Working with Shannon. Talking about the time it takes to use the AI agents. Talking with Shannon about the type of AI tool we are trying to use and setup. We spent some time talking about concerns and possible constraints, moving forward.

My constraints:

- I'm willing but should we do this? I really want it to work - I can see the potential - We need a master trainer that can help others

- We need to communicate - Steve and I

- My time

- Already put in a ton of time

- Questions on funding and budgets

- Train the mother AI engines and models or the mini model? Which one(s)?

- I need some direction - too many things on the plate and I don't know what to prioritize

- The training pieces are all over (on the web and in certain docs) also in my head - No one spot for all training currently

- Things keep changing

- I am saving my conversations with the AI agent, but it takes time to format them, and if it (the training) doesn't stick, then what?

- Putting in energy but not getting the output that is desired - we can only go fast for so long - we can keep chipping away at things, but that is it

- There are lots of parties involved, adilas, Nxtlinq, our users, and the actual AI agent

- We have a vision and we can build a number of things out... we have plans - those plans will take months and months - kinda feeling pressured

- Do we really want to have meetings every two days

- Tons of potential... how far are we willing to push it?

AI agent constraints:

- I've done x hours on the mini model, currently not retaining anything

- Storage at the database level - For them and for us - How can we help this process?

- How can we help with storage - Known needs - System level, corp level, and user level

- They are retaining the chat messages - At the Nxtlinq level - We are not

- Question - How do we really train this agent or entity?

- It, the agent, is giving bogus answers - For example - How do I get rid of an element of time - How can I tell how many elements of time I have (see stored chats from 11/25/25 - Some other user, not from me)

- Can it do what we need it to do

- Cut off from the mother, meaning the agent or mini model - Already over a year behind the data on the mother AI engine

- We don't know the model that it is based on... It really likes the security and blockchain stuff - Meaning what was the original purpose of the AI agent.

- We just need a simple bot that can translate natural language into adilas tools and prompts

- Not retaining information long term... Currently only per person, per session, and that's it

- How are they doing with the crawler to pick up content

- Demands - too much on us, build, train, market, deploy, and it doesn't even work all the way yet

- People aren't wanting it... what it currently is and can do... It take too long, it requires too many two-part authentication stuff, and it can't do advanced or multi step things...

- It is being used less and less - People want it, but if it gives them the wrong answer and isn't fast enough, they will move on

- Too slow, not helpful enough (wrong answers), too many auth steps, can't do complex actions, etc.

- Say you aren't using it... then you decide to use it... and it takes multiple steps to do the auth stuff... you have a quick question and you want an answer in a couple of seconds... you then have to do a multi-step 2-5 minute process to get your answer. Pretty soon, you say, no, not doing that... that is from our users.

At the end of the session with Shannon, we were talking about our simplified AI plan - 1. Teach it how to do some basic or system navigation, 2. Get it all trained up so that it can help with training and support, 3. The ideal is that it knows everything and can help with the higher consulting level and business level questions.

Shannon and I were also talking about the jellyfish model and helping to stabilize the team and their roles. There is a need there.