How Microsoft Copilot Can Save Your Property Maintenance from Costly Misdiagnoses
Discover the Power of Digital Assistants in Streamlining your Repair operations
⚡ Tl;dr
- AI-powered solutions like Copilot, organisations can reduce repair misdiagnoses, streamline work order processes, and improve customer satisfaction.
- Guidance on developing and integrating Copilot, this article highlights its potential to empower customer service teams, ensuring repairs are handled accurately and efficiently from the first point of contact.
🛠️ The Crucial Role of Customer Service in Repair Management
In nearly every organisation, you will find a hive of activity centred around a Customer Service operation. Agents tirelessly juggle calls, emails, and chat messages, expected to handle every customer interaction seamlessly.
From my observations, almost every Housing Association, Local Authority and Repairs Contractors, there is a Customer Service team responsible for raising work orders for property repairs. This process often requires a special code, known as a Schedule of Rates (SOR) or a Repairs Ordering Schedule (ROS), typically well-versed by Surveyors or Contractors.
One common challenge organisations face is the misdiagnosis of repairs at the initial point of contact, leading to the wrong contractor or operative being dispatched. This can result in frustrations for both customers and contractors, such as not having the right tools, materials, or skills to address the issue.
I wanted to understand how a Copilot could assist agents in diagnosing an issue and assigning the appropriate SOR/ROS code. This blog explores the creation of a Copilot — a digital assistant designed to streamline the repair code process, automate work orders, and enhance the efficiency and accuracy of initial diagnoses. The Copilot aims to ensure repairs are handled correctly from the first point of contact.
💻 Choosing Microsoft Copilot for Development
Microsoft CEO Satya Nadella recently highlighted the transformative potential of Copilot Studio, a platform allowing organisations to tailor Copilot for Microsoft 365 or create custom copilots. This innovation has already been embraced by over 10,000 organisations, since its release back in March 2023 (MarketBeat).
Microsoft leads in low-code/no-code development, empowering users to create apps, automate workflows, analyse data, and now, build custom copilots. Over 230,000 organisations have utilised AI capabilities in Power Platform, marking an 80% increase quarter-over-quarter (MarketBeat).
New data shows that the productivity gains are real. Through a combination of surveys and experiments, we learned that Copilot significantly boosts productivity, creativity, and time savings. Remarkably, 77% of early users reported they didn’t want to go back to working without it (Microsoft).
👷♂️ Building the Copilot
To get started and create your first Copilot with Copilot Studio, you need:
- A license for each user, also known as a per user license (or Copilot Studio User License as referred to on the Microsoft 365 admin center), should be assigned to individual users who need access to create and manage copilots.
- A license for your organisation, also known as a tenant license (or Copilot Studio as referred to on the Microsoft 365 admin center), should be acquired by the tenant administrator. This license cannot be assigned to individual users.
Step 1
When you create Copilots with Copilot Studio, you author and edit topics. Topics are specific conversation paths in a Copilot that allow the user and Copilot conversation to feel natural and flow appropriately. The first topic I set up was “Repair Issue Code”.
Step 2
To equip the Copilot with the repair codes and descriptions, I uploaded a PDF containing the codes and descriptions to Copilot Studio. I could have also connected to Dataverse, where the codes were stored in an entity, ensuring the Copilot had access to up-to-date and comprehensive information
Note: Currently, generative answers can only process files up to 3 MB in size. Larger files can be stored in SharePoint and are returned by a Graph API search, but aren’t processed by generative answers. If files exceed the 3 MB limitation, you might want to explore alternative architectures, such as connect your data to Azure OpenAI for Generative Answers.
Step 3
Set the ‘Create generate answers’ step to only use the selected knowledge source and turn off the ‘Allow AI to use its own general knowledge (preview)’. An important part of the process is to set a prompt, as shown in the screenshot below.
Step 4
Training the Copilot: By feeding it past repair issues and their corresponding codes, I was able to validate the responses against actual codes used on work orders.
Important! Train Your Model: Actively train your Copilot by upvoting good responses and downvoting bad ones. This feedback loop will help improve its accuracy over time.
Step 5
Testing took up the largest amount of time whilst working on this proof of concept. I did get some unexpected responses from time to time but this was often due to the prompt and not upvoting good responses. So with your own projects, it’s important to test to ensure that your model was refined as necessary.
🔍 Does the Copilot only provide codes?
You can ask the Copilot to provide examples of a repair issue, and it will offer both the description and the corresponding code.
You can also give the Copilot a repair code, and it will provide a detailed description of the associated repair task. This feature helps ensure clarity and understanding, allowing customer service agents to verify codes and ensure the correct actions are taken.
Even vague descriptions will work; its natural language processing capabilities can interpret and understand incomplete or imprecise inputs, ensuring that the appropriate repair codes are still generated accurately.
📲 Using the Copilot
Users can access the Copilot directly from within Microsoft Teams, without needing to switch between different applications. This convenience ensures that they can quickly respond to repair issues as they arise.
You can also configure Microsoft Teams chat in Customer Service so users never need to leave the Dynamics 365.
⚙️ Copilot to automate work order creation
During this development, I focused on using Microsoft Copilot Studio. My aim was to develop a Copilot capable of generating accurate repair codes based on input from Customer Service agents. With Power Automate, the Copilot’s capabilities would expand to include automating the entire work order process. Here’s how it could work:
- When the Copilot generates a repair code based on the agent’s input, this would trigger a predefined flow in Power Automate.
- The flow would automatically create a new work order in the system, eliminating the need for manual entry by the agents.
- The work order would be populated with all necessary details, including: The repair description provided by the customer. The SOR/ROS code generated by the Copilot.
- The system would then assign the work order to the correct contractor or tradesperson based on the provided repair code.
- End to end automation could be achieved through providing the available appointment days and times to the customer so they could book and reschedule their own appointments. Alternatively, the work order could be picked up internally and then scheduled on the customers behalf.
This integration not only reduces the workload on Customer Service agents but also increases the likelihood of repairs being correctly diagnosed and resolved on the first visit, significantly enhancing overall customer satisfaction.
🌟 The Vision
With the implementation of this Copilot, Customer Service agents are empowered to handle repair requests with the expertise of a seasoned Surveyor or Contractor, transforming the efficiency and accuracy of property maintenance processes.
This innovation marks a significant step forward in how Housing Associations, Local Authorities and contractors manage their future repair operations, ultimately leading to better outcomes for customers.
Imagine a world where repair work orders are created automatically, freeing up more time for agents to focus on customer experience and being able to spend more time on complex cases.
📚 References
For more information on creating and optimising your Copilot, refer to the following resources:
Feel free to reach out in the comments if you have any questions or need further assistance with your Copilot project. Happy building!
