Making sure that relevant information flows throughout an organization and that knowledge isn’t kept siloed within individual teams has been one of the biggest challenges for product organizations. Miro and Spotify are just two of the many examples of companies to have successfully implemented an architectural approach to avoid pigeonholing.
Their choice of creating cross-functional teams that are able to see a feature’s creation from start to finish, in a true end-to-end approach, comes from the practical realization that removing barriers is actually more beneficial to the efficiency, effectiveness, and improvement of a product.
Perhaps more recently, the missed opportunity from setting up processes and structures that allow this exchange of information has come to the attention of conversational AI professionals, too. Finding strategies to educate others about AI in general, and ways to make conversation design in particular more known and open towards other teams is a first, crucial step in that direction.
In this article, we explore the idea of taking an architectural approach to solving the issue of knowledge diffusion in CAI-forward organizations, considering:
- How Miro’s AMPED cross-functional team structure makes it a true “product org” and Spotify’s Agile approach
- The calibrated perspective of Esha Metiary, a renowned Senior Conversation Designer, on the best organizational structure for a CAI-reliant organization
- The experience of a conversational AI professional working in a telco that has embraced cross-functional team arrangements
Taking inspiration from product orgs’s architectures: Miro and Spotify
If you follow our LinkedIn posts, you’ll know that we love taking inspiration for our conversational AI-centered discussions from adjacent fields, like UI, UX and, yes, also Product. Especially when it comes to themes concerning end-to-end-oriented organizational architecture and cross-functional teams, product orgs can offer a lot of inspiration to more recent and less structured fields, like conversational AI.
An incredibly successful example of a product organization that has fully embraced this approach is Miro. In a conversation on Lenny’s Podcast, their CPO, Varun Parmar discusses their “architecture” as the definition of what they mean by “product org”, pointing to the importance of their organizational structure in guaranteeing that barriers are fully removed and product culture is fully embraced.
Their architectural approach to ensure a diverse and cross-functional representation is referred to as AMPED (Analytics, Marketing, Product, Engineering, and Design). Meaning: each of their teams is composed of 1+ person for each of these functions. This setup enables a variety of perspectives to contribute, balancing aspects like end-user experience and enterprise requirements, facilitating effective course corrections and fostering collaboration across different streams.
As described in this article by Ashley-Christian Hardy, Spotify also employs a distinctive approach based on the Agile framework, by organizing its teams into small, self-contained units, each responsible for a specific functional aspect, such as Search or Recommended artists. These squads, resembling Scrum teams, operate with a dedicated product owner who supplies user stories for development.
The squads, co-located and unified under a long-term mission, possess a comprehensive skill set and tools to independently handle end-to-end processes, encompassing design, development, testing, and production release. This setup fosters autonomy and expertise, akin to mini start-ups.
Furthermore, to encourage inter-squad interaction, Spotify integrates shared lounges as well as regular informal team gatherings, where squads share ongoing projects. Overseeing this dynamic ecosystem is the Tribe Leader, tasked with cultivating a conducive environment for all squads to thrive. Spotify employees have suggested that the way the company organizes itself has evolved since 2011, when the Spotify model was set in place, but the basic framework still appears to be the one described.
These two successful organizations are testimonies of how a cross-functional approach to team structuring can lead not only to a better organization, but also to effective and efficient release of high-level outputs.
Could this approach benefit chatbot and voicebot teams and the other departments of organizations that have implemented conversational AI?
Let’s hear the perspective and experiences of Esha Metiary, Senior Conversation Designer, and of a CAI professional working in a prestigious telco with cross functional teams.
Esha Metiary, Senior Conversation Designer, on how CAI-reliant organizations can (or should) structure their teams
Esha Metiary: “As long as you can work from bottom up, there is nothing wrong with traditionally structured companies, with a CEO at the vertex, management teams in a middle layer, and the operational teams below. This can be a successful approach if every team’s lead regularly interacts with the other team leads, discusses what each of their teams is going through and, then, one of them reports to someone at the level above. This approach ensures that the higher levels are informed about what decisions are made below them.
However, from a content output perspective, I think it's much more useful to have a multifaceted team, where, e.g., there is:
- a conversation designer;
- a developer;
- a web editor;
- a content manager;
- someone from customer support.
Having all these disciplines together, you’d be able to create the best output, because the lines would be much shorter, you could iterate much faster, and there’d be a lot less trouble in getting information from others.”
— — — — — — — — — — — — — — — — — — — — — — — —
The experience of a CAI professional working in a prestigious telco with cross functional teams
How is your team structured?
"We work with the Agile methodology, so we have a cross-functional squad dedicated to our chatbot, composed of:
- 2 Developers
- 2 Testers
- 1 Conversation designer
- 1 Scrum master
- 1 Product owner
Each member of the conversational AI Squad has their own Chapter. Meaning: developers from the conversational AI Squad and all other Squads create a Chapter, me and all other Product Managers form another Chapter, and so on.
Within each Chapter there will be a form of hierarchy, so that, e.g., the developers working within my Squad report to managers from their Chapter, that are higher up in their own technical hierarchy. Every Chapter will have their own leads, ways of reporting, and organize their own meetings to keep everyone within it aligned."
What are the benefits of this approach?
"An example of the way this structure helps us move fast and get better results is that we have two testers within the squad. This means these two people know everything about the chatting experience, the journeys we’ve implemented, the business we’re in and are able to spot every kind of detail and possible ways to improve. Without this knowledge and contextual understanding, you will not be able to get as good results. "
How do you split your time and resources between different tasks?
"For the two developers, I’ll sometimes split them between the cognitive part of the journey, i.e. working on the utterances, and the button part of the journey. On the other hand, for the testers, I tend not to tell them what to do, or how to divide the work.
We only have one conversation designer and they mostly spend their time on new intents, so there’s not a lot of retrospective work to be done."
Does this organizational structure favor interactions with other teams?
"This is actually a problem we've had before, as we used to be totally disengaged from the marketing team. The relationship we had with them was mediated by the communications team, who would let us know what kind of scripts, new offers, or new products the marketing team would have liked to publish on the chatbot. The problem with this approach is that you might also not be updated, or not hear the feedback from this unit at all, as they didn’t know a lot about the chatbot.
Since I joined, I have started to create this relationship with the marketing team and invited them to engage with our chatbot regularly, to give us feedback. Having this channel of communication with them has meant that they will now let us know what needs to be added, removed, changed in the responses or intents that the chatbot tackles, and that we know about their initiatives."
How do you keep rapport with other teams within the organization?
"With the marketing team specifically, we have monthly meetings. Now, they have awareness of the full flow, so, when they are testing and put themselves in the shoes of the customer, they know what kind of replies they should be receiving. During the meeting, we can discuss any discrepancies they notice.
We also keep close ties with the technical support team, as they can help us if there’s any emergent issue arising or that we see occurring from the data, and they communicate with the provider of the NLU system."
And if you’re still hungry for knowledge, follow us on LinkedIn for weekly updates on the world of conversational AI, or check out our one-on-one interview with Esha Metiary, where she shares all her tips and strategies around CAI and conversation design evangelism.