IGNITE UPDATE NOVEMBER 2023

🚀🚀🚀 In connection to the Copilot Studio Announcement at Ignite 2023 we have VERY exciting news for Timon! 🚀🚀🚀

Your friendly little "M365, Teams and beyond " Copilot, Timon, is getting an AI-boost!

First of all, you need to know about the current state in November 2023. Basically, there are two options to Ai-ify your Timon Instance:

Option 1 - Timon AI Hybrid - Using your private Azure OpenAI service in your tenant (Multiple languages available)

$200/month + $3.5/1000 responses*

• Timon with Azure OpenAI + Human knowledgebase
• 10+ language interfaces
• Multi-language responses when using AI, regardless of language interface
• Your own framing
• Data stays in your tenant
• Azure subscription and Azure OpenAI service needed for AI to run in your own tenant
• Knowledge is vast = Azure OpenAI GPT model can handle any topic. Like ChatGPT, knowledge is limited to training set of September 2021

Option 2 - Timon AI Only - (US region and English only)

$200/month*

• Timon with Microsoft Copilot Studio integrated Azure OpenAI service.
• English only
• United States is required as a region for your copilot
• Data processed in US
• Azure subscription is optional (not required)
• In default configuration: Knowledge is based on your choice of data sources, not the Azure OpenAI service. (This means that evergreen live public URL sources, internal SharePoint sites, Internal OneDrive, and more data source options are available).

*Both options will include 25 000 message capacity per tenant/month Details will follow here soon

[!TIP] Some more Info:

Option 1 - Add Azure OpenAI topic to Timon

Adding Azure OpenAI Services gives Timon the possibility to use a LLM (Large Language Model) and it's training data to provide answers in cases where normally Timon wouldn't have a answer.

The cool thing about Azure OpenAI in direct coparision to ChatGPT, for example: Azure OpenAI runs in your Subscription, the data / prompts will never leave the bounadaries of your Microsoft environment. Additionally it lets your user's indirectly use a LLM and that might bring a lot of fun! Instead of user's asking ChatGPT and data possibiliy leaving the corporates boundaries, your users can trust Timon with their query 🤖✅ All data will stay in your environment hosted on Azure.

Curious?

We got you covered -> just follow the Link: Add your Azure OpenAI Instance

Option 2 - Copilot Studio -> Conversational Boosting

Using the latest GPT Models of Azure OpenAI Services Timon can stay up to date with it's knowledge. Usually a GPT model is trained up to a specific date in time and therefore wouldn't know about anything past that date. With the Conversational Boost feature in Copilots we can circumvent this limitation by providing the links to the M365 Documentation in the Copilot Config. The Copilot will then 'crawl' the documentation and stay up-to-date with it's data.

Currently this feature is limited to your bot copilot being in English and hosted in the United States.

Ok, that's a pity for many of our global customers outside the US currently, but what else? Let's see what the PVA documentation has for us...:

Additional geo hosting and language options are expected to become available in the future (no dates disclosed).

• This capability may be subject to usage limits or capacity throttling.
• Responses generated by the generative answers capability aren't always perfect and can contain mistakes. The system is designed to query knowledge from the website of your choosing and to package relevant findings into an easily consumable response. However, it's important to keep in mind some characteristics of the AI that may lead to unexpected responses:
  • The corpus upon which the model has been trained doesn't include data created after 2021.

We have implemented mitigations to prevent the model from using its training corpus as a source for answers, however it is possible for answers to include content from websites other than the one you selected.

• The system doesn't perform an accuracy check, so if the selected data source contains inaccurate information it could be shown to your chatbot users. We have implemented mitigations to filter out irrelevant and offensive responses, and the feature is designed not to respond when offensive language is detected. These filters and mitigations are not foolproof. Note You should always test and review your bots before publishing them, and consider collecting feedback from your bot's users.

Ok, ok - noted! BUT HOW?

We got you covered -> just follow the Link: Copilot Studio