Aira: Global Services AI Risk Assessor

This document outlines the process of presenting a prototype for Aira, an AI-driven risk assessor designed for global services. The presentation highlights real-world challenges faced by implementation teams and how Aira can aid in assessing risks in a timely manner.

Step 1

Greetings, team and Todd. This is Devika. I was one of the finalists in the Zebra Strup challenge, and I am here to present my prototype on Aira, the global services AI risk assessor. As a principal SDM in the global services team, assessing human risk is a key aspect of my daily work.

Step 2

Given my experience, it was fitting to choose this topic for the prototype presentation. Global services implementation teams encounter real-world challenges daily as they interact with customers, executive sponsors, C-suite executives, vice presidents, and directors. These projects are increasingly enterprise-scale, involving multiple workstreams and vendors.

Step 3

For example, in FY26 active projects, costs have increased significantly beyond original SOW estimates. Approximately 14,000 investment hours have been dedicated to completing projects and satisfying customers. Statistically, this has led to over three months of average go-live delays. Currently, half of the active projects in GS Noam are flagged as yellow or red.

Step 4

The critical question is whether we are effectively identifying risks or allowing them to escalate, contributing to these metrics. Can AI offer enhanced risk assessment capabilities? This is where Aira can be instrumental.

Step 5

Aira is intended as a real-time risk assessor. In this prototype, I have selected two key risk aspects encountered in my role as an SDM. The first, crucial area is solution risk. From the onset of a customer journey—through requirements gathering, workshops, RTM, SDD, testing, and launch—customers frequently raise questions regarding requirements.

Step 6

Whether dealing with a comprehensive BRD, RTM, or ad-hoc inquiries, solution and product risks are present at each step. The time required to assess these risks and implement mitigation steps is substantial, often extending from days to weeks. The question is whether the real-time assessor can reduce this timeframe to mere minutes.

Step 7

The second aspect of risk involves customer sentiment concerning budget, timeline, and scope. Is it possible for the assessor to build a risk dashboard? My focus is on these two risk types, with particular emphasis on solution risk. The prototype will demonstrate how it can assess solution risks by evaluating customer business requirements using an agent like Glean, starting with verified data sources.

Step 8

Rather than relying on publicly accessible knowledge centers or internet-based sources, this process utilizes verified internal sources from Zuora to provide a comprehensive fit-gap assessment. It involves data from Zuora KCs, Google Drive, Zendesk tickets, and Salesforce COCs.

Step 9

Additionally, it examines the Jira roadmap for P&D to thoroughly assess fit-gap against specific business requirements or RTMs, ultimately providing a risk score from a technical assessment perspective. This process, which traditionally requires days of coordination with various stakeholders, aims to be completed in minutes.

Step 10

The goal is to provide real-time assessments when an SA is addressing customer inquiries. For instance, when faced with questions like, "I asked ChatGPT if Zuora can handle XML invoicing and it said yes, can you confirm?" this prototype can utilize verified internal sources to deliver holistic answers, transforming architects into risk assessors, not just solution designers. Let's explore the prototype, which features two developed agents.

Step 11

As we transition from prototype to production-ready stages, we plan to develop a comprehensive risk assessor. At a high level, I'll walk through the technical risk prototype and the customer sentiment risk dashboard. Let's examine Aira, the technical risk assessor.

Step 12

While Aira operates, I'll explain the thought process and planning involved in understanding the agent's capabilities.

Step 13

We will select a customer BRD, a Zuora RTM, or an ad-hoc requirement for demonstration.

Step 14

For this demonstration, I will choose a BRD from an actual customer and proceed with the evaluation.