Prioritize risks and identify workarounds

Actively manage risks throughout the project lifespan:

• Identify risks early: During the planning phase and at each sprint's start, brainstorm what could go wrong or block progress. For example:

  • "Integration with a service might be delayed,"
  • "We're unsure if the natural language model can distinguish these two intents,"
  • "Legal approval is needed for the HR answers."

  List these risks.

• Assess impact and likelihood: Note which risks would have the biggest impact. A risk of "Application Insights might show personal data" could be high impact (compliance issue). Likelihood might vary. Gauge and then prioritize addressing the high-likelihood, high-impact risks.

• Plan mitigations: For each major risk, assign a mitigation or contingency plan. For example:

  • Risk: Non-Microsoft API for order tracking might not be ready by testing phase.
  • Mitigation: Build a stub or mock service to simulate responses so development and testing can proceed. Integrate the real API when available. Alternatively, prepare a fallback ("Sorry, order info is currently unavailable, please contact support") in case the live integration fails, to handle it gracefully rather than just show an error to the user.

• Track blockers daily: In agile practice, discuss blockers in stand-up meetings. If something is stopping progress (for example, waiting for credentials, a customer connector isn't available yet, unclear requirements), escalate or resolve it as soon as possible.

Workaround strategies for blockers

Sometimes, despite all preparation, you encounter a blocker that you can't immediately resolve. For example, an external dependency or unsolved technical issue might block progress. Instead of halting the project, modern management finds creative workarounds:

• Reduce scope temporarily: If a feature is too complex or blocked, deliver a simplified version.
• Manual process as bridge: Automate what you can, and introduce a manual step where you must. If an AI model can't classify a rare request type, route those cases to a human agent or send an alert to a subject matter expert while you retrain the model. If an automatic approval flow isn't ready, have someone handle those approvals manually. Document this workaround and mark it in the backlog so the team remembers to replace it later. The goal is to avoid project delays by having an interim way to achieve the outcome.
• Parallel spikes and research: Dedicate a time-boxed "spike" to investigate a risky area, like evaluating a new API or testing a complex regular expression for entity extraction. This approach doesn't directly produce user functionality, but it reduces risk of the unknown. If the spike results are good, you proceed. If it turns out not viable, at least you know early and can pivot to an alternative approach without having promised that feature. This strategy is essentially a workaround for uncertainty.
• Communicate and align: Keep stakeholders informed about any significant blockers and your mitigation plan. Transparency prevents panic.

Proactive risk management and prioritization mean you're always working on what matters most first.

Assumptions and concerns

Capturing assumptions and concerns early helps the team identify potential risks. Together, these elements form the foundation of a successful Copilot Studio project. They ensure clarity from the start and a shared vision of what the agent is meant to achieve.

Capture any assumptions or concerns early to inform design and risk management:

• Example assumptions:

  • The solution supports out-of-the-box deployment across all required channels.
  • Generative AI reduces the need for manual knowledge management.
  • The solution provides state-of-the-art intent recognition and entity extraction.

• Example concerns:

  • Limited control over LLM-generated answers.
  • Potential internal resistance to automation.
  • IT preferences for on-premises hosting or specific compliance needs.

Next step

Build your cross-functional project team with the roles needed to deliver and operate the agent successfully.

Build a cross-functional team