Ask Altrinum and AI

Why this area exists#

AI access in Altrinum is controlled at the tenant level. The AI section in Workspace Settings decides whether the workspace can use Ask Altrinum and shows the shared credit counters currently tracked for the tenant.

Confirmed settings and fields:

• Enable Ask Altrinum
• Enable AI Ask Optimization
• Maximum Ask Adjustment (%)
• Minimum Prior Gifts
• Current AI Credit Balance
• Lifetime AI Credits Purchased
• Lifetime AI Credits Used

The Giving Form editor now uses these values as workspace defaults inside the Gift options step. Individual giving forms can inherit them as-is or override them locally when a specific campaign should behave differently.

What the toggle changes#

Turning on Enable Ask Altrinum does more than save a preference.

Confirmed effects:

• the tenant is considered AI-enabled
• the Ask Altrinum navigation item can appear for the workspace
• the Ask Altrinum page checks this tenant setting before it allows use

If the setting is off, the workspace should expect Ask Altrinum to stay unavailable or show an AI-disabled message.

What Ask Altrinum is for right now#

The Ask Altrinum page is oriented toward straightforward analytics and planning questions in these areas:

• donations
• events
• programs
• emails
• prospects

The page includes its own usage guidance and example prompts once the feature is enabled.

Ask Altrinum can also start contextual creation flows. Supported creation flows include donation form drafting and event drafting from natural-language requests. Create email mode produces assistive fundraising copy from a prompt, but does not save or send a campaign. Draft records use workspace context already available in Altrinum.

Creation is intentionally conservative. Ask Altrinum may ask for required missing details, and AI-created donation forms are saved as drafts only with status=draft and is_active=false. It cannot publish the form, activate the public page, create designations automatically, or bypass review.

For designation handling, Ask Altrinum matches only active designations inside the current workspace. A clearly named designation/fund in the prompt can be matched by name or slug. One active safe default can be used as the recommended default. Multiple active designations trigger a short choice question instead of an open-ended fund prompt.

For content, Ask Altrinum uses AI-generated copy after deterministic creation rules decide a donation form draft can be saved. It can polish the title, intro copy, giving copy, thank-you copy, and optional goal or matching copy when requested or clearly implied. It must not invent statistics, beneficiary counts, tax/legal claims, matching sponsors, match rates, or deadlines. Tax-related language should stay cautious unless stronger wording is grounded elsewhere. If AI copy generation is unavailable or invalid, Ask Altrinum falls back to safe starter copy and records that fallback in internal metadata.

If Ask Altrinum asks a donation-form setup question, the next answer continues the current drafting session. This keeps a designation or title answer from being treated as a brand-new reporting question.

Event drafting uses tenant branding, tenant timezone, recent event titles and descriptions, common event types, common locations, recent registration structures, waitlist settings, ticket patterns, and recent custom registration questions when those records exist.

Event creation is also conservative. Ask Altrinum asks for required missing details such as date, start time, location or online format, and whether registration is free, paid, or invitation-only. AI-created events are saved with status=draft and visibility=hidden; public event routes should not expose them until staff review and publish manually.

For event content, Ask Altrinum uses AI-generated copy for starter descriptions, registration CTA copy, confirmation copy, and online-event notes after deterministic event setup rules pass. It must not invent addresses, speakers, sponsors, ticket prices, tax/receipt claims, capacity limits, deadlines, accessibility claims, food details, or parking details. If AI copy generation is unavailable or invalid, Ask Altrinum falls back to safe starter event copy and records that fallback in internal metadata.

Create email mode uses safe defaults when staff provide a natural-language prompt: it infers audience and tone where possible, otherwise uses general supporters and a warm tone. The result is an assistive draft only with subject line options, preview text options, body copy, and CTA suggestions. It is not saved as a campaign, not scheduled, and not sent.

Event announcement email creation from an event draft is a review-time next action. Ask Altrinum can suggest it, but event prompts do not create email campaigns automatically.

What AI Ask Optimization changes#

When Enable AI Ask Optimization is turned on, Altrinum may adjust public donation suggested amounts for eligible logged-in donors.

Confirmed runtime behavior in this pass:

• baseline suggested amounts still start from the giving form amount_options
• runtime settings resolve from the giving form first when that form disables inheritance and saves local AI Ask Optimization values
• otherwise runtime settings fall back to the workspace defaults in Workspace Settings → AI
• anonymous visitors keep the baseline amounts unchanged
• ineligible logged-in donors keep the baseline amounts unchanged
• eligible logged-in donors can receive adjusted suggested amounts using the persisted tenant_constituent_prospect_scores.explanation_factors_json.ask_optimization recommendation
• Maximum Ask Adjustment (%) hard-limits how far adjusted amounts can move away from the giving form baseline
• Minimum Prior Gifts requires a minimum completed giving history before adjusted suggested amounts can be shown

If the feature is off, or if the runtime data is missing or uncertain, the public donation flow falls back to the form-configured baseline amounts.

Donation attribution for AI Ask Optimization#

When an optimized public donation is created, Altrinum now stores a small donation-level attribution snapshot in donations.ai_ask_optimization_attribution_json.

This snapshot is computed server-side at donation creation time. It does not come from Blade state or the request payload.

The stored payload is intentionally focused and currently includes:

• whether optimization was applied
• the resolved tenant_user_id
• the identity source used for resolution such as authenticated_member
• the prospect score row used for the decision
• the persisted ask recommendation values used by the resolver
• the runtime limits used by the resolver such as maximum adjustment and minimum prior gifts
• the completed gift count checked at runtime
• the baseline and adjusted suggested amount values used for that donation

If optimization is not applied, the donation keeps ai_ask_optimization_attribution_json empty.

How attribution is used for reporting#

This donation-level attribution exists so later reporting can measure AI Ask Optimization against actual gifts instead of only against prospect score recommendations.

That makes it possible for future reporting work to answer questions such as:

• how many donations were created while optimization was applied
• which strategies or confidence levels were most common on completed gifts
• whether optimized donations trended differently from non-optimized donations
• which giving forms or source codes were involved when optimized asks were shown

Because the stored payload matches the resolver output, reporting can use the same normalized context that the public donation flow actually used.

What current reporting can and cannot measure#

Current AI Ask Optimization reporting is intentionally conservative.

It can measure:

• completed donations with stored optimization attribution
• completed donations without stored optimization attribution
• optimized donation revenue by giving form or tracking source
• optimized donation breakdowns by stored strategy or confidence

It cannot measure:

• who saw optimized suggested amounts but did not donate
• impression or exposure counts
• conversion rates from optimization exposure to donation

Those questions would require exposure tracking that is not part of the current design.

Usage notes visible in the implementation#

• The settings page does not expose AI provider setup for tenant admins here.
• The credit counters are read-only in this tenant-facing settings page.
• The AI safety modal explains that Prospect Intelligence and Ask Altrinum stay inside the workspace data boundary and that OpenAI API processing is used for some AI work.
• Public donation ask optimization now includes donation-level attribution persistence when the optimization is actually applied.

Operational notes#

• This is an organization-wide decision. Turning the feature on changes the workspace, not just one user.
• Staff should test the feature with a simple question after enabling it rather than waiting for a high-pressure reporting moment.
• If your team uses Prospect Intelligence questions in Ask Altrinum, make sure prospect scores have been refreshed first.

Common mistakes#

• enabling AI and not verifying that the Ask Altrinum page now appears
• assuming the settings page also manages model or provider selection
• ignoring the shared nature of the displayed credit balance
• asking prospect questions before Prospect Intelligence has been refreshed

What to test afterward#

• confirm the Ask Altrinum navigation item appears
• open the page successfully
• run one simple question
• check the visible shared credit counters

Related topics#

• Prospect Intelligence
• Ask Altrinum Tutorial
• Ask Altrinum How-To