Looking to take your AI workflows from one-off skills to a comprehensive system? Snag your spot for this event we're hosting with our friends from Button to learn how!
🗓️ You’re invited: Join us for a (free) special event with Ditto!
Date: July 28, 2026
Time: 11:00 am–12:00 pm CDT
Join Jolena Ma for a session on how content designers can build reliable content systems to power better product copy at every step of the workflow.
Register now: https://lnkd.in/eGEtE69F
🎥 This event will be recorded and shared with registered attendees within 24 hours after the session ends.
Looking for content design resources? Curious about upcoming events? Be sure to subscribe to our new LinkedIn newsletter. ✨
Inside this edition:
• Button 2026: Introducing Intensive Day (Sep 23)
• Free watch party: Naming product features for a global audience with Urmi Rahman
• Free special event: Building content systems for AI workflows with Ditto
• New on the Button blog: Drawing diagrams to persuade and align by Bridget Lawrow
• Smart perspectives from the community
Have a resource you’d like to see featured next time? Send us a message or email us at hello@buttonevents.com.
Notice anything different? 👀
Our team has been shipping so fast, we needed a website upgrade to properly capture all the incredible things we're bringing to your product copy workflows.
Here are a few of our favorite new additions...
Go check out the shiny new dittowords.com to see everything that's new.
We're thrilled to announce Ditto as one of our Lead with Tempo sponsors!
Ditto have strived to improve content designers' experiences so it's a match made in heaven for our community...we look forward to partnering with them over the next couple of months and bringing some great content your way. Watch this space!
We assigned three engineers to Ditto's generative features, full time.
But not to build them out.
To keep them honest — and make sure they actually work.
Building the feature is no longer the hard part.
The hard part is keeping outputs reliable as customer edge cases emerge, product context changes and grows, and new workflows create new ways for quality to drift.
Here's what they've done in the last month alone:
- Built an eval harness that tests every prompt and model change against real cases, and gated it in CI so nothing ships until it passes.
- Kept growing the test set from real, reported customer failures rather than hypotheticals.
- Automated a quality report that finds patterns in rejected style guide suggestions and turns them into prompt fixes.
- Added an inline way for users to reject a suggestion and say why, and started reviewing those reasons regularly.
- Caught our own rejection metric inflating itself: refreshes were silently logging unseen suggestions as rejected.
- Shipped a model upgrade that only paid off because the prompt and test work came first.
- Fixed a bug where our AI localization was pulling in rules meant for another language.
- Widened the context generation can draw on, including design-system context it couldn't see before.
- Started capturing the exact inputs behind every generation, so a regression traces back to its cause.
- Throttled how often edit suggestions refresh, so a burst of edits can't blow past the model's rate limits.
- Built dashboards and alerts for token spend and rejection rate by workspace.
Not listed: dozens of experiments that we thought would improve output quality, but turned out to just add more noise and degrade quality.
With AI features, you can't just own the initial build. You have to own its output over time. If you're shipping something non-deterministic and just assuming it works… it probably doesn’t.
(And yes, this applies to all of the internal-facing agents, .md files, and tools your team is shipping too.)
Didn't snag a ticket for Config 2026? Or was your schedule just a bit too packed to see everything you wanted?
Catch Jessica Ouyang and Jolena Ma's 20-minute talk on Config's Maker Stage, now up on Youtube.
Jess and Jo break down how teams can build an AI content system to generate and ship better copy, right alongside the tools your team is already working. Because the way we build and ship products has changed. Copy workflows need to change with it.
https://lnkd.in/gd_bDHTF
Still buzzing from Config last week!
It felt really special for Jessica and me to get to represent content on the Config stage.
To us, words feel like they matter more than ever right now.
With this talk, we wanted to share what's actually working -- practical solutions we've heard from 100+ product teams figuring out how to keep their content accurate, on-brand, and useful as more people and agents are writing it.
If you missed it live and want to watch, here’s the recording! https://lnkd.in/gVt3kEVD
(And thank you to everyone who messaged us afterwards or came to our booth to chat! Everyone seems to just be figuring things out as they go, and it feels good to be talking about it together.)
Giving a talk *and* running a booth at Config means that you organically get exposed to a LOT of conversations. Enough to lose your voice by the end of the week 😅
After a weekend catching up on some much-needed sleep, here’s what stuck with me:
1. Systems were king.
Whether it was design systems, design ops, or content ops, the underlying mechanics were top of mind for everyone.
Systematization seemed to be the answer teams kept arriving at independently. And the question had shifted from "why?" to "how?"
> How do we maintain quality at scale?
> How does my team keep governance when work is non-deterministic?
> How can I audit what’s actually being shipped—without lowering the bar for craft, and without leaving the precious details up to an agent to decide?
Systems went from “nice to have” to P0 overnight.
2. Design teams (esp content designers) are actively reaching beyond what’s theirs.
Every conversation I had with a content designer expanded fast. To localization. To design systems. To engineering.
Cross-functional collaboration has always been good practice. Now it's closer to survival, especially as everyone moves up a layer. A designer who doesn't understand what it takes to ship the product (even as AI lowers that barrier) risks not getting their work shipped at all.
Design managers I talked to understood silos to be a vestigial organ of the traditional process. To them, the areas of overlap had only gotten more obvious as teams worked more closely together.
Ultimately, I think this is a good thing. Product text has never existed in a silo anyway.
3. Figma is no longer *the* tool. It's one of them.
Yes, this was Figma’s flagship event. And yes, teams are still using Figma as a core place for design work. But most design leaders I talked to described the canvas as one surface among many where work happened: agents like Claude Code or Codex, design system tooling, developer tooling, the codebase itself.
Design definitely isn’t dead. It's just not always happening in the canvas.
Everyone I talked to thought about these surfaces as constantly interdependent parts. New tooling won't get evaluated in 2026 on how well it works in isolation. It has to move the existing agentic workflow forward.
4. The online discourse is easily more alarmist than the conversations in the room.
Being at an event like this in-person—talking to people from 7AM through dinner—makes everyone real. And in real life, there’s plenty of optimism, curiosity for new ideas, and problems to solve (rather than just post about). People will say "honestly, I don't know, we're still figuring it out" across a table in a way they never will in my feed.
Reality has an incredible amount of detail. A lot of it gets flattened online.
This was our third year at Config, and it's become a kind of annual checkpoint for how much both Ditto and the wider industry have changed.
What stood out to you this year?
Another stellar year at Config in the books. 💪
Thank you to everyone who came for a chat, stayed for a demo, and left with some swag. Events like these remind us exactly why we work so hard – for you, our users.
Already looking forward to the next one! (But first, a few really big sleeps)
