Bonterra provides software for nonprofit organizations, public-sector teams and corporate grant makers. Its portfolio spans fundraising, donor management, case management and grant making. The people using these products often work in small teams with limited time and little specialist support.
Persistent leveraged Claude Code and worked closely with Bonterra to develop Bongentic, a reusable agentic AI platform that serves as the foundation for Que, Bonterra’s AI teammate. Rather than creating separate assistants for each product, the team built a shared AI infrastructure once and made it available across Bonterra’s portfolio, enabling scalable and consistent AI adoption across products.
Their work is varied but consistently demanding. A fundraiser may need to draft a donor appeal and build an audience. A case manager may need a quick summary of a participant’s recent activity before a meeting, pulled from case notes scattered across dozens of records. A grants manager may need to assess whether an opportunity is worth pursuing. Until recently, this work got little help inside the Bonterra products where it already happens.
The wider sector faces similar pressure. Sage’s 2025 survey of more than 350 nonprofit leaders found that 41% cited a lack of process automation and organizational efficiency as a leading operational challenge, while 35% identified manual, time-consuming reporting.
Building a Reusable AI Foundation
Persistent made technical contributions to the agentic AI platform that powers Bonterra Que the sector-informed, ethical and proven AI built for social good. The objective was not to create a separate agent experience for every product. It was to build shared infrastructure once, then make it available across Bonterra’s portfolio.
Que is now live in EveryAction, Network for Good, Apricot, DonorDrive, OneCause, CyberGrants and Nonprofit Hub. It supports product-specific work such as drafting donor communications, building audience segments, generating fundraising forms, summarizing participant case histories, identifying possible duplicate records and offering context-specific guidance. Each product can use tailored skills while benefiting from the same underlying safety, quality and delivery foundation.
Que works directly inside the tools teams already use, either in a chat panel or built into the screen itself. Staff can ask for help in plain language without leaving their work. Its skills work securely within the organization’s own database informed by patterns learned across the wider Bonterra Network helping it provide guidance that reflects the organization’s unique context rather than generic advice.
AI built directly into the workflows people already use not another tool to manage.
Keeping People in Control
Human control is central to the experience. Que can draft, suggest and prepare work, but it does not send an email, save a record or submit a form without review and approval from a staff member. Each skill includes this review step, so efficiency never comes at the cost of accountability. Bonterra also reviews every Que skill through a formal AI governance process before it reaches any customer.
That approach reflects a wider adoption need. Salesforce’s 2025 Nonprofit Trends Report found that 55% of nonprofits were actively using or piloting AI, while 63% identified data privacy and security as concerns.
Built for trust: AI that serves people ethically, with humans always in control.
Fundraising outreach doesn’t stop at email. Many organizations now reach donors by text as well and each channel has its own demands: an email needs to fit a template and sound like the organization, while a text has to say the same thing in far less space. Que offers a dedicated drafting skill for each. For email, Que starts from the organization’s own past emails, borrowing tone and content from close matches and falls back to its saved brand voice when nothing similar exists. It also fills in the organization’s chosen template section by section, so the draft arrives ready to review and if the team switches templates, Que moves the content into the new layout. For text messages, Que keeps every draft within Network for Good’s 150-character limit and flags emoji, which can cut the usable length to about 70 characters. Either way, fundraisers can refine a draft just by asking “make this friendlier,” “shorten it” before Que saves it to Network for Good.
Fundraising analysis grounded in real numbers
Most Network for Good organizations have years of fundraising history donor counts, revenue, retention, average gift size but turning that history into answers has usually meant exporting data to a spreadsheet or waiting on a manual report. Que’s fundraising analysis skill answers plain-language questions like, “Where should I focus my fundraising efforts?” or “How does my donor retention compare to organizations like mine?” using up to five years of the organization’s own data. It shows the answer as a clear chart and, where it helps, compares results with the average for similar organizations. Ask about several metrics at once and Que plots them together so trends are easy to spot. The numbers come straight from the data, not AI estimates.
A shared technology foundation
Bonterra’s in-house agentic platform provides the shared technology foundation behind these Que experiences. Amazon Bedrock hosts Claude models and applies Guardrails before responses reach users. The platform can gracefully route certain requests to lighter, faster models when needed. Amazon DynamoDB stores conversations, tool results and organization-specific configuration, while Amazon S3 and S3 Vectors support memory and product knowledge bases.
Infrastructure that makes every product better built once, reused everywhere.
Business Impact
From Isolated AI Features to a Reusable Product Capability
- Que is live in EveryAction, Network for Good, Apricot, DonorDrive, CyberGrants, OneCause and Nonprofit Hub.
- Each product rollout inherits shared guardrails, review controls and platform capabilities.
- In Apricot, Que’s Image Reader, powered by Claude, reduces manual document-entry work that previously took 15–45 minutes per document to a review process measured in minutes.
Network for Good shows what that reuse looks like in practice. Beyond email and text, staff use Que’s Donor Segmentation skill to build donor lists from a plain-language request, Product Support to get quick answers without leaving the screen, Donor Stewardship to draft personal notes to individual donors and Personalized Coaching to plan campaigns and get expert guidance in seconds. These skills run on the same Que platform delivering 44% more in donation totals compared to non-Que users. Whenever a skill produces a draft, note or list, a person reviews it before anything is sent or saved.
Persistent assisted Bonterra to create more than a single AI feature. The Que platform establishes a reusable pattern for supporting real work across multiple products: embedded assistance, accessible UX organization-specific context, shared safeguards and human approval before consequential actions become final.
For Bonterra, that creates a foundation for faster delivery of new AI capabilities and more accessible experiences for nonprofit teams.
The most trusted AI in social good.
Conclusion
The use of agentic AI powered by Claude has enabled Bonterra to move from isolated AI features to a scalable, portfolio-wide capability. It automates and simplifies high-effort nonprofit workflows, improves access to data-driven fundraising insights, and supports more personalised donor engagement. The strongest documented outcome is 44% higher donation totals among Que users compared with non-Que users. By combining shared infrastructure, embedded workflows, guardrails and mandatory human approval, the platform also enables faster rollout of new AI capabilities without compromising accountability and trust.
