Principles

Build for the creators

Voice AI tools should build with the end users – the creators – in mind. Keeping individuals productive is more important than generating perfect reports or metrics.

Opinionated software

Voice AI needs to be opinionated. It's the only way the product can truly do the heavy lifting. Flexible systems let everyone invent their own workflows, which eventually creates chaos as implementations scale.

Create momentum – don't sprint

Find a cadence and routine of working with voice technologies. In cycles, you can decide priorities and assign responsibilities. The goal is to maintain a healthy momentum with your teams, not to rush towards the end.

Meaningful direction

Even if your daily schedule is filled with smaller tasks, it's important to understand and remind everyone of the purpose and long-term goals of your voice AI work. Milestones, projects, and initiatives are all essential to consider when you plan your weekly schedules.

Aim for clarity

Don't invent terms if possible, as these can confuse and have different meanings in different teams. Use clear, consistent language in your voice interfaces.

Say no to busy work

Your voice tools should not make you the designer and maintainer of them. A tool should work for you, not the other way around. Remove or automate "work around work," so you can focus on what really matters.

Simple first, then powerful

Teams have different needs depending on their size. A voice AI tool should be simple to get started with and grow more powerful as you scale.

Decide and move on

There isn't always a best answer. Sometimes the most important thing is to make a decision, and move on.

Practices

Set strategic product initiatives

Ambitious goals are the only way to make a significant impact with voice AI. Companies should focus on them when they define their high-level direction. Reserve some space on your product timelines for unplanned work that comes up unexpectedly and allow your product plans to change if needed.

Connect daily work to larger goals with projects

All voice AI projects and work should directly correlate to these goals. Review projects and their target dates during initiative meetings and pull from projects as you plan cycles.

Work in n-week cycles

Cycles create a healthy routine and focus teams on what needs to happen next. 2-week cycles are common in voice AI development. They're short enough to not lose sight of other priorities, but long enough to build significant features.

Keep a manageable backlog

You don't need to save every feature request or piece of feedback indefinitely. Important ones will resurface, low priority ones will never get fixed. A focused backlog makes it easier to plan cycles for voice AI development.

Mix feature and quality work

All voice systems have issues, more than we can ever fix. Include bugs and other fixes as part of your cycles. Invest in tooling as it is a force multiplier if done right.

Specify project and issue owners

Every voice AI project should have a named owner who's responsible for writing the project brief and delivery. Even if other people are involved, the responsibility should lie with a single person.

Write project specs

Aim for brevity. Short specs are more likely to be read. The purpose of a spec is to briefly communicate the "why," "what," and "how" of the voice AI project to the rest of the team.

Understand your users

Collect user feedback and use it as a research library when developing new voice features. Try to spot trends. Reading feedback, even complaints, is an opportunity to get to know your users and build exactly what they actually need.

Scope issues to be as small as possible

Break down voice AI work into smaller parts and create an issue for each one when possible. Ideally you can complete several concrete tasks each week. It feels great to mark issues as done.

Measure progress with actual work

The clearest way to see whether something is complete or not is to show the actual changes in the voice system. When the tasks are scoped small, your changes will be easier to test and review.

Run cross-functional teams

Voice designers and engineers should work together on projects, creating a natural push and pull. The best voice AI creators often have a talent for both design and implementation.

Write a changelog

Writing a changelog benefits both internal and external communication. Internally, it helps your team track progress on voice features. Externally, it keeps users informed about what's new in your voice AI system.