accomplish any other goal, really, but to help people recover some time and work-life balance,” he says. “I think that led to us reducing our attrition rate significantly across engineering, and across the company. For me, that's a very convenient way to prove the value."
For Headspace, the benefits of 3B adoption aren't just adding up — they're multiplying:
Hundreds of hours saved and reallocated
Frictionless redeployment of team members on more impactful work
Optimized LLM spend
Reduced attrition rates and increased productivity
These "intangible" benefits are all well and good, but Chris is thinking about return on investment, too: "I'm already confident that I've hit my goal, which is an ROI of 2x to 4x what this cost. That's what I want to achieve in at least the first year, and then we'll reassess as we go."
Governance without slowing people down
Between HIPAA laws and GDPR requirements, technical and governance teams are in close contact throughout the compliance process — which means the automations and connections themselves are core protected assets that require careful management.
This added an extra layer of challenge for Chris while implementing launch-gating and review functions: "Everything ties into our compliance team. We added checks and balances to ensure people don't have access to the secrets on the data stores there. We're also cautious about what kind of gate gets done, in case of duplication or data access problems."
Reviewing automations before launch, and triaging them before anything went live, required the technical and InfoSec teams at Headspace to work together when choosing the right solution for the job.
"Automation tools are very high-risk, just because of what they're able to do. Under the hood is a repository of connectors, basically a set of API keys. Luckily, there's a lot of security and governance around who can access them and create things with them. But they also allow you to essentially manage a lot of the secrets. There's a scariness about the complexity involved. For users, they don't have to worry about needing those credentials or that level of access."
I was glad Chris mentioned this, especially the complexity for non-technical users and the iceberg of security vulnerabilities it opens up.
We designed Tines 3B with security built in from the ground up. There are no credentials in the scripts at all, precisely because the people building with AI tools don't have decades of experience in credential security. That won't stop them from building, though, which is why LLMs don't have access to credentials in 3B either.
Everything is managed securely, down to flexible role-based access controls governing who can access which connectors and where they can build.
Where this is headed: from experimentation to enterprise-wide adoption
As a product manager, I've loved watching Headspace's usage of Tines 3B extend past Chris's initial builds, even in the short time since this webinar was released.
Here's how Chris and the team at Headspace sum up where things stand today:
Tines 3B is helping Headspace move beyond AI experimentation and into real adoption — enabling teams to build solutions that solve meaningful business problems, and deliver real outcomes.
Teams across Accounting and Finance, RCM, Marketing, Revenue Operations, Sales Operations, IT, and Information Security can build their own apps, agents, and automations, while Tines 3B gives us the visibility and control needed to govern those workflows at scale.
We appreciate the flexibility to use established tools like Claude Code and Codex for building, while Tines 3B provides the operational and governance layers necessary to ensure safety, compliance, and scalability.
We see 3B changing how teams work, moving from isolated automations to more collaborative, enterprise-wide workflows. That balance of flexibility and control is what makes 3B so valuable.
Chris Oh, Senior Director, AI Enablement, Headspace
As AI enablement at Headspace continues to grow, the breadth of functions now building is proof that adoption has moved well past a single pilot use case. Enabling this new generation of builders to keep using their preferred tools and LLMs — with Tines 3B providing the governance layer — keeps everything running safely.
This security-first yet flexible approach represents a larger shift: from isolated, single-owner automations to collaborative, enterprise-wide workflows. Which brings us full circle back to Chris's initial problem, and the core mission behind Tines 3B: taming wild code.
Early lessons for teams starting out
These are some of my favorite takeaways from Chris, in his own words:
Let people try, even if it doesn't work — "What's different about 3B is it's very good for people who tend to be intimidated by building with AI. It won't misinterpret you. It asks you exactly what you want it to do, and then you just press the button you want.”
Target the work nobody wants to do — "I don't want to do data entry or validation for eight hours a day. For my team, I'd love to say, 'Hey, let's reduce this so you can pursue more fun, enriching stuff.'"
Centralize before you scale — "Let's say you need a simple content translation automation. Third-party agentic automation providers will quote you, like, $25K to build it, then $5K a month to run it. Using a centralized tool lets you manage, maintain, and refine everything in one place, versus using 100 automations across 80 different platforms that are overly complex for what you need."
While it's still early days for Chris and the team at Headspace, one thing is already clear: governed AI adoption works. Headcount continues to grow despite all those time savings and reallocations — the clearest signal yet of 3B's potential to eliminate muckwork and enable more transparent, impactful AI usage across the enterprise.