Otto: AI Consulting Agent Reshapes Enterprise Automation Requirements Analysis

Otto is an AI consulting agent that interviews stakeholders through multi-modal methods such as voice interaction and screen sharing, automatically maps workflows, and generates automation implementation plans. It aims to solve pain points in enterprise automation projects like deviations in requirement understanding, high documentation costs, and knowledge loss, helping enterprises with digital transformation.

Original author/maintainer: mwzhu Source platform: GitHub Original link: https://github.com/mwzhu/otto Source release time/update time: 2026-06-10T23:45:36Z

In enterprise digital transformation, the failure rate of automation projects remains high. Research shows that over 60% of RPA projects fail to achieve expected results, with the core reason being "deviations in requirement understanding".

Pain points of traditional requirement collection:

• Business-IT gap: Business personnel find it hard to describe process details, while IT personnel lack scenario understanding
• High documentation cost: Manual sorting of interview records and drawing flowcharts is time-consuming and labor-intensive
• Risk of knowledge loss: Key business knowledge relies on "key personnel", and staff turnover easily leads to gaps
• Frequent requirement changes: Static documents struggle to adapt to the fast-changing market environment

Otto targets these pain points and proposes a new paradigm of requirement discovery where AI directly "enters" the business scene.

Otto integrates three perceptual modalities to achieve comprehensive scenario understanding:

Voice Interaction

• Context memory: Maintains coherence in multi-turn dialogues and links previous concepts
• Follow-up guidance: Automatically clarifies ambiguous answers to dig deep into underlying needs
• Multi-language support: Cross-language teams communicate in their native languages, and the system organizes uniformly
• Emotion perception: Identifies hesitation and emphasis to capture implied meanings

Screen Sharing

• Operation sequence: Recognizes system switches and execution actions
• Interface elements: Captures interactive components like form fields and buttons
• Data flow: Tracks information acquisition and transfer paths
• Exception handling: Records error responses and waiting states

Process Mapping

Automatically generates structured documents: flowcharts, role matrices, data dictionaries, and pain point annotations.

Multimodal Perception Layer

• Automatic Speech Recognition (ASR): Supports accent adaptation and domain terminology
• Natural Language Understanding (NLU): Extracts entities, relationships, and intents to build semantic graphs
• Computer Vision (CV): Recognizes screen UI elements and text
• Temporal modeling: Understands operation sequences and identifies loops and conditional branches

Knowledge Representation and Reasoning

• Process ontology: Defines core BPMN concepts (activities, gateways, etc.)
• Constraint rules: Records business rules and compliance requirements
• Confidence scoring: Prompts manual confirmation for low-confidence content

Automation Solution Generation

• Technical selection recommendations: RPA, API integration, low-code, etc.
• Workload estimation: Predicts cycles based on historical data
• ROI analysis: Quantifies expected benefits
• Risk list: Identifies technical difficulties, dependency risks, etc.

Scenario 1: AS-IS Analysis Before New System Launch

Replaces traditional consulting firms, completes multi-department interviews within days, generates process documents, and shortens project cycles.

Scenario 2: Rapid Evaluation of RPA Projects

Remotely connects to the customer environment, quickly identifies automation opportunities, generates proposals, and reduces costs by over 80%.

Scenario 3: Knowledge Inheritance and Training

Records the experience of senior employees, converts it into training materials for new employees, and mitigates knowledge loss.

Scenario 4: Continuous Process Optimization

Acts as a "resident consultant", regularly interviews to track process changes, and discovers optimization opportunities.

• Minimal collection principle: Only captures process-related interfaces and blurs sensitive information
• Real-time authorization: Each screen sharing requires explicit user consent and can be interrupted at any time
• Data localization: Original recordings and screenshots are processed locally and not uploaded to the cloud
• Audit logs: Fully records collection scope and processing steps to meet compliance audits.

Dimension	Traditional Interview + Documentation	Otto AI Agent
Time Cost	Weeks	Days
Accuracy	Depends on interviewer's experience	AI standardized processing
Tacit Knowledge Capture	Difficult	Achieved via screen sharing
Scalability	Limited by manpower	Can handle multiple projects in parallel
Cost	High (consulting fees)	Low (software license)

Otto represents the deep application trend of AI in the enterprise service field. Future versions may implement:

• Real-time collaboration: Multiple people participate in interviews, and AI coordinates discussions to integrate opinions
• Simulation verification: Automatically generates test cases to verify the accuracy of process understanding
• Automatic implementation: Deeply integrates with RPA tools to automate the entire process from requirements to deployment

Otto provides enterprises with a low-threshold, high-efficiency requirement discovery solution to accelerate digital transformation.