AI Agents Training: What Each Agent Does, How to Customize It, and How to Turn It On
1. Understand the purpose of AI agents in Pete 5:03
The training is about AI agents start to finish: what each one does, how to customize it, and how to enable it.
The core idea is not replacing your team.
AI is meant to cover the gaps when humans are busy.
It helps with missed calls, web leads, text replies, follow-up, and call scoring.
The main business problem being solved is coverage and speed:
People can only handle one call at a time.
Leads can sit too long when the team is busy.
Long-term follow-up and coaching often get delayed.
2. Learn the AI agent lineup 6:45
The training introduces the main agents in the Pete ecosystem:
Front Desk Finn — answers inbound calls.
Rapid Reach Rose — calls sellers when a lead comes in.
Texting Taylor — handles text threads.
Follow Up Fay — makes follow-up calls from workflows.
Mentor Max — scores transcribed calls.
Partner Pete — lets you ask questions in plain language.
You can use:
the full Team Pete package, or
individual add-ons for just one role.
Some agents are usage-based:
voice agents and outbound calling agents have per-minute usage
texting has per-message-segment usage
Partner Pete is included with the add-on
3. Understand roles vs. personas 9:35
A role is the agent type itself, like Front Desk Finn or Follow Up Fay.
A persona is a customized version of that role.
Example: “After Hours Finn” vs. “First Call Finn.”
Each persona can have its own:
name
description
voice
knowledge files
booking/scheduling setup
This is the main way to make the AI sound and behave like your business instead of the default Pete version.
4. Customize an agent persona 10:04
Open an existing Pete-branded agent to review its settings.
You can see and adjust things like:
scheduling templates
appointment booking behavior
team routing
knowledge base links
voice selection
You can preview the voice, change speech speed, and test outbound calls.
Pete-branded agents are usable as-is, but you cannot directly edit them.
To customize, you need to copy or create your own version.
5. Create your own agent from an existing one or from scratch 11:34
Use the plus button to create a new agent.
You can start from an existing agent if you want a head start.
The system prompt is the agent’s instruction set.
You can rename the agent.
You can edit the prompt to change behavior.
The prompt may include:
comments marked with
#variables in curly braces like
{{company_name}}or{{caller_name}}
You can also copy one of your own agents if you want to make small changes without losing the original.
6. Add knowledge files to give the agent background information 16:08
Think of the prompt as the script.
Think of the knowledge base as background information.
Knowledge files can include things like:
zip codes you buy in
counties you buy in
solutions to common problems
internal business details
Once linked, the agent can use that information automatically when answering questions.
This helps the AI avoid giving generic answers and instead respond based on your actual business data.
7. Configure Front Desk Finn for inbound calls 17:27
Front Desk Finn answers inbound calls and can:
introduce itself
collect callback name, phone number, email, and message
book appointments using scheduling templates
You can decide where it routes calls:
to the AI first
to the team first
to voicemail
You can also choose whether it handles:
all calls
only after-hours calls
overflow calls
repeat callers only
This is controlled at the campaign and phone number level.
8. Set inbound call routing rules by campaign 18:11
In campaign settings, you can choose who answers first.
Options include:
Front Desk Finn answers before the team
the team rings first
direct voicemail handling
You can also choose whether AI handles:
all leads
only new leads
repeat callers differently
This lets you create different experiences for different phone numbers.
Example: one number for new leads, another for long-term follow-up.
9. Use Rapid Reach Rose for speed-to-lead follow-up 20:51
Rapid Reach Rose is designed for web leads and speed-to-lead.
The flow can be:
Try to call a human first.
Wait a set amount of time.
If no human connects, have Rose call the lead.
You can set the wait time before Rose calls, such as 3 or 5 minutes.
Rose can be customized to:
book appointments
collect basic contact info
simply say the team is busy and request a callback
You can also limit Rose to after-hours only if that fits your process better.
10. Configure Texting Taylor for inbound and outbound SMS 23:23
Texting Taylor works in two ways:
replies to incoming text messages
sends outbound texts from workflows
You can create a persona for Taylor with a detailed prompt.
On a campaign, you can enable SMS AI and choose which Taylor persona responds.
Replies usually go out in about a minute, with a slight delay so they feel natural.
You can turn Taylor off on a specific property/contact if you want to take over manually.
11. Use opt-out handling and text controls 25:39
Texting Taylor can recognize opt-out intent.
Example: if someone says “don’t text me anymore,” the system can stop future texts.
If the person says “stop,” it will stop texts to that number.
You can also configure what happens after opt-out:
stop active workflows
update phase or phase status
mark the contact in a specific way
On a property record, you can see whether Taylor is active:
green = enabled
red = disabled
yellow = following campaign settings
12. Use Texting Taylor in outbound workflows 28:58
In workflows, you can trigger an outbound SMS using Texting Taylor.
This is useful for long-term follow-up sequences.
Instead of sending a fixed template, Taylor can:
review conversation history
use context from the lead/property
send a more natural probing message
You can choose between:
very specific templated wording
more flexible AI-generated wording
This gives you control over how automated or conversational the follow-up should be.
13. Use Follow Up Fay for outbound calls 31:47
Follow Up Fay handles outbound follow-up calls.
The persona prompt can be customized for your style and goals.
A key instruction is to let the seller talk most of the time.
The example mentioned aiming for the AI to talk only about 20% of the time.
Fay can be used for:
long-term follow-up
post-appointment follow-up
checking in after an acquisition manager visit
You can also generate a prompt from an existing script or document instead of starting from scratch.
14. Trigger Follow Up Fay from workflows 33:26
Follow Up Fay is launched through workflows.
Example: on day 32 at 11:00 AM, initiate a call for follow-up.
Fay uses the full context available in Pete:
notes
conversation history
property details
prior interactions
You can steer the call with the persona prompt.
A practical use case is post-appointment follow-up:
ask how the appointment went
gather feedback on the acquisition manager
uncover details that may not have come up during the visit
15. Use Mentor Max to score calls automatically 35:29
Mentor Max scores transcribed calls using a rubric or prompt.
The system can now do this automatically instead of requiring manual clicking.
In phone settings, you can enable:
scoring every transcribed call
auto-detecting the call type
Mentor Max can classify calls such as:
initial lead intake
long-term follow-up
offer negotiation
offer presentation
rent inquiry
maintenance request
speed-to-lead
This matters because different call types should be scored differently.
16. Review Mentor Max scoring results and use them for coaching 38:03
Mentor Max provides a score and written feedback.
Example feedback may include:
whether motivation was uncovered
whether urgency was discussed
whether next steps were clear
This gives immediate coaching insight without listening to every call.
You can also use these scores to build KPIs and track trends over time.
Example: average score by person or by call type
Example: whether performance is improving or declining
17. Use Partner Pete for natural-language help and analysis 39:33
Partner Pete is the conversational assistant layer.
It lets you ask questions in plain language.
It can answer based on the property you are currently viewing.
Example questions:
“What dispo strategy makes the most sense?”
“Which leads haven’t been touched since the first call?”
“What are my best lead sources?”
It can also help with:
deal analysis
property updates
notes
field updates
analytics across your database
18. Use Partner Pete to update records by voice or chat 41:21
You can tell Partner Pete to update notes or fields conversationally.
Example actions it can help with:
update ARV
update rehab estimate
move a property to another record
add notes from a conversation
This is especially useful on mobile.
The idea is to reduce manual data entry and let you interact with Pete like an assistant.
19. Ask questions about business performance and deal history 42:19
Partner Pete can also be used for analytics.
Example questions:
Which are my best rentals?
Which are my best flips?
Which areas do I get stuck on?
How is a team member doing on offer negotiations?
It can use historical deal performance and database context to answer.
This is positioned as a fast, built-in alternative to building a separate analytics system.
20. Discuss practical concerns and how AI fits with human teams 43:14
The group discussed how much customization is needed.
The answer: you can keep it close to the default or tailor it heavily.
Differences in wording, market, and process are expected.
The system was intentionally built so you can:
create from scratch
modify existing agents
add your own business language and knowledge
21. Review pricing and subscription structure 44:37
The agents are additional subscriptions.
General pricing discussed:
most agents are about $79/month each
Partner Pete is $99/month
Team Pete package is about $319/month if you want more than three
There are also usage-based charges:
per-minute fees for voice agents
per-text-segment fees for SMS
The team emphasized that the usage pricing is relatively low compared to some other AI tools.
22. Address concerns about AI mistakes and monitoring 47:57
A major concern is: what if the AI says the wrong thing?
The response was that monitoring matters.
Mentor Max helps by scoring AI calls too, not just human calls.
That gives you feedback so you can:
adjust prompts
update knowledge files
change voice or behavior
compare performance over time
The team emphasized that humans also make mistakes, so AI should be monitored like any other team member.
23. Show real-world use cases for AI inside Pete 57:00
The team shared examples of how they use AI in real operations:
property management call handling
rental availability questions
maintenance triage
better intake questions than a human might ask
Example: if a tenant says the toilet is not working, the AI can ask follow-up questions to narrow down the issue.
This shows that the agents are not just for acquisitions—they can support multiple business functions.
24. Learn how to purchase and enable the agents 58:52
To buy agents, go to Settings > Subscription.
The account owner can enable add-ons by clicking the plus button.
The system will prorate the current month and add the subscription to the monthly renewal.
You can also contact support or the sales team for help choosing the right setup.
The same subscription area is where you manage add-ons like:
AI agents
Buyer Search
AI REI Content Suite
25. Wrap-up and next steps 01:02:48
The session ended with a reminder that this training was moved up by request.
If people have ideas for future Thursday trainings, they should send them in.
The main takeaway:
AI agents can cover gaps in your process
they can be customized to your business
they can be monitored and improved over time
The team invited follow-up questions after the session.