What are the use cases for MCP?

There is an interesting thread here where user tom talks about what they've done with the bob MCP and this part of the thread in particular shows an illustration of what you could do for data querying- https://heartcorehr.hibob.com/ai-innovators-315/have-you-integrated-bob-with-any-ai-friendly-tools-5739?postid=23683#post23683 and then further down in the thread shares some more examples and links to their blog, where they talk about in more detail the hows to of MCPs, HiBob and n8n.

We haven't connected the Bob MCP to anything yet, as we're waiting for the per-user authorisation, which I have read is coming at some point this year. A lot of our use cases revolve around querying data, taking actions, and more customised dashboarding. We have held off diving too much into this, as with Bob's out-of-the-box AI chatbot assistant coming, some of the use for the MCP may be resolved in that. The second hold up for us is the big question of where we want people to be searching for info and doing actions, and we don't know if that is Guru or Hibob or Slack or somewhere else (we are using Claude for our AI). Until we have a clear view on what our AI strategy looks like, we probably won't connect any MCPs. 

One interesting use case that we are considering is connecting HiBob to our knowledge database, which is a tool called Guru. What this will allow us to do is then use Guru's AI agent to perform actions and source data. A lot of our Guru knowledge base resolves around policies, so an example use case we are considering is for Annual leave.

• If any employee is asking Guru for information about annual leave, currently, it would tell them that the policy says if you're an Australian employee you get 20 days per year
• unless you're part-time, in which case it's 20 days prorated
• unless you're a nurse, in which case you get 25 days
• unless you're a part-time nurse, in which case its 25 days prorated
• and then it's up to the person to determine what category they fall into and, if they are part-time, work out the amount of days. 

With the data flowing from Bob, Guru should be able to read:

• this employee is a nurse
• who works three days a week
• so based on the policy, they would get 15 days of annual leave per year, no requirement for the employee to do any maths or work out which category they fall into. 

Then we would build it so that it Guru chat would prompt to offer the actual available balance in hibob based on accruals (eg 12 days) and then it would prompt them to see if they want to put in a time-off request and what those dates might be and it would do this for them. This is all hypothetical; we have not built or tested this. We are just building out what a use case might look for once we get to the point that we can consider connecting it. Connecting with Guru for us will also allow connections to other places. So, for example, putting in a Jira ticket if they need special dispensation to have their laptop work from a different country while they're on annual leave, connecting into Slack to send their manager a message to say, "Hey, look out for this time off request that's coming through," and many other potential options. That is all if we decide that Guru is going to be the place that we base our AI web from for people-related actions/query-ing.

I expect that's probably not what you're looking for, as it isn't an actual live example, but it's the closest I have to share, as we are not yet ready to build it out but want to imagine what might be possible. 

I see multiple use cases:

• General analytics, as an HR analyst being able to query via natural language rather than remembering which report to use. Connecting to both HiBob data and other operational data will make analysis better and more relevant.
• Having company documents and policies stored in Bob read by AI to provide accurate answers to employees.
• Being able to streamline various processes for admins
• Routine data hygiene checks done autonomously

The problem is always the data model, and ensuring all new features are also available in the APIs and via MCP. Without parity in UI and machine use of the data, it is difficult / impossible to adopt quickly. 

Great response ​@Amy Ward  and ​@Agnar! 

Some things I’ve thought about were how to celebrate events. Using the MCP and API’s, we could build a multi-agent workflow to celebrate/highlight events like those that want their birthday celebrated and work anniversaries. It could build an out a jira board for new hires (our tech team lives on jira and tend to not notice our HiBob task lists for those areas) as well as a leavers task list. The Bob agent is very good for getting info that it can reach right now but going further with an HR on Demand function could be helpful. A digital HR agent that could talk to multiple areas and recommend training for situations based on what they’ve done in the system and our other learning system Thought about building a compliance agent to ensure no compliance deadlines or tasks are missed. Lastly, connecting some of our external learnings into HiBob (Microsoft Learn which can be pulled through their API). 

Once, I get access- I plan on using it to replace the current WFP dashboard- it’ll be easier to track headcount, open positions- track attrition roles- etc. At least that’s the goal once I get access to MCP-