Generative AI for Business Productivity

What generative AI can realistically do today

Illustrative workplace use cases rather than unsupported claims about universal adoption.

LLMs, multimodal AI, AI assistants, and AI agents

Productivity versus automation versus augmentation

Common misconceptions and practical limitations

ChatGPT and ChatGPT Enterprise

• Capabilities and limitations of the models available in the actual training account; avoid fixed release-number prerequisites.
• Deep research
• Data analysis
• File uploads
• Projects and selected agent workflows, subject to the assigned seat and workspace controls.
• Review applicable enterprise privacy and administration settings; do not assume personal and organisational accounts have the same protections.

Anthropic Claude

• Available Claude models and task suitability; model releases and limits vary.
• Claude for professional work
• Claude for long-context analysis
• Claude for writing, summarisation, reasoning, and coding support
• Claude Tag and Slack collaboration examples where an organisation has enabled and configured access.
• Claude Enterprise and Team use cases

Google Gemini

• Workspace productivity use cases
• Research and summarisation workflows
• Multimodal interaction

Microsoft Copilot

• Microsoft 365 productivity workflows
• Word, Excel, PowerPoint, Outlook, and Teams scenarios
• Business user adoption considerations

Choosing the Right AI Tool

• ChatGPT versus Claude
• Claude versus Gemini
• Copilot versus standalone AI assistants
• Tool selection by business task
• Compare cost, permissions, privacy, output quality and usability using a defined task.

Claude Workflow Examples

Evaluating Claude for a Business Task

• Long-form writing
• Large document review
• Policy-style drafting support; specialist legal review is still required where appropriate.
• Evaluate reasoning-related outputs against task evidence rather than relying on confident language.
• Summarisation with comparison to the original source.
• Natural business communication

Task-Based Comparison with ChatGPT

• Compare writing style on the same prompt and audience.
• Compare supported conclusions and errors, not an assumed universal reasoning ranking.
• File and document workflows
• Enterprise collaboration scenarios
• Practical selection guidelines

Claude and modern workplace collaboration

• Claude/Claude Tag in Slack: optional demonstration of configured access, channel context and organisational permissions.
• Team-based AI interaction
• Context-aware assistance and the risks of unintended information exposure.
• Task tracking and summarisation
• Review owner-configured access controls and activity visibility.

Prompting fundamentals

• Role
• Task
• Context
• Format
• Constraints
• Examples
• Evaluation criteria

Prompting for better outputs

• Clear instruction design
• Structured response formats
• Iterative refinement
• Tone and audience control
• Asking for assumptions and limitations
• Reducing vague and generic responses

Business prompt patterns

• Executive summaries
• Email drafting
• Proposal support
• Meeting preparation
• Report writing
• Policy drafting
• Customer response drafting
• Brainstorming and ideation
• Decision comparison
• Risk identification

Writing and communication

• Professional emails
• Reports and memos
• Presentations
• Meeting notes
• Minutes of meeting
• Executive updates
• Customer-facing communication

Research and analysis

• Topic exploration
• Market scanning
• Competitor review
• Document summarisation
• Long-document analysis
• Source checking
• Insight extraction

Data and spreadsheet support

• Data cleaning concepts
• Formula assistance
• Trend interpretation
• Chart explanation
• KPI analysis
• Management reporting

Creative and strategic work

• Campaign ideas
• Product positioning
• Training content
• Internal communication
• Scenario planning
• Problem framing

Complete two small guided tasks; use demonstrations or discussion for the remaining examples. Check facts, calculations and suitability before using an output.

Understanding AI agents

• Assistant versus agent
• Task delegation
• Tool use
• Multi-step workflows
• Human approval points

Practical agent use cases

• Research agent
• Report preparation agent
• Meeting follow-up agent
• Customer support assistant
• Data analysis assistant
• Internal knowledge assistant

Model Context Protocol and connected AI

• MCP as a protocol for connecting AI applications with tools and context; it is not itself an autonomous agent or an access-control guarantee.
• Why tool connectivity matters
• AI access to business systems
• Permission boundaries, authentication and review of connected-data access.
• Enterprise integration considerations

Retrieval-augmented generation (RAG): retrieving relevant information to support an answer, with source, access and quality checks; retrieval does not guarantee correctness.

Accuracy and verification

• Hallucinations
• Fact-checking
• Source validation
• Human review
• Check supporting evidence and uncertainty; a model’s stated confidence is not a verified probability.

Privacy and confidentiality

• Sensitive data handling
• Customer data
• Internal documents
• Enterprise policy alignment
• Data leakage risks

Governance and compliance

• Acceptable use policies
• Approval workflows
• Auditability
• Bias and fairness
• Copyright awareness
• Risk management

• Identify a suitable use case and a manageable first experiment.
• Create reusable prompt examples and simple team guidelines.
• Define a baseline, review quality and record actual time/effort rather than assuming a productivity gain.
• Assign approval and escalation responsibilities.
• Decide whether to continue, revise or stop the pilot after evaluating evidence.