Jump to content

Bromcom AI

supportman

By supportman
in MIS Systems

 Share

Recommended Posts

supportman

supportman

Posted
  • Author
Posted

I dunno, it could be good.

 

Imagine being able to ask an MIS chatbot "how many students you have on roll" or "Which staff are off sick today".

 

Some of those might actually be quite useful, rather than multiple menu clicks or slow widgets.

Marci

Marci

Posted
Posted (edited)

I don't disagree with the usefulness of it... (and indeed, can already do what you imagine thru PowerBI via it's natural language Insights engine if you view reports directly in PowerBI service / via PowerBI mobile app rather than within Bromcom's embedded viewer [ie: individual / pro / premium licensed users rather than single-pro-license for designer & shared licensing thru embedded viewer] - I can literally ask Siri for both those things _now_ and she'll just tell me, along with a tap-thru to go straight to that specific BI report in the PowerBI Mobile App)

 

But still needs the process accounting for within the scope & context of GDPR etc.

Edited by Marci
paulkerton

paulkerton

Posted
Posted

I can certainly see the benefit of natural language models being built into an MIS. Especially for less technical staff who want to retrieve assessment information without having to have the technical chops to use tools like Looker or Power BI.

 

Eventually you could use AI and natural language models to do things like arrange cover, or pull room occupancy statistics.

 

It's potentially an exciting area of development though I think like OpenAI and Bard it's going to be baby steps forward.

localzuk

localzuk

Posted
Posted
This keeps sending my brain back to Automated Decision Making & Profiling with a l'il flashy red light and a distant klaxon... eg: at what point is the line crossed...?

https://ico.org.uk/for-organisations/guide-to-data-protection/guide-to-the-general-data-protection-regulation-gdpr/automated-decision-making-and-profiling/what-does-the-uk-gdpr-say-about-automated-decision-making-and-profiling/

From reading it, there would still be the human doing the work - as the entire thing is about the member of staff asking it stuff. It isn't a totally automated process making decisions. The member of staff still has to go "yes, that is right, we'll use that", making the ultimate decision.

DeGrimmy

DeGrimmy

Posted
Posted

I've just had a demo at BETT - Interestingly, it's termed a "Personal Assistant" in Bromcom. It brings up a chat box on the right hand side and it gives you some suggestions or frequently asked questions. You can also type in your query at the bottom, like "Which staff members might retire soon" and it returns with a list of names who are retirement age and who will be in the next 2 years.

 

The guy demoing it said that it uses ChatGPT - I asked if it's self-contained and doesn't send data anywhere (I'm afraid I'm not very clued up about AI and ChatGPT) and he said that it uses ChatGPT for the language model, but then kicks it back to the database to pull the information you need. Hopefully someone who is a bit hotter than me on this sort of thing can ask the right questions!

  • Thanks 2
Bromcom_John

Bromcom_John

Posted
Posted
I've just had a demo at BETT - Interestingly, it's termed a "Personal Assistant" in Bromcom. It brings up a chat box on the right hand side and it gives you some suggestions or frequently asked questions. You can also type in your query at the bottom, like "Which staff members might retire soon" and it returns with a list of names who are retirement age and who will be in the next 2 years.

 

The guy demoing it said that it uses ChatGPT - I asked if it's self-contained and doesn't send data anywhere (I'm afraid I'm not very clued up about AI and ChatGPT) and he said that it uses ChatGPT for the language model, but then kicks it back to the database to pull the information you need. Hopefully someone who is a bit hotter than me on this sort of thing can ask the right questions!

 

Feel free to head over to the stand for a demo if you want to see what the potential is.

In answer to your question, and I can get actual technical updates from the team once they arrive and I can chat with them (real chat, not AI).

The implementation is entirely self contained where your own schools data is concerned. It is going to reach out to the web for certain things, like in your example of the retirement age staff. It will have reached out to the web for that data in terms of what retirement age "is" but then securely into your data to identify which of your staff are within that scope.

Everything is controlled within the existing data usage agreements and there will not be any use of your data to provision requests from other schools, unless you are already signed up to the data service agreement.

One of the big directions of development is to use the AI capability to live translate data within communications. Have a parent of an EAL student with German as their mother tongue, the system will live translate the message a d send it in Gmrnan. In reverse, the incoming message in German will be translated back into English.

 

It really is massive area of potential development.

  • Thanks 2
DeGrimmy

DeGrimmy

Posted
Posted
I've just had a demo at BETT - Interestingly, it's termed a "Personal Assistant" in Bromcom. It brings up a chat box on the right hand side and it gives you some suggestions or frequently asked questions. You can also type in your query at the bottom, like "Which staff members might retire soon" and it returns with a list of names who are retirement age and who will be in the next 2 years.

 

The guy demoing it said that it uses ChatGPT - I asked if it's self-contained and doesn't send data anywhere (I'm afraid I'm not very clued up about AI and ChatGPT) and he said that it uses ChatGPT for the language model, but then kicks it back to the database to pull the information you need. Hopefully someone who is a bit hotter than me on this sort of thing can ask the right questions!

I spoke to someone at the Arbor stand who was more customer service facing. They had a live demo at 12pm yesterday, but I unfortunately missed that. Instead, I was shown a slideshow of gifs. I asked the same questions as I did at Bromcom but they didn't have the answers to my questions because it was "quite new to them as well". Still, from the gifs, it looks quite similar to Bromcom's AI offering, but instead of just posting the results in the chat box, it actually generates a proper report for you.

FragglePete

FragglePete

Posted
Posted

We managed to see the Arbor demo of their 'Ask Arbor' system yesterday - and although made clear it's a very early it did look quite impressive with what they showed. Essentially within a Student Profile page click the Ask Arbor button - then proceeded to type just a couple of sentences along the lines of ...'Doing very well but can sometimes be chatty in classroom' - it then proceeded to produce 4 paragraphs of detailed information from the data of each subject for that student based on the initial 'seed' which was well written and could pass as a perfectly good 'report' or individual progress update of the student.

 

A question was asked if there was an additional cost for this and they did believe they're would be which would allow a certain number of 'credits' - as each individual question does involve a small cost with OpenAI. Again, early days and very open to the fact it's still all in development, nothing decided yet, etc, etc.

 

Pete

Bromcom_John

Bromcom_John

Posted
Posted

and my technical colleague provided me answers as well

 

1> When a user types in a question we only send the question to ChatGPT or an Azure bot to understand what the user is looking for. Once we get a ‘machine understandable’ version of the user question we then go to our database to pick up data from the respective school. This is the returned to user. ChatGPT is therefore completely agnostic about the data we have in our database. At no point is our data or database exposed to ChatGPT or any third party.

2> Should the answer to the user’s question be found in the internet then we obviously avoid going to our database and rely on ChatGPT’s answer to show to the customer.

 

So in both cases ChatGPT never has access to our data or database. This ensures we are compliant with GDPR and protect our customers’ data.

  • Thanks 1
  • 1 month later...
Marci

Marci

Posted
Posted

(1) No - not til Phase 2 of rollout. In current state, it's just a regular ChatGPT session embedded into Bromcom, and doesn't touch your own data therefore needs no permissions.

(2) Not as they stand, no, but I've raised this.

CHiLL

CHiLL

Posted
Posted (edited)
(1) No - not til Phase 2 of rollout. In current state, it's just a regular ChatGPT session embedded into Bromcom, and doesn't touch your own data therefore needs no permissions.

(2) Not as they stand, no, but I've raised this.

(1) is pretty evident when you try and use it! I asked "How do I take a register" and it told me to make a list of names of people you expect to attend, for those who arrive late add them and mark accordingly and write who is absent with the reason why if provided. While it's not wrong, a true tailored experience would explain how to do it in Bromcom. Clearly it's a work in progress (so why add it when it's not ready?).

Edited by CHiLL
localzuk

localzuk

Posted
Posted (edited)

The notice that comes up has flagged some serious concerns about compliance with data protection rules for us. I've pushed it to our DPO to investigate. Specifically, the blurb says that OpenAI would be a data controller etc... Which, as they're in the USA and we don't have any direct contract with them raises some serious GDPR concerns.

 

We're leaving it disabled for the time being. At least until our DPO can provide us with some advice.

 

I think it fits into the wider questions about AI in education as well. There's a much larger conversation going on regarding it all, and I don't think we as schools should leap in and get it wrong.

 

So in both cases ChatGPT never has access to our data or database. This ensures we are compliant with GDPR and protect our customers’ data.

As it stands, the notice we're shown if we click on the icon as admins does not make this clear. It is very apparent that Bromcom is implementing this new feature quickly, and the notice tries to remove any responsibility from Bromcom - which is very concerning.

Edited by localzuk
Theblacksheep

Theblacksheep

Posted
Posted (edited)
(1) No - not til Phase 2 of rollout. In current state, it's just a regular ChatGPT session embedded into Bromcom, and doesn't touch your own data therefore needs no permissions.

(2) Not as they stand, no, but I've raised this.

 

I understand it doesn't currently, but when integrated will it use the existing permission structure to provide answers. It's OK for HR to ask about staff age profiles, but maybe not regular staff.

 

The HTs I have spoke to want chats recorded, mainly for safeguarding purposes.

 

I also wonder with questions about 'how to take a register' could be fed with instructions related to Bromcom. The support documents are incomplete at best, so it would be useful for staff to be able to learn Bromcom from BromcomAi.

 

 

The notice that comes up has flagged some serious concerns about compliance with data protection rules for us. I've pushed it to our DPO to investigate. Specifically, the blurb says that OpenAI would be a data controller etc... Which, as they're in the USA and we don't have any direct contract with them raises some serious GDPR concerns.

 

We're leaving it disabled for the time being. At least until our DPO can provide us with some advice.

 

I think it fits into the wider questions about AI in education as well. There's a much larger conversation going on regarding it all, and I don't think we as schools should leap in and get it wrong.

 

As it stands, the notice we're shown if we click on the icon as admins does not make this clear. It is very apparent that Bromcom is implementing this new feature quickly, and the notice tries to remove any responsibility from Bromcom - which is very concerning.

 

There is a difference between web and API use, with API being a better option from a DPA perspective, explained here:

 

https://simpliant.eu/insights/GDPR-requirements-when-using-chatgpt-api

 

API usage data is opt-in and it's deleted after 30 days:

 

https://openai.com/policies/api-data-usage-policies

Edited by Theblacksheep
Bromcom_John

Bromcom_John

Posted
Posted (edited)

Morning all,

 

This is the phase 1 release which is mostly a proof of concept that the AI tools are going to be functional and accessible within the MIS environment.

Even once we begin the rollout of Phase 2, which is when the systems will start being able to give responses based on the schools own data, the CHatGPT subsystems will not be exposed to school specific data.

So just to repeat, that when using Bromcom AI, no Management Information System (MIS) data or user information is transmitted to OpenAI within the MIS platform.

It is important to note that in this phase 1 release, all information voluntarily entered by the user in the free text chat box within the Bromcom AI panel ie "the question", is transmitted to OpenAI, similar to the behaviour of ChatGPT.

We absolutely acknowledge the paramount importance of user privacy at all times and will update the relevant policy terminology to reflect this.

 

As far as I am aware the team are aiming to have the functional prototype of "phase 2" ready to show at our Shard event in a few weeks.

Edited by Bromcom_John
  • Thanks 1

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share
Go to topic listing


×
×
  • Create New...