Listen to the Intuit story or read the edited story below
Writer is the full-stack generative AI platform for enterprises. We empower your entire organization to accelerate growth, increase productivity, and ensure compliance.
Writer transforms work by delivering high-quality outputs that are accurate, compliant, and on-brand. Our platform consists of Writer-built LLMs, a Knowledge Graph that connects our models to your internal data sources, AI guardrails to enforce your rules, a flexible application layer, and an ecosystem of robust APIs and integrations. Our enterprise-grade platform doesn’t use your data in model training and complies with SOC 2 Type II, HIPAA, PCI, GDPR, and Privacy Shield.
Tina O’Shea is the Director of Content Design and Design Tech at Intuit, and leads a team of 60 content designers, strategists, and architects working on QuickBooks. For nearly nine years, Tina has worked at Intuit developing and executing the QuickBooks voice in marketing, product, and self-help.
Read or listen to the story of how Tina and the Intuit team use Writer across six functions to increase employee productivity, drive internal brand consistency, and say “yes” to more.
Since speaking to us for Intuit’s first customer story, your team has made huge strides in its use of generative AI. We’re excited to chat again! Tell us a bit about your role and how it’s evolved over time.
I’ve been at Intuit now for going on nine years, so a lot has changed. When I first started, we were a content design team of about 10 and we were beginning to grow. We really needed to dedicate some of our resources to keep us all in sync. Around that time, I created a role for content systems, and we are now about 60 content designers and a systems team of around four people.
The systems folks are responsible for managing our voice resources, our style guide, and all of our guidelines, processes, and tools. Writer was one of the first tools that we brought in to help us all be in sync, especially as we grew.
The Intuit homepage
What is your philosophy on unifying Intuit’s businesses, including the ones you acquired?
Even before we were dealing with acquisitions, we had 16 different style guides all around the company, because people not finding what they’re looking for tend to spin up their own thing. Some were just Google Docs and people keeping notes on Post-its or on desktops.
We pulled everything together into a public-facing guideline. Now, we get about 25,000 views a month. That’s obviously a lot more than just my team, or even just the folks at Intuit who are looking at it. We get people from outside coming to see what it looks like to have content guidelines.
“Writer apps allows us to ensure better quality when we’re not there to handcraft that message.”
Tina O’Shea Director of Content Design and Design Tech
What led you to look for an AI solution to automate those guidelines?
We were growing bigger, and as any content team out there knows, there’s never enough of us correcting all of the mistakes that get made.
We set ourselves a big, bold goal, and that was not just to take responsibility for content quality within our own group, but outside of our own team. That meant we were going to need tools that actually put the style guide into our employees’ hands where they were writing.
Generative AI follows the same premise. There’s not enough of us to go around. We can’t be everywhere. We can’t do everything. So having generative AI helps people who don’t necessarily have a writer on their team to at least start the process of brainstorming and having a few iterations.
We are dealing with micro-content on interface screens. It’s hard to do and it takes a lot of thought. You still have to follow the style guide. People tend to write way too long on those screens, so we find that generative AI is really good for shortening things, tightening it up. It gives us a bunch of new ways to say something in a different way.
Error messages are an example of a type of content that we would love to be able to handcraft every single time, but the reality is we just can’t. Often, by the time an error message needs to be written, it is because a developer who is about to release something has realized that we didn’t catch every single possible case of what could go wrong. They need to write the error message because they are not going to be able to slow down their work to go find someone to help them out.
We’ve had content designers come up with a formula for writing error messages, how it should be structured, how long it should be, what the headline should be, the subhead, the call to action. Using Writer apps allows us to ensure better quality when we’re not there to handcraft that error message.
“We feel like we can now start to say ‘yes’ to some of those things that we were always having to turn away.”
Tina O’Shea Director of Content Design and Design Tech
Developer documents and “help” or “how to” articles are longer-format pieces that we often do not have enough resources for. There aren’t enough high-quality writers to go around the company. Since they are pretty standard format, we are able to come up with the template for what it should look like, and then have Writer help enforce that.
We use the Writer API with GitHub, which is where developers are working and publishing their documentation. The developers are not writers, and writing developer docs is their least favorite thing to do.
Having an app to help them get to the next step has been really useful. Also, because of this API, when they go to check in that documentation, there is a check to make sure that everything is on brand, on terminology, that it’s following our style, and correct any mistakes before they publish.
Using a custom app in Writer to draft error messages
Our conversational UI team has also been using generative AI and Writer to come up with utterances, or all of the multiple different ways a customer might ask a question. It used to be that they had to use their brains to come up with 100 different ways that a customer might ask the same question to train the help bot. But now, they’re able to use generative AI to iterate on the multiple ways a customer might utter a certain question.
It’s saving us a ton of time.
We use generative AI often with emails and IPD.
IPD is in-product discovery. These are spots inside the product where we can let customers know that there’s a new feature or that there’s a feature that they have that they haven’t been using that could help them out. This is the kind of content that often we’ve got marketers who are asking us for a gazillion different variations of it.
It’s really time-consuming and it’s quite repetitive, so it’s not fun work. But we do want it to be high quality, and we do want to make sure that it follows our style guide. We want it to follow our rules for readability and brand.
Using a custom app in Writer
How has your team responded to the adoption of generative AI?
We have 1,000 occupied seats. We’ve got more people who want to use the tool than we have seats for. Writers and any content designers fill up the first 100 seats, and we fill those other 900 with marketers, help writers, designers, and developers.
Within Intuit, there are a lot of different brands, for example: TurboTax, Credit Karma, QuickBooks, and MailChimp. How has Writer helped with the different nuances between these brands?
For a few years, we have been working to see what the overlaps are between the voice of each of our different brands. It’s understandable, especially when you’ve had a bunch of acquisitions.
But we are building an ecosystem. We want Intuit to be the financial platform for small businesses and consumers out there. The more we can bring those voices together, the better we sound like we are a cohesive company.
Each one has a bit of its own personality, but we have been on a journey to identify the overlaps between them and begin to lean into those a little bit harder, because the more that we can bring them together, then the better these products will work together when customers begin to see them at the same time.
We have been interested in using the Knowledge Graph feature to upload our own training materials with high quality content that we know is on-brand and have that material become the training for writers, LLM, for anything that gets generated.
“Developers are not writers, and writing docs is their least favorite thing. [Writer] helps them get to the next step and makes sure everything is on brand and following our style before they publish.”
Tina O’Shea Director of Content Design and Design Tech
What impact has generative AI had on your company?
There were a couple of different types of marketing requests that we have been in the habit of saying “no” to. These are often large campaigns that have multiple variations to them. An email campaign that has 10 variations for different kinds of audiences and that might be accompanied by in-product messaging, SMS, and multi-part marketing campaigns where the content might be very short. It might seem very small and not like a large asset, but when you want 10 variants on everything, it’s the kind of work that we had just been saying “no” to.
We really feel like we can now start to say “yes” to some of those things that we were always having to turn away.
We were really interested in figuring out how we could thread some of these pieces together into a workflow so that these multi-part campaigns would require a little bit less individual oversight and management.
For example, if we know a social element, an IPD element, an email element, and an SMS element, then we know that we can provide one brief and generate all of the assets that we need.
One goal [to expand usage of Writer] is with the Figma plugin, which will be for designers who are being asked to do more and more these days. They are being asked to become more like generalist designers, meaning that they need to be able to do it all. They’re supposed to know the interactions, the visuals, the content; they’re supposed to do it all. It’s not an easy job. This is why we’ve got specialists all over the company who do this work. But as they are being asked to do more and more of it, having the tool right in Figma for them will be useful.
“We have 1,000 occupied seats. We’ve got more people who want to use Writer than we have seats for.”
Tina O’Shea Director of Content Design and Design Tech
What can you tell us about the work you have done to incorporate inclusive language into your style guide?
We drove over 800 changes throughout the product to remove outdated terminology and replace it with terminology that’s more inclusive.
One example was “master admin.” It’s one of those terms that many developers have used for a long time and didn’t realize what its roots were and didn’t really think about how it might affect some people. Part of our job was not just coming up with the new guidelines, but going around the company and explaining to people why words matter and it’s important to get it right. Once we were able to have those conversations, then people were eager to change and got on board with making these changes.
