I organized practical IT documentation, developed a daily shared-drive directory catalog, and moved from Cody AI experimentation to a working Grok bot that can communicate through Microsoft Teams and use an existing knowledge base as a reference.
Support knowledge was spread across documents, file shares, and employee procedures. People could have difficulty locating a current guide or the right file. Some reference materials were not being interpreted correctly by an AI retrieval tool. A useful assistant needed both accessible source information and a communication channel employees already used.
I organized and refined the knowledge base, converted documents when parsing failed, wrote SOPs that were later reviewed and implemented, and built a script to map the network shared drive into a regularly refreshed directory reference. I first experimented with Cody AI conversations through Microsoft Teams, then configured and trained a Grok bot that successfully worked through Teams while having access to the existing Cody knowledge base.
My work. Diagnosed why an AI assistant was not reliably parsing certain spreadsheet and CSV materials, then converted affected references to readable document formats.
Recorded outcome. The format-related retrieval problem was addressed in the documented support work. This did not, on its own, establish a complete AI support deployment.
My work. Created and refined scheduling SOPs, a conference-room Microsoft Teams guide, and directions for connecting a remote VoIP phone to Wi-Fi.
Confirmed outcome. The SOPs and guides were reviewed and implemented. The earlier “pending review” status is no longer accurate.
My work. Initially tested Cody AI's ability to converse with team members and reply inside Microsoft Teams. I subsequently created and trained a Grok bot, configured a Teams connector, also added the bot as a standard user, and made the organized Cody knowledge base available to it.
Confirmed outcome. The Grok bot was successfully made to work for Teams-based communication using the available knowledge resources. This case does not claim every potential question or action has been independently tested.
My work. Developed a script to map the company network drive into a text-based directory tree each day and made that current location reference available as part of the knowledge resources.
Confirmed outcome. The directory catalog supports questions such as “Where can I find the current procedure?” without requiring someone to manually search through folders. The case does not claim live file-content indexing, guaranteed retrieval accuracy, or verified automated synchronization of every downstream knowledge source.
Earlier test. Separately tested a quality-assurance process using Microsoft Forms, Teams links, and response delivery.
Recorded outcome. A prototype was tested. This Forms workflow is distinct from the later working Grok conversational assistant; a complete Forms migration was not established.
How a coworker's question can reach source knowledge, with a separate process for keeping file-location information current. These are functional relationships, not a disclosure of private configuration or connector permissions.
Reviewed scheduling procedures, conference-room Teams instructions, VoIP setup steps, and previously organized knowledge materials.
A custom file-inventory script produces an updated list of network-drive paths for file-location lookups.
Organized content and directory information made accessible to the trained Grok assistant.
A connector and standard Teams user setup provide a familiar place for employees to ask questions and receive replies.
Employee: “Where can I find the setup guide for a remote office phone?”
Knowledge assistant: “I can use the reference directory and knowledge base to help identify the relevant guide.”
The project owner reports reviewed and implemented procedures, a working Teams-based Grok assistant, and a script-generated daily shared-drive catalog that was supplied as a knowledge resource. The earlier Forms workflow remained a prototype, and initial Cody AI Teams experiments preceded the working Grok setup.
Actual assistant-answer accuracy, the identity/permission controls behind the connector and standard user, and automatic refresh behavior inside Cody were not independently tested for this public case.
No real company drive paths, employee names, ticket messages, bot credentials, account identifiers, or confidential documents are published. For the separate conversational IT ticket intake system and specialist bot delegation workflow, see the related projects.