Share via

Conversational AI project best practices

  • Article

Adoption of low-code technologies like Copilot Studio can help customers envision and deliver solutions with less management and fewer maintenance costs.

Engagement principles

Customers approaching a Copilot Studio project should align project delivery with their own company culture and objectives. However, there are some core principles that are adopted again and again among successful Copilot Studio projects.

Principles

An icon with gears to represent a flywheel. An icon of a box with a checklist inside of it. An icon of a pencil writing on a checklist.
Engage your flywheel Keep it simple Keep goals in focus
When the Copilot Studio agent goes live, the "customer feedback flywheel" takes effect quickly. Actionable customer feedback comes in almost instantly. Delaying the publishing of your agent for more topics or processes means foregoing valuable user feedback. Encourage the team to see going live as the beginning of your project, not the finale. One of the core value propositions of Copilot Studio is iteratively delivering a conversational AI experience quickly. After the project goes live, maintainability is key to staying agile. Tooling can allow users to create complex conversation flows. However, extensibility with the Bot Framework enables flexible interactions, therefore decide if complexity is always necessary. It’s easy when implementing new technology, like an agent, for developers to focus on the minor details of implementation. Define your key business goals, for example to deflect 30% of customer queries or score above 4.0 customer satisfaction. Goals can keep the project moving and ensure correct feature prioritization.

Best practices

  1. Focus on delivering immediate value

    Building agents is an iterative initiative. You learn from your users what they're looking for. Use these observations to drive further investment.

  2. Establish role clarity

    Make it clear who's the decision maker, so there's no ambiguity in delivery.

  3. Identify risks early and plan contingency accordingly

    Occasionally customers run into technical issues when implementing specific requirements or integrating with other systems. These issues can take time to diagnose.

  4. Agree on issue priorities

    Define a blocker, a must have, and a nice to have. Decide how aggressively you want to push for workarounds, while the underlying issues are addressed. Workarounds are a natural part of projects as they allow the project to make progress, while a longer-term solution is found.

  5. Determine your Application Lifecycle Management (ALM) strategy

    Move to the production-level process as early as possible. Don’t linger in development phases, when there are timelines to be met.

Note

For more information about analyzing usage, see Combine web and agent analytics.