<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=1094218042721276&amp;ev=PageView&amp;noscript=1">
Skip to content

AI Baked In - the Good Kind Entry 98

How can nonprofits use HubSpot AI agents without sacrificing trust?

HubSpot AI agents can answer common questions, use approved knowledge and CRM context, complete limited tasks, and route complex situations to the right person. A carefully scoped agent gives donors, volunteers, members, event attendees, and staff timely support while protecting the human relationships and sensitive decisions that should remain with people.

Reduce repetitive work without removing human connection

Nonprofit teams routinely answer questions about background checks, year-end giving letters, recurring gifts, event parking, membership renewals, and where to mail a check. Agents can handle approved, repeatable requests so staff can focus on fundraising, programs, relationships, and mission-critical work.

  • Donors
  • Volunteers
  • Members
  • Program participants
  • Event attendees
  • Sponsors
  • Community partners
  • Board members
  • Advocates
  • Grant funders

What is an AI agent?

An agent goes beyond generating a text response. Depending on its tools, permissions, and configuration, it can combine language understanding with organizational knowledge and approved actions.

Agent capability Possible nonprofit use
Understand and clarify Interpret a question and ask an appropriate follow-up
Search knowledge Find an approved answer or resource
Use CRM context Recognize relevant status or prior activity
Take an approved action Create a task, start a workflow, or route a request
Escalate Transfer the conversation and its context to a person
Important boundary

Keep sensitive decisions and high-risk situations with people

Agents should not independently decide program eligibility, beneficiary services, major-donor strategy, legal or tax matters, financial hardship, safeguarding, employment, medical needs, or crisis responses.

They should never request, expose, or transmit full payment-card information, passwords, background-check details, protected health information, or other sensitive data through an inappropriate channel.

Support volunteers through common questions

  • How to apply
  • Where to complete orientation
  • Whether a background check is required
  • How to submit documents
  • Where to find opportunities
  • How to select a shift
  • What to wear
  • Where to report
  • Who to contact about an absence
  • How to record hours

The agent can use approved documentation and direct volunteers to the correct form, portal, scheduling page, or staff member.

Guide approved volunteer-onboarding steps

  1. Confirm the volunteer role of interest.
  2. Explain the application process.
  3. Link to required forms.
  4. Provide background-check instructions.
  5. Direct the person to orientation.
  6. Share approved training resources.
  7. Identify available shifts.
  8. Confirm next steps.
  9. Escalate special circumstances.

The agent may explain the process, but it should not expose or interpret background-check results. Suitability, exceptions, accommodations, safeguarding, and eligibility stay with authorized staff.

Answer routine donor-service questions

  • How to donate
  • Where to mail a check
  • How to donate stock
  • How to join recurring giving
  • How to update a payment method
  • How to change or cancel a gift
  • Where to find giving history
  • How to request an acknowledgment
  • How matching gifts work
  • Who handles major gifts

The agent can provide approved instructions and connect the donor with a person when the relationship or request needs personal attention.

Retrieve year-end giving information securely

  1. Authenticate or appropriately verify the requester.
  2. Identify the relevant donor record.
  3. Confirm that an approved document exists.
  4. Trigger a secure delivery process.
  5. Record the request.
  6. Create a task if review is needed.
  7. Escalate discrepancies to finance or development.

An agent should not change giving totals or make tax determinations. Questions involving deductibility, valuation, hard and soft credits, donor-advised funds, in-kind gifts, or tax treatment require authorized staff.

Payment security

Send recurring donors to a secure management experience

An agent can explain how to update an amount, frequency, payment method, pause, or cancellation and then direct the donor to an approved payment-management page. It must never ask for full card details in ordinary chat, email, forms, or CRM notes.

Provide immediate event support

  • Registration and tickets
  • Location and parking
  • Accessibility
  • Schedule and attire
  • Meal selections
  • Guest policies
  • Sponsorships
  • Volunteer shifts
  • Auction-item delivery
  • Cancellation policies

Time-sensitive sources need active maintenance. An outdated start time, parking location, accessibility instruction, or cancellation policy can create immediate confusion.

Help members find the correct next step

Agents may answer questions about membership options, benefits, eligibility information, renewal dates, payment instructions, account access, event registration, directories, continuing education, and chapter contacts.

Disputes, exceptions, refunds, and eligibility decisions should generally be escalated to staff.

Provide approved program information with strict limits

  • What the program provides
  • Where services are available
  • When applications open
  • Which documents are required
  • How to begin applying
  • Where to find emergency resources
  • Which team handles the request
  • When a person must take over

An agent should not independently determine eligibility, medical needs, legal status, safety, crisis intervention, or access to essential services unless the organization has designed and approved a compliant process.

Support staff with internal agents

  • Locate policies
  • Find approved templates
  • Summarize CRM activity
  • Identify process owners
  • Retrieve program information
  • Find campaign resources
  • Explain procedures
  • Prepare meeting summaries
  • Surface incomplete records
  • Route support requests

Internal access still requires controls. An agent's ability to locate sensitive donor, employee, beneficiary, or financial information does not mean every staff member may receive it.

Build every agent on approved knowledge

  • Knowledge-base articles
  • Frequently asked questions
  • Volunteer manuals
  • Event guides
  • Membership policies
  • Donation instructions
  • Payment documentation
  • Program descriptions
  • Internal procedures
  • Training resources
  • Approved email templates
  • Website content

Each source needs an owner, approval status, and review date. Outdated or contradictory information can produce a confident but incorrect response.

Use CRM context to avoid generic answers

With appropriate permissions and identification controls, an agent may recognize that someone is already registered for an event, has an active recurring gift, completed volunteer orientation, is approaching membership renewal, already received an acknowledgment, or has an assigned support request.

The agent should use only the minimum information necessary for its approved task.

Allow only defined and approved actions

Possible action Example limit
Create a support ticket Use approved categories and routing rules
Update a property Restrict the fields and values the agent may change
Create a task Assign only to designated teams or owners
Start a workflow Use a tested process with suppression and exit rules
Send a link Use a current, approved destination
Schedule a meeting Respect availability and relationship ownership
Record a request Do not imply the underlying change is complete
Escalate Preserve context and route to the correct person

Design the agent to know when to stop

  • A person requests human help
  • The agent lacks confidence
  • Sensitive information is involved
  • A record appears incorrect
  • The request falls outside scope
  • The constituent is upset or distressed
  • A refund or exception is requested
  • Eligibility or safeguarding is involved
  • A major donor needs attention
  • Legal, medical, tax, or crisis guidance is requested
  • A complaint may be involved
  • Authentication fails

The handoff should preserve conversation context so the constituent does not have to start over.

Support high-touch relationships without replacing them

For major donors, agents can answer logistics, schedule meetings, retrieve approved resources, record requests, create follow-up tasks, notify relationship managers, and provide service outside office hours.

For high-volume, lower-dollar programs, agents can provide timely self-service while keeping a reasonable path to human support available to every donor.

Give each agent a narrow, documented purpose

  • Answer volunteer-onboarding questions
  • Provide current event logistics
  • Guide secure recurring-gift management
  • Answer common membership questions
  • Retrieve internal process documentation
  • Route inquiries to the correct team

“Help with anything” is not a safe or useful scope. A focused responsibility is easier to test, govern, monitor, and improve.

Authentication and identity

Verify identity before revealing or changing personal information

Public questions such as an event start time or mailing address may require no identification. Sending a giving statement, showing donation history, changing a recurring gift, updating contact details, or confirming membership status does.

Knowing a donor's name or possessing an email address may not be sufficient verification for a sensitive request.

Apply data minimization to every interaction

  • Limit accessible CRM properties
  • Restrict sensitive records
  • Define shareable information
  • Limit update permissions
  • Require authentication for sensitive actions
  • Set conversation-retention rules
  • Define storage periods
  • Document external data recipients
  • Support deletion and corrections
  • Apply controls at least as strict as human access

Test ordinary, unusual, and adversarial conversations

  • Common questions
  • Unclear questions
  • Incorrect assumptions
  • Sensitive requests
  • Requests for personal data
  • Angry or distressed language
  • Out-of-scope requests
  • Outdated information
  • Conflicting knowledge
  • Attempts to override instructions
  • Requests for a person
  • Misspellings and incomplete details

Confirm that the agent answers accurately, refuses inappropriate requests, and escalates correctly.

Monitor performance after launch

  • Questions answered
  • Resolution and escalation rates
  • Unanswered questions
  • Incorrect responses
  • Constituent satisfaction
  • Repeated knowledge gaps
  • Completed and failed actions
  • Average response time
  • Human workload reduced
  • Privacy or security concerns
  • Corrective follow-up needs
  • Mission value created

Review performance continuously and after changes to programs, events, policies, integrations, permissions, or knowledge sources.

Confirm available agent capabilities

HubSpot agent functionality varies by subscription, product, seat, channel, language, permissions, region, credits, connected systems, and rollout.

  • Available Breeze Agents
  • Supported communication channels
  • Accessible knowledge sources
  • CRM data access
  • Supported records and properties
  • Available actions
  • Human-routing behavior
  • Identity-verification options
  • Usage and credit limits
  • Access permissions
  • Conversation retention
  • Required integrations or APIs

Best practices for nonprofit AI agents

  1. Give every agent a narrow and documented purpose.
  2. Use approved, current knowledge with named owners.
  3. Provide clear human-escalation paths.
  4. Minimize access to personal information.
  5. Verify identity before sensitive disclosure or changes.
  6. Never collect full payment-card details in ordinary conversations.
  7. Keep background-check details restricted.
  8. Route legal, tax, medical, crisis, and safeguarding issues to people.
  9. Test attempts to move the agent outside its scope.
  10. Preserve context during escalation.
  11. Monitor incorrect and unresolved responses.
  12. Review usage, permissions, and security regularly.
  13. Tell constituents when they are interacting with AI.
  14. Keep human assistance reasonably accessible.
  15. Measure mission value, not only reduced staff time.

Give staff more time for the work that needs people

With reliable knowledge, limited permissions, secure actions, identity controls, human escalation, and continuous monitoring, agents can handle repetitive questions and approved self-service without displacing the relationships, judgment, and care at the center of nonprofit work.