How to build a full-stack Conversational AI team

In the full-stack concept, a single person performs all the tasks in an end-to-end delivery model. This isn’t a new idea. For example, full-stack engineers were all the rage about 15 years ago, about the time when companies began to replace websites with web-based applications featuring tightly integrated front- and back-ends.

The allure of full stack is powerful. If one person manages end-to-end delivery, accountability is clear, speed of communication is fast, and people costs are low. Such benefits are particularly attractive for startups looking to move fast, or large companies looking to build innovation teams to exploit new technologies quickly. But as teams, technologies, and companies mature, the shortcomings of full stack may become a concern.

Stacks eventually become complex. Full-stack talent, be they engineers or conversation designers, struggle to keep up with complex technology. If the stack in question is new or its components closely coordinated, a single person can be an expert in all of it. But as technology evolves (as it always does), it's very difficult for any one person to be an expert in the full stack.

You can’t stack tasks.

The need to be able to stack tasks in a workflow (as opposed to working on them in sequence) is often overlooked. For example, an efficient conversational AI delivery team should be able to plan Use Case D while building Use Case C, and testing Use Case B while optimizing Use Case A. Such task-stacking is impossible if one person is responsible for all the tasks.

Full stack talent is by definition special and hard to find. This also makes it very brittle: lose one or two people from a team, and work grinds to a halt. A team organized around full stack talent is also very difficult to flex: it's much easier to outsource a narrowly-defined specialist role than a full-stack role.

To be clear: there are lots of good reasons to embrace the full-stack concept. This is why so many conversation designers invested time and resources in mastering end-to-end conversational AI delivery, and why so many companies have come to rely on them. However, as a company, you need to assess whether the costs of relying on full-stack conversation designers outweigh the benefits.

The good news is that an alternative concept already exists: the full-stack team. This is an end-to-end delivery team consisting of specialists in planning, design, building, and optimization AI assistants.

They understand language, psychology, ánd technology. Everyone in the team has their strengths and they can lean into those. This modular approach turns all the shortcomings of the full-stack conversation designer on its head:

• Each of your specialists is given the opportunity to do one or two things extraordinarily well, and extraordinarily fast. This is great for output and morale.

• Complexity is less of an issue, allowing your team to embrace new technology faster and with less risk.

• Tasks can be stacked and coordinated across timeframes or even deliverables without reducing quality or burning out team members.

• Your team as a whole is more resilient and can be easily flexed up or down to meet demand.

Contrary to conventional wisdom, a full-stack team of conversational AI specialists at different seniority levels isn’t necessarily more expensive than a team of full-stack conversation designers. Nor does having more people in the mix mean a loss in efficiency or quality. The trick is to have thoughtful processes based on best practices that allow team members to be more productive and more happy.