Chapter 2: Chatbots: Tasks and tips

Introduction

This chapter introduces practical selection and use of AI chatbots and image generators for everyday individual IT work.

• Task-based AI tool selection: Explain why choosing an AI assistant by the specific IT task matters, and describe the chapter’s use-case-based approach rather than treating all AI tools as interchangeable.
  Why it matters: Matching a tool to the job helps you get more useful results in day-to-day IT work.
• Individual IT AI capabilities: Describe the kinds of day-to-day IT assistance the chapter will examine, including web searches, error-message analysis, image analysis, voice commands, task automation, and AI image generation.
  Why it matters: These capabilities show the range of practical work that AI tools can support for an individual IT professional.
• Practical scope and limitations: State that the chapter considers both what current AI tools can and cannot do well for individual IT professionals, while leaving organization-specific security and compliance decisions outside its main scope.
  Why it matters: Keeping the scope clear helps you apply the examples realistically without assuming one tool choice fits every organization.

Key terms: AI assistants, AI chatbots, use cases, web searches, analyzing images, voice commands, automating tasks, AI image generator, conversational models, security

2.1 Chatbots and conversational AI

This section frames the evaluation of AI chatbots by contrasting benchmark rankings with real-world performance and previewing the practical criteria used in the following reviews.

• Benchmark and real-world evaluation: Distinguish benchmark-based chatbot evaluation from day-to-day evaluation, explaining why a strong benchmark ranking is only a performance snapshot rather than the complete test of a model.
  Why it matters: Benchmarks can help compare models, but practical work reveals whether a chatbot is useful when you need it.
• Practical chatbot assessment criteria: State that the upcoming hands-on chatbot discussions focus on reliability, feature sets, and practical use cases for IT professionals.
  Why it matters: These criteria give you a practical basis for choosing among chatbots instead of relying only on broad popularity or rankings.
• Variable model performance: Describe the section's observation that even highly rated models can have occasional off days because of high usage or resource constraints, while generally still producing impressive results.
  Why it matters: It helps you interpret an unexpectedly weak response as possible variability rather than assuming a model's usual performance is always identical.

Key terms: benchmarks, Hugging Face, lmarena.ai, Chatbot Arena, day-to-day scenarios, reliability, feature sets, practical use cases, IT professionals, high usage

2.1.1 OpenAI ChatGPT

This section evaluates ChatGPT’s practical IT capabilities and shows when to use its web search, vision, voice, custom GPTs, and advanced models.

• ChatGPT web research: Explain how ChatGPT’s SearchGPT combines live web results with its knowledge, provides citations, and helps IT professionals obtain current information.
  Why it matters: Current, citable research can prevent work based on outdated packages, APIs, rates, or documentation.
• Focused search verification: Describe why requests to verify many dependencies or facts should be split into small groups, such as three packages at a time, when using ChatGPT web search.
  Why it matters: Smaller verification requests reduce the chance that errors will slip through when each item needs separate searching and checking.
• Vision task workflow: Describe ChatGPT’s image-analysis uses for IT work and explain why separating a visual task into stages, such as transcription followed by translation, can improve results.
  Why it matters: Vision analysis can turn screenshots and other difficult-to-read visuals into useful technical information more efficiently.
• Voice interaction tradeoffs: Explain the benefits and limitations of ChatGPT’s paid voice conversations, including hands-free, web-enabled assistance, language support, usage limits, and interruption problems.
  Why it matters: Knowing the tradeoffs helps you use voice assistance safely and realistically during mobile or multilingual work.
• Custom GPT specialization: Distinguish custom GPTs for repeated, environment-specific routine tasks from regular ChatGPT for novel or complex problems that require broader flexibility.
  Why it matters: Choosing the right kind of assistant avoids repeatedly supplying context while preserving flexibility for unpredictable production issues.
• Advanced model capabilities: Describe the role of paid advanced models such as o1 and o1 mini for complex technical problem-solving, and state the practical tradeoffs in ChatGPT’s overall offering.
  Why it matters: Matching difficult work to advanced models and account capabilities helps set realistic expectations about cost, context size, and model strengths.

Key terms: SearchGPT, web search, citations, dependencies, verification, vision tasks, visual analysis, transcription, translation, advanced voice

2.1.2 Anthropic Claude

This section explains why Claude is especially useful for context-aware writing, software architecture, and computer-control tasks, while outlining its usage limits and trade-offs.

• Context-aware technical writing: Explain how Claude supports advanced, natural technical and professional writing by producing precise content that accounts for context, rationale, and the intended audience.
  Why it matters: This helps you select Claude when clear, nuanced documentation matters as much as technical correctness.
• Custom writing style adaptation: Describe how to establish a custom Claude writing style by supplying company documentation, style guidance, or multiple examples, and explain how this supports consistency across future documents.
  Why it matters: A reusable style lets a team produce consistent documentation even when several people contribute to it.
• Architecture-aware solution design: Explain how Claude can support software architecture rather than merely generate a code change, including analyzing existing patterns, considering security and downstream effects, and designing for future requirements.
  Why it matters: Considering system-wide consequences and future integrations can prevent costly redesigns after a seemingly small feature change.
• Computer Use automation trade-offs: Describe Claude's Docker-based Computer Use capability as natural-language-driven interaction with a system, give an example of a multi-step workflow it can automate, and explain when script generation may be the more cost-effective choice.
  Why it matters: You can weigh powerful UI automation against the security, speed, and API-cost implications before granting an AI system access.
• Claude plans and limitations: State the practical differences between Claude's free and Pro tiers and summarize key strengths and limitations, including constrained usage, missing features, and the need to monitor remaining quota.
  Why it matters: Knowing Claude's limits helps you choose an appropriate plan and avoid running out of capacity during a complex task.

Key terms: context-aware analysis, technical documentation, Concise style, Explanatory style, Formal style, custom styles, style guides, writing requirements, software architecture, authentication patterns

2.1.3 Google Gemini

This section assesses Google Gemini’s multimodal capabilities, Google integrations, and limitations, concluding that inconsistent performance makes it hard to trust for professional work.

• Gemini multimodal capabilities: Describe Gemini’s ability to process images, videos, text, and code, and identify notable capabilities such as visual generation, multilingual support, real-time AP news, and custom models on paid plans.
  Why it matters: Gemini’s broad feature set helps you recognize when it may be a useful fit for multimodal or Google-centered tasks.
• Professional reliability concerns: Explain why the authors found Gemini difficult to trust for professional work: its performance was hit or miss and it could give incorrect or confusing responses, including seeming to forget capabilities it has.
  Why it matters: A tool’s advertised features are less valuable in IT work if its behavior is not consistent enough to rely on.
• Gemini search and Android integration: Describe how Gemini can search Google’s index and summarize current results within a chat, and explain its role as the default Android assistant on eligible Android 10+ devices with at least 2 GB of RAM, replacing Google Assistant while allowing users to switch back.
  Why it matters: These integrations can make Gemini convenient for current-information requests and Android use without moving among separate apps.
• Gemini coding tradeoffs: Distinguish Gemini’s potential advantage for large codebases through its large context window from the authors’ view that it is not among the best AI coding assistants and is weaker than specialized tools for writing and complex thinking.
  Why it matters: Matching an AI tool to a task means weighing a useful strength, such as large-context code handling, against its limitations relative to specialized assistants.

Key terms: Gemini, multimodal, custom AI models, real-time news, Associated Press, inconsistent performance, hit or miss, professional work, capabilities, Google Search

2.1.4 Microsoft Copilot suite

This section evaluates Microsoft Copilot products, emphasizing Azure Copilot’s integrated troubleshooting strengths while identifying the focused scope, uneven maturity, and licensing limits of the wider suite.

• Azure Copilot troubleshooting integration: Explain how Azure Copilot assists with Azure errors directly in the Azure interface. Describe how it uses the error context and Azure environment to confirm the affected resource and suggest resolution paths.
  Why it matters: Azure Copilot can reduce the manual effort of interpreting detailed cloud error messages and searching documentation.
• Microsoft 365 information retrieval: Describe the main business-oriented role of Microsoft 365 Copilot: finding and organizing company documents and information across Microsoft applications such as Teams and SharePoint. Distinguish this ecosystem-focused capability from the broader complex-task versatility attributed to ChatGPT and Claude.
  Why it matters: Matching the tool to company information and workflow tasks helps set realistic expectations for Microsoft 365 Copilot.
• Copilot scope and limitations: Explain why individual Microsoft Copilots are focused on narrow tasks within their host applications, and describe the trade-off: focus can support intended-task performance and control company AI costs, but can also cause rejected requests and limited customization. State that performance and maturity are inconsistent across applications.
  Why it matters: Knowing that each Copilot is application-specific prevents overreliance and helps you choose a standalone chatbot when broader capability is needed.
• Data-analysis Copilot capabilities: State the useful data-oriented Microsoft Copilot capabilities highlighted in the section: Python-powered analysis and automated reporting in Excel and Power BI, with automated Power BI reports requiring Premium capacity. Describe their Premium licensing requirement and the recommendation to explore them for enterprise data-analysis work.
  Why it matters: Excel and Power BI are among the more promising Microsoft Copilot options when an organization has the required access and a data-analysis need.
• Context-aware GitHub Copilot chat: Describe GitHub Copilot chat as an example of seamless, context-aware AI integration: it appears on GitHub pages and changes according to what is onscreen to help with repository searches, code understanding, and documentation.
  Why it matters: Context-sensitive assistance illustrates a more intuitive integration model that other platforms may adopt.

Key terms: Azure Copilot, Ask Copilot, operation IDs, error context, affected resource, resolution paths, Microsoft 365 Copilot, Microsoft ecosystem, Teams, SharePoint

2.1.5 Summary comparison of AI assistants

This section compares the assistants by task and emphasizes choosing among them as their fast-changing features evolve.

• Task-based assistant selection: Explain how the section matches ChatGPT, Gemini, Claude, and Copilot to different kinds of work, while recognizing that the best choice depends on the task.
  Why it matters: Matching an assistant to the job helps you use its relative strengths rather than treating every chatbot as interchangeable.
• Assistant feature volatility: State that the comparison table is only a snapshot because AI assistant features and capabilities change often.
  Why it matters: A current comparison is useful, but you should recheck capabilities rather than assuming a tool will remain unchanged.

Key terms: ChatGPT, Gemini, Claude, Copilot, Microsoft-centric environments, Table 2.3, AI assistants

2.2 Text-to-image AI models

This section introduces text-to-image AI as a sometimes-useful but currently limited way to create visuals from written descriptions.

• Text-to-image model purpose: Explain that text-to-image AI models generate visual images from text descriptions, and describe their limited but potentially useful role in everyday IT work.
  Why it matters: This framing helps you identify when image generation may be useful without treating it as a central IT tool.
• Current image generator limitations: Describe the main current limitations of text-to-image generators: unreliable prompt adherence, edits that often produce new images instead of modifying the original, and refusals that may claim a requested output was produced.
  Why it matters: Knowing these limitations helps you set realistic expectations and avoid relying on generated images for exact or repeatable results.
• Conditional image generation value: State that, despite their significant current shortcomings, text-to-image generators can produce impressive results when they work.
  Why it matters: This balanced view supports experimentation while preserving appropriate caution about the output.

Key terms: text-to-image AI models, text descriptions, visuals, prompt adherence, edits, original image, refuse, impressive results, limitations

2.2.1 OpenAI DALL-E and 4o

This subsection introduces OpenAI’s DALL-E image generation, illustrates its use for a simple logo, and notes its integration into GPT-4o and related tools.

• DALL-E text-to-image generation: Explain that DALL-E is OpenAI’s image-generation tool and that it converts text descriptions into sophisticated images. Describe the kinds of work it can support, including visual design and the authors’ use of simple logos and diagrams.
  Why it matters: Text-to-image generation lets IT professionals create useful visual assets from a written description.
• Logo prompt outcome: Describe the logo-generation example: the prompt requested a simple relaxed, relieved-looking cloud with a simple yellow sun in the background, and DALL-E closely followed the request. State that the resulting cloud logo is shown in Figure 2.9.
  Why it matters: The example shows that a concrete visual description can produce a usable image even when expectations are low.
• GPT-4o image integration: State that DALL-E was easy to use through OpenAI apps and the web interface despite not being best-in-class at the time. Explain that image generation is now built directly into GPT-4o, making it faster, more accurate, and integrated with tools such as Microsoft Copilot.
  Why it matters: Knowing where image generation is integrated helps readers choose accessible ways to create images in their existing AI workflows.

Key terms: DALL-E, OpenAI, image generation, text descriptions, prompt, relieved emoji, cloud, yellow sun, Figure 2.9, GPT-4o

2.2.2 Midjourney

This section presents Midjourney as a precise, artist-oriented image generator and illustrates its strengths with a “beautiful API” image.

• Midjourney visual strengths: Describe Midjourney as a generator known for artistic and hyperrealistic images, with detailed controls for visual style, composition, aspect ratio, and upscaling.
  Why it matters: These capabilities help you recognize when Midjourney is a strong fit for image work requiring creativity and visual precision.
• Beautiful API example: Explain how Midjourney was used for a presentation called “Designing Beautiful APIs,” producing an imagined API architecture that combined organic forms with clean, logical lines of connectivity and functionality.
  Why it matters: The example shows how an abstract technical idea can be expressed as a visually striking image.
• Subscription cost assessment: State that Midjourney costs $10 per month and explain that the authors normally did not use AI art often enough to justify that ongoing cost, although reactivating it for this section reminded them of its quality before OpenAI’s 4o.
  Why it matters: This illustrates that a tool's quality must be weighed against how frequently you will actually use it.

Key terms: Midjourney, artistic images, hyperrealistic images, visual styles, composition, aspect ratios, Designing Beautiful APIs, API architecture, connectivity, functionality

2.2.3 Google Gemini Imagen and Meta Imagine

This subsection briefly identifies Google Gemini Imagen and Meta Imagine as accessible text-to-image models and notes Meta Imagine’s chat integration.

• Gemini Imagen and Meta Imagine: Describe Google Gemini Imagen and Meta Imagine as text-to-image models available through Google Gemini and Meta’s respective tool suites, and state that both offer surprisingly good output quality.
  Why it matters: Knowing the available image-generation options helps you recognize which AI tools may fit a task.
• Meta Imagine chat integration: State that Meta Imagine is conveniently integrated into Facebook and Instagram chats.
  Why it matters: Its placement in familiar chat applications can affect when and how easily you can use the tool.
• Business-use experience limitation: State that the authors have not yet used Google Gemini Imagen or Meta Imagine for business purposes.
  Why it matters: This distinguishes the tools' reported accessibility and output quality from firsthand business-use experience in the book.

Key terms: Google Gemini Imagen, Meta Imagine, Google Gemini, Meta suite of tools, Facebook, Instagram, chats, business purposes

2.3 Prompts used in this chapter

This section records example prompts used to write and edit the chapter across several AI assistants.

• Chapter-authoring prompt examples: Describe the kinds of chapter-writing and editing tasks the authors gave to AI assistants: researching a feature, refining an explanation, planning screenshot placement, and editing document capitalization.
  Why it matters: These examples show that AI assistants can support both research and practical editorial work while producing a technical chapter.
• Assistant-specific prompt assignments: State which assistants were used for the listed tasks: ChatGPT for the Claude code-interpreter question, Claude for wording and Copilot-screenshot questions, and Cline or Roo for the capitalization edit.
  Why it matters: Matching an example task to its assistant makes the prompt record concrete and reproducible.
• Capitalization edit cost: State that the Cline/Roo task to find Title Case instances and convert them to sentence case cost $0.30.
  Why it matters: The stated cost gives a concrete example of the expense of a small AI-assisted editing task.

Key terms: ChatGPT, Claude, Cline, Roo, code interpreter, system prompts, Copilot, chapter02.adoc, Title Case, sentence case

2.4 Summary

This summary compares major chatbot and image-generation tools so readers can choose based on required capabilities, acceptable output quality, and budget.