Making AI Easy to Use

Nov 12, 2024

Fitting AI into Everyday Work

For AI to be truly helpful, it needs to be easy for everyone to use in their daily tasks. Right now, using AI can be tricky:

  • Too Many Tools: AI helpers often feel separate from the apps we use every day.
  • Hard to Learn: Getting started with AI, even with tools like bot builders, can take a lot of training.

These issues make it hard to see the real benefits of AI. Our solution is to build AI directly into our platform, so it works smoothly across all the services you use.

Think about building a helpful chatbot:

  • We want users to easily pick pre-made setups (presets) for common tasks.
  • Custom settings made in the builder should automatically work with the AI.
  • Adding information should be simple (like uploading a conversation script to help the AI learn).

While these ideas fit our current bot builder, we know it can be complex. Our goal is to make these new AI features easy to understand, even if you're new to this.

People like using tools like ChatGPT because it's clear what goes in and what comes out, and you can adjust it easily. Just having powerful AI isn't enough if people don't understand how to use it or feel limited by the interface.

Making AI feel "intuitive" is more than just good design. We need a clear plan that considers how people actually work. When we introduce new AI features, we need to think about the old tools or habits people might need to change.

Our 3 Steps to Better AI Experience

We're improving the user experience (UX) for AI in three main stages. These steps often build on each other, but we can be flexible:

  1. Make AI Simple: Our tools are better because using AI is straightforward.
  2. Make AI Practical: Our tools are better because our AI is easy to adjust and use daily.
  3. Make AI Powerful: Our tools are better because they help you get more done with AI.

Why User Experience is Key for AI

A good user experience makes it easy to go from what you want to do to using the system. When you use AI, there are usually three parts to your request:

  • Command: What you want the AI to do.
  • Constraints: Any rules or limits you need to set.
  • Context: The specific situation or background information.

For example, telling an AI assistant:

You are an AI assistant for a bank.

Interview a sales candidate.

Focus on their motivation, understanding of the job, and if they're okay with cold sales.

Breaking it down:

  • Command: Interview a sales candidate for a bank.
  • Constraints: Act like AI, not a human.
  • Context: Focus on motivation, job understanding, and cold sales.

AI helps by handling some of this structuring for you.

How We're Improving the Experience

Let's look at making AI better at each stage:

1. Make AI Simple

The first step is making AI easy to find and use right where you work. ChatGPT did this well by offering a simple chat window instead of just code. If AI is easier to access in our tools than in others, that's a big win. This could mean simpler ways to choose AI models or having an AI helper guide you when building bots.

2. Make AI Practical

Once AI is easy to access, we need to make it practical for everyday tasks. Common problems people face are:

  • Forgetting the exact words needed to get the same good result again.
  • Getting results that are almost right, but with small errors (like formatting).
  • Spending too much time checking the AI's work, making it feel less helpful.

Fixing these issues makes AI much more practical. When AI gives consistent results quickly and with less hassle, it's easier to rely on. If our tools make it easier to get results than hiring help or asking support, people will value that.

3. Make AI Powerful

This is the hardest but most rewarding step. We want to design the experience so you naturally use AI to get more value from our platform. This often means more complex background work, but the goal is to make our tools feel like the best way to achieve your goals.

Where We're Focusing AI First

To bring the most value quickly, we're prioritizing adding AI to the parts of our platform people use most. Based on feedback and usage, here's our focus:

  1. Dialogues: Helping support teams by analyzing past chats, suggesting replies, summarizing conversations, and checking customer happiness.
  2. Analytics: Providing deeper insights by predicting trends, creating instant reports, and suggesting improvements based on data.
  3. Messages: Making marketing and outreach more effective by writing personalized messages, targeting the right groups, and sending at the best times.
  4. CRM: Improving customer relationship management by predicting task success, suggesting personalized actions for sales teams, and recommending the best ways to contact leads.
  5. Bot Builder: While important, we'll enhance the bot builder after improving the areas above. Why?
    • The builder is often used by more technical people, and we need to simplify its basic design first.
    • Improving areas like Messages or CRM directly impacts key business goals like customer engagement and sales.
    • While the builder already uses some AI, making it simpler in other areas first will provide broader value sooner.

Making the Bot Builder Better with AI

  • Simple: Put AI tools right inside the builder, so you don't have to switch apps.
  • Practical: Let the AI automatically understand the context. For example, select parts of your bot's logic, and the AI automatically uses that information in your prompt. You could then just type, "Create a hiring chatbot..." without copy-pasting. A handy feature would be a "Refactor" option that sends a specific part's details straight to the AI.
  • Powerful: Show helpful AI suggestions directly in the interface, hinting at what's possible. The key is then giving clear options to apply those suggestions based on the full picture, making the AI work for your specific needs.

Making Sure the AI Works Well

Beyond being easy to use, the AI needs to produce high-quality results. We need solid ways to:

  • Check Quality: Regularly test the AI's output.
  • Monitor Performance: Keep an eye on how the AI is doing over time.
  • Fix Issues: Improve the AI's behavior through adjustments (like fine-tuning or better instructions).

Just fixing problems isn't enough. Doing all three like checking, monitoring, and fixing creates a cycle that leads to truly great AI.

AI Hiring Bot

Our AI hiring bot helps HR managers quickly evaluate candidates. It's valuable because it keeps everything in one place. Initially, it worked well, but as we added more complex rules, sometimes the quality dropped:

  • Changes in one area sometimes broke another.
  • Improving the AI's instructions became harder.
  • Instructions got too long and complicated trying to cover every rare situation.

Monitoring and Fixing

Good monitoring and ways to fix issues are essential. We need reliable systems for testing and watching the AI. Different kinds of tests have different costs, which affects how often we run them (e.g., quick checks for small changes, bigger tests for major updates).

While improving AI often involves gathering lots of data, having automatic quality checks helps everyone continuously make the AI better.

Conclusion

To make AI a helpful part of everyday work, we need to:

  1. Make it simple to access and understand.
  2. Make it practical by reducing manual effort.
  3. Make it powerful by giving users control and flexibility.
  4. Ensure high-quality results through careful testing and monitoring.