Skip to main content
Xoxoday Plum delivers contextual AI optimisation through application-layer mechanisms — including prompt templates, admin controls, and usage analytics — without modifying or retraining the underlying foundational language models.
Xoxoday Plum takes a contextual optimisation approach to AI performance, working at the application layer rather than modifying foundational large language models. This means the AI behaviour visible in reward recommendations, nudges, and automated workflows is shaped entirely through configurations your team controls — not through model retraining cycles.

Prompt Design and Templates

Xoxoday Plum’s AI layer is guided by structured prompt templates that define how queries are framed and how responses are surfaced to end users. Admins can configure these templates to align with company tone, reward catalogue scope, and programme goals. Changes take effect immediately without any model downtime or engineering overhead.

Admin Controls and Business Rules

Business rules set at the admin level act as guardrails for AI-generated suggestions. If your organisation runs a performance recognition programme integrated with Darwinbox or SAP SuccessFactors, admins define which employee segments receive AI-driven reward nudges, which reward categories are prioritised, and how frequently suggestions appear. These controls ensure Xoxoday Plum behaves consistently with your HR policies and budget guidelines. Interaction patterns also feed back into the system over time. As employees engage with reward suggestions — accepting, dismissing, or modifying them — Xoxoday Plum refines future outputs based on what resonates.

Validation Through A/B Testing and Analytics

Xoxoday Plum validates AI performance through A/B testing rather than model retraining. Two variants of an AI-generated reward suggestion can be surfaced to different employee cohorts simultaneously, with engagement, redemption rates, and programme participation tracked as outcome signals. Usage analytics provide a continuous feedback loop, helping programme managers identify where suggestions land well and where prompt templates need adjustment. For organisations using Microsoft Teams or Slack for recognition workflows, analytics capture engagement data from those surfaces too — giving admins a fuller picture of AI recommendation effectiveness across channels.

A Practical Example

Consider an organisation running a quarterly sales incentive programme through Xoxoday Plum, integrated with Workday for headcount data. Xoxoday Plum surfaces personalised reward suggestions based on role, tenure, and past redemption behaviour. Admins use business rule controls to limit suggestions to a specific catalogue tier, while A/B testing runs in the background comparing two messaging approaches. At the end of the quarter, the analytics dashboard identifies which variant drove a higher redemption rate, and the winning prompt template is promoted — no model retraining involved. This application-layer model keeps your AI-powered rewards experience aligned with evolving business rules, workforce changes, and programme goals without any dependency on model update cycles.
Learn more: Xoxoday Plum Help Centre — AI LLM

How does Xoxoday Plum generate AI reward recommendations?

Learn how Xoxoday Plum’s AI engine surfaces personalised reward suggestions based on role, behaviour, and programme context.

How does Xoxoday Plum handle data privacy for AI features?

Understand how Xoxoday Plum protects employee data used by AI features, including compliance with ISO 27001 and SOC 2 Type II standards.