How OOTI Turned Financial Data into Natural Conversations with AI

Introduction

OOTI is a specialized ERP for architecture firms, bringing together project management, financial tracking, and resource planning. Their users are architects, not analysts, yet the data they manage is rich and detailed.

To make that data more accessible, OOTI partnered with Rakam to build a conversational AI assistant. This assistant lets users ask natural-language questions about margins, progress, or billing, and get precise, contextual answers from their ERP data.

The Challenge

Architects rely on OOTI to manage time, budgets, and billing across client projects. But making sense of this data often required navigating dashboards, running reports, or interpreting spreadsheets.

They faced key usability challenges:

• Users needed faster access to project data without relying on finance teams

• The system had to adapt to industry-specific language and workflows

• Answers had to be clear, reliable, and context-aware

• The assistant needed to handle sensitive financial information with care

"Architects don’t speak in filters and tables. They ask questions. And we wanted to answer them, instantly, and in their own words."

The Vision

The goal was to let any user, regardless of technical or financial training, query their data naturally, and trust the responses.

OOTI wanted to:

• Make project data navigable through conversation, not interfaces

• Align responses with financial reality and terminology

• Avoid misleading outputs with strong safeguards

• Offer visibility into how answers were sourced and calculated

The Solution

Working with OOTI’s team, Rakam designed a conversational agent that connects directly to live ERP data and responds using clear, structured business logic, wrapped in plain language.

1. Conversational Access to Financial Metrics

Users can ask questions like "How much margin do we have left on Project A?" or "What’s the status of unpaid invoices this month?" and receive detailed answers.

• The AI understands project codes, phase structures, and budget lines

• It returns clear responses with supporting figures and explanations

• Filters and breakdowns are applied automatically based on query context

"You don’t need to export a report or find a dashboard. Just ask."

2. Language Tuned for Architecture Practices

The assistant uses domain-specific vocabulary that matches how architects think about time, cost, and delivery.

• Recognizes terms like “fees,” “man-hours,” “pre-studies,” and “AVP”

• Adjusts phrasing based on user role (manager, partner, staff)

• Can handle embedded references to timelines, clients, or project types

"We trained the system not just on data, but on the way our users talk."

3. Built-in Controls for Compliance and Accuracy

Given the financial nature of the data, strong safeguards were essential.

• Sources are always cited, with links to original tables

• Sensitive terms trigger verification or clarification prompts

• A human-in-the-loop system allows for review where needed
  
  

The Outcome

The assistant now lives inside OOTI’s ERP, helping firms interact with their data as easily as asking a colleague a question. It saves time, boosts clarity, and makes project oversight accessible to the whole team, not just those with reporting skills.

• Teams get faster insights, fewer report requests, and better oversight

• Decision-making is closer to real time, even on the go

• Users feel more empowered and in control of project health

"We made data speak the way our users do, and that changed the relationship they have with it."