TL;DR: watch the demo video
A couple of years back, most AI bets were about creating armies of robots taking over manual labour. It turns out that knowledge work is where the real disruption was. The rise of LLMs like ChatGPT, Claude and Llama has changed everything and every product team should be exploring what it means for them.
Here’s what happened to us.
Reconsidering what’s possible
First, let’s set some context: our team is building a goal-tracking platform for teams and most of our customers use it to manage their OKRs and connect them back to their projects.
So, a lot of our initial effort went into shipping the best set of goal-tracking features we could have:
- A custom editor to write goals with a doc-like experience
- Dashboards designed specifically to make progress and confidence easy to monitor
- Custom views built for alignment
- Workflows, integrations, API, etc…
Basically, we built a complete SaaS platform that is great to use once you know what your goals are.
But, we quickly realised that many teams were looking for guidance to help them write their goals. They’d come to Tability hoping that the platform could turn their ideas into a set of effective OKRs. And that makes sense! Adopting OKRs can be challenging and many orgs find it difficult to switch from an output-centric focus to an outcome-driven culture.
We started by writing guides and creating templates to help, as it’s pretty challenging to provide effective goal-setting support within a tool. Too many of the issues are context specific and every customer is working on a different problem and strategy.
We needed something more, and found the perfect answer with OpenAI and their API.
Experimenting with ChatGPT
If there’s one takeaway from this post, it is that all product teams should make it mandatory to experiment internally with LLMs. It’s going to be really hard otherwise to understand what’s possible and allow creative solutions to flourish.
For months, we were spending hours every week trying to think about all the content that we could create to help our users understand how to write great goals. This was a classic pre-AI thought process where you’re trying to scale knowledge solutions through humans. But the big issue with this approach is that you either have to niche hard to offer relevant examples (ex: how to set goals for design teams in B2B SEO SaaS) or you have to offer generic advice that is helpful to a bigger crowd but lacks domain-specific examples.
But, once you start using LLMs, it becomes clear that you now have a brand new tool that is able to offer a beautifully tailored experience to every single user. It’s like having a colleague next to you that can answer any question no matter what function, industry, or role you’re in. It might hallucinate at times, yes, but for the most part, it’s an incredible asset that can drastically improve things for the end user.
We talk about UX issues to reference problems related to the usage of a tool.
I believe that we’ll soon introduce the concept of User Knowledge issues (UK) to reference problems related to getting the right data into a tool – regardless of how easy it is to use said tool.
Now let’s talk about how we integrated LLMs capabilities into our platform.
Integration AI into our existing features
Here’s the simple process that we used:
- Step 1: map the user journey – and make sure to highlight all the friction points
- Step 2: talk to users to get qualitative insights about the friction points (don’t just rely on your intuition)
- Step 3: brainstorm AI-specific ways to solve these issues
- Step 4: try solutions in the LLMs/AI tools
- Step 5: build it!
This flow looks obvious, but it’s crucial to start with the user journey if you want to get interesting ideas in step 3 – otherwise it can quickly turn into a set of “AI summaries for X” mimicking what you saw in other platforms.
In our case, interviews with OKR champions highlighted that teams struggled to find the right way to write their goals, even when they were really bought on the idea of using a framework like OKRs. So the first thing we did was to go to ChatGPT and see if it could turn ideas into measurable goals, or if it could rewrite poorly drafted OKRs for us.
And it could.
🤯
So we then went back to our user journey map and looked at the corresponding screens. Most of the changes would happen in our editor, and the great thing about it is that we already made it pluggable.
And this is another point that I want to make: this isn’t about putting a cosmetic UI layer on top of ChatGPT. This is about taking a SaaS platform that does many great things by itself, and figuring out the right way to augment the user experience with AI.
You absolutely need to invest in building the right product first.
Fast forward to today, and our new AI features allow people to:
- Get feedback and suggestions to improve their OKRs – and implement changes in 1-click
- Auto-detect metrics to set the right targets
- Chat with our AI to find the best way to turn their ideas into a measurable strategy – including action items
Check it out in the video at the top of the article, and you can also try for free by signing up at https://tability.app/signup.