AI and Communication

Summary

AI is transforming how engineers communicate. This chapter covers large language models, AI writing assistants, and prompt engineering — teaching you to use AI as a productivity multiplier without replacing human judgment. You will learn about multimodal AI, AI-powered presentation tools, summarization, and translation. Critically, the chapter also addresses AI limitations, the ethics of AI-generated content, fact-checking AI output, and how to integrate AI tools into your communication workflow responsibly.

Concepts Covered

This chapter covers the following 16 concepts from the learning graph:

1. AI in Communication
2. Large Language Models
3. AI Writing Assistants
4. Prompt Engineering
5. AI-Generated Content
6. Human-AI Collaboration
7. AI Content Review
8. Multimodal AI
9. AI Presentation Tools
10. AI Summarization
11. AI Translation Tools
12. AI Limitations
13. AI Ethics Communication
14. Fact-Checking AI Output
15. AI Workflow Integration
16. Future of AI Writing

Prerequisites

This chapter builds on concepts from:

• Chapter 1: Communication Foundations
• Chapter 4: Brevity and Conciseness
• Chapter 5: Understanding Your Audience
• Chapter 7: Designing Presentations
• Chapter 11: Technical Writing

---

The AI Shift in Technical Communication

Every tool you have studied in this textbook — clarity frameworks, structure models, brevity principles, audience analysis — remains essential. What has changed is the speed and scale at which you can apply them. AI in communication refers to the use of artificial intelligence systems to assist with creating, editing, translating, summarizing, and delivering messages. AI does not replace the communication principles you have learned. It accelerates them.

Consider the workflow of a software engineer writing a design document. Before AI tools, the process was entirely manual: outline, draft, revise, get feedback, revise again. Today, that same engineer can use AI to generate a first draft from bullet points, restructure sections for a different audience, check for jargon, and summarize the document into an executive brief — all in minutes rather than hours. The engineer's judgment about what to say, who to say it to, and why it matters remains irreplaceable. The mechanical work of producing polished text is where AI provides leverage.

This chapter teaches you to use that leverage responsibly and effectively.

Understanding Large Language Models

A large language model (LLM) is a type of artificial intelligence trained on vast quantities of text data to predict and generate human language. LLMs such as GPT-4, Claude, Gemini, and Llama power the AI writing tools you encounter today. Understanding how they work — even at a high level — helps you use them more effectively and recognize their limitations.

LLMs work by learning statistical patterns in language. During training, a model processes billions of sentences and learns which words and phrases tend to follow others in various contexts. When you give an LLM a prompt, it generates a response by predicting the most likely sequence of words that should follow your input, given all the patterns it has learned. It does not "understand" your question the way a human does. It produces text that is statistically consistent with the patterns in its training data.

This distinction matters for technical communicators. An LLM can produce grammatically correct, well-structured prose on almost any topic. But it can also produce text that sounds confident and authoritative while being factually wrong — because it is optimizing for plausible language, not verified truth.

flowchart LR
    A[Training Data<br/>Billions of documents] --> B[Model Training<br/>Learn language patterns]
    B --> C[Large Language Model]
    D[Your Prompt] --> C
    C --> E[Generated Response<br/>Statistically likely text]
    E --> F{Human Review}
    F -->|Accurate| G[Use Output]
    F -->|Inaccurate| H[Revise or Discard]

The diagram above illustrates the fundamental LLM pipeline. Notice that the process does not end at "Generated Response." Every output requires human review before it can be trusted. This principle — AI generates, humans validate — is the central theme of this chapter.

AI Writing Assistants

An AI writing assistant is a software tool powered by a large language model that helps users draft, edit, rewrite, or improve written content. Examples include ChatGPT, Claude, GitHub Copilot (for code), Grammarly (for grammar and style), and Jasper (for marketing copy). These tools differ in their specialization, but they share a common architecture: you provide input, the tool processes it through an LLM, and you receive suggested text.

AI writing assistants excel at several tasks that are directly relevant to technical communication:

• Drafting: Generating a first draft from an outline, notes, or bullet points
• Rewriting: Adjusting tone, formality, or reading level for a different audience
• Editing: Identifying grammar errors, passive voice, wordiness, and unclear phrasing
• Expanding: Turning a brief description into a detailed explanation
• Condensing: Reducing a long document to its essential points

However, an AI writing assistant is not a replacement for your communication skills. It is a power tool, and like all power tools, it amplifies whatever you bring to it. If you give it a vague, poorly structured prompt, you get a vague, poorly structured draft. If you give it a clear outline with specific instructions about audience and purpose, you get a draft that is far closer to what you need.

Think About It

An AI writing assistant is like a GPS navigation system. It can calculate routes faster than you ever could, but it does not know where you want to go. You still have to set the destination (your purpose), choose whether to optimize for speed or scenery (your audience and tone), and recognize when the suggested route takes you somewhere wrong (fact-checking). The driver, not the GPS, is responsible for arriving safely.

Prompt Engineering

Prompt engineering is the practice of crafting effective inputs — called prompts — to get the best possible outputs from a large language model. It is one of the most immediately practical skills you can develop for working with AI tools. The quality of your prompt is the single biggest factor determining the quality of the AI's response.

Effective prompts share several characteristics. They specify the role you want the AI to adopt, the context it needs, the task you want it to perform, and the format you want the output in. Compare two prompts for the same task:

Weak Prompt	Strong Prompt
"Write about our new API."	"You are a technical writer for a developer tools company. Write a 300-word overview of our new REST API for a developer audience. The API handles user authentication using OAuth 2.0. Use a professional but approachable tone. Include a brief code example in Python."
"Summarize this meeting."	"Summarize the following meeting transcript in 5 bullet points. Focus on action items and decisions made. Each bullet should name the responsible person and the deadline."
"Fix my email."	"Rewrite the following email to be more concise and professional. The recipient is a VP of Engineering who has limited time. Keep it under 150 words and lead with the key request."

The strong prompts apply communication principles you already know — audience awareness, purpose clarity, structure, and brevity — to the act of instructing an AI. Prompt engineering is not a new discipline separate from communication. It is communication, directed at a machine audience.

The following framework, called CRAFT, provides a repeatable structure for building effective prompts:

• Context: Provide background information the AI needs to understand the situation
• Role: Specify who the AI should write as (e.g., technical writer, project manager, instructor)
• Audience: Describe who will read the output (e.g., developers, executives, customers)
• Format: Define the structure you want (e.g., bullet list, three paragraphs, table, email)
• Tone: Indicate the style (e.g., formal, conversational, persuasive, neutral)

AI-Generated Content and Human-AI Collaboration

AI-generated content is any text, image, audio, or video produced partially or fully by an artificial intelligence system. As AI tools become more capable, the line between human-written and AI-generated content is blurring. This raises both practical and ethical questions that you must navigate as a professional communicator.

The most effective approach is not to choose between human writing and AI writing, but to combine them. Human-AI collaboration is a workflow in which humans and AI systems contribute their respective strengths to produce a better result than either could achieve alone. The human provides judgment, domain expertise, creativity, and ethical reasoning. The AI provides speed, consistency, pattern recognition, and tireless iteration.

A productive human-AI collaboration typically follows this pattern:

1. Human defines the goal: What is the purpose, audience, and key message?
2. AI generates a draft: The human provides a structured prompt; the AI produces raw material
3. Human reviews and refines: The human evaluates accuracy, tone, completeness, and appropriateness
4. AI assists with polish: The human uses the AI for grammar checking, reformatting, or alternative phrasings
5. Human approves the final version: The human takes responsibility for the finished product

flowchart TB
    A[1. Human Sets Goal<br/>Purpose, audience, key message] --> B[2. AI Generates Draft<br/>From structured prompt]
    B --> C[3. Human Reviews<br/>Accuracy, tone, completeness]
    C --> D{Satisfactory?}
    D -->|No| E[Refine Prompt or<br/>Edit Manually]
    E --> B
    D -->|Yes| F[4. AI Assists Polish<br/>Grammar, formatting]
    F --> G[5. Human Approves<br/>Final responsibility]

This cycle can repeat multiple times. The key principle is that the human is always in the loop and always accountable for the final output. An AI-generated report with your name on it is your report. If it contains errors, you are responsible — not the AI.

AI Content Review

AI content review is the process of using AI tools to evaluate and improve existing content. This goes beyond spell-checking. Modern AI tools can analyze your writing for:

• Clarity: Flagging sentences that are too long, too complex, or ambiguous
• Consistency: Detecting shifts in terminology, tone, or formatting
• Completeness: Identifying missing sections or gaps in logic
• Audience fit: Assessing whether the vocabulary and complexity match the intended reader
• Bias: Highlighting language that may be exclusionary or culturally insensitive

AI content review is particularly valuable for long documents — technical reports, specifications, and user manuals — where human reviewers may miss inconsistencies that accumulate across dozens of pages. However, AI review should complement human review, not replace it. An AI tool can tell you that a sentence is complex, but only a human reviewer can tell you whether the complexity is necessary for precision.

Multimodal AI and AI Presentation Tools

Multimodal AI refers to AI systems that can process and generate content across multiple modes — text, images, audio, and video. Early AI tools worked only with text. Modern multimodal models can analyze an image and describe it in words, generate images from text descriptions, transcribe and summarize audio recordings, and produce video content from scripts.

For technical communicators, multimodal AI opens new possibilities:

• Convert a written technical explanation into a diagram or flowchart
• Generate slide visuals from presentation outlines
• Transcribe and summarize recorded meetings or lectures
• Create accessibility descriptions for images in documentation

AI presentation tools are a specific application of multimodal AI focused on creating and delivering presentations. Tools such as Gamma, Beautiful.ai, and Tome can generate entire slide decks from text prompts, apply consistent design templates, suggest visual layouts for data, and even create speaker notes. These tools apply the design principles from Chapter 7 — visual hierarchy, whitespace, consistency — automatically.

The following table compares how AI assists across different communication modes:

Communication Mode	AI Capability	Example Tool Category	Human Role
Text	Draft, edit, rewrite, summarize	Writing assistants	Set purpose, verify accuracy
Images	Generate, edit, describe	Image generators	Ensure relevance and appropriateness
Audio	Transcribe, translate, summarize	Transcription tools	Verify transcript, add context
Video	Generate captions, create clips	Video editors	Review quality, approve content
Slides	Generate layouts, suggest visuals	Presentation builders	Ensure narrative flow, verify data

However, the convenience of AI presentation tools carries a risk. When you can generate a full slide deck in two minutes, the temptation is to skip the audience analysis, narrative structure, and key message distillation that Chapter 7 taught you. The result is a presentation that looks polished but says nothing meaningful. AI tools produce form efficiently. Substance still requires human thought.

AI Summarization

AI summarization is the use of AI to condense long documents, conversations, or data sets into shorter, more digestible summaries. There are two types:

• Extractive summarization: The AI selects and combines the most important sentences or phrases from the original text. The output uses the author's own words.
• Abstractive summarization: The AI generates new sentences that capture the meaning of the original text. The output is a paraphrase, not a direct excerpt.

AI summarization is valuable for engineers who must process large volumes of information — research papers, meeting transcripts, incident reports, Slack threads, and email chains. Instead of reading a 30-page report, you can ask an AI to summarize the key findings, decisions, and action items.

The risk with AI summarization is that the model may omit important nuances, misrepresent the source material, or introduce errors in the paraphrase. Always verify AI summaries against the original source when the stakes are high. A summary of a casual team meeting can tolerate minor inaccuracies. A summary of a legal contract or safety-critical specification cannot.

AI Translation Tools

AI translation tools use language models to translate text between languages. Modern AI translation has improved dramatically over older machine translation systems, producing more natural and contextually appropriate output. Tools such as DeepL and Google Translate now handle idiomatic expressions, technical terminology, and complex sentence structures with reasonable accuracy.

For engineers working in global teams, AI translation is a practical necessity. You might use it to:

• Translate documentation for international users
• Understand error reports filed in another language
• Communicate with overseas team members who prefer their native language
• Localize user interfaces for different markets

AI translation works best with clear, well-structured source text. The clarity principles from Chapter 2 — simple sentences, precise vocabulary, logical structure — make your writing easier for both human readers and AI translators to process. Jargon, idioms, and culturally specific references are where AI translation most often fails.

Pro Tip

If you know your writing will be translated — by AI or humans — use controlled language: short sentences, active voice, consistent terminology, and no idioms. Write "The system failed to respond within the timeout period" instead of "The system dropped the ball." Controlled language doubles as good technical writing practice regardless of translation needs.

AI Limitations

Understanding AI limitations is not optional — it is a core professional skill. Engineers who use AI tools without understanding their boundaries will eventually publish incorrect information, make flawed decisions, or lose credibility. The major limitations of current AI systems include:

• Hallucination: LLMs can generate text that is factually incorrect but sounds authoritative. They may cite papers that do not exist, invent statistics, or describe processes that do not work. This happens because the model is generating plausible-sounding text, not retrieving verified facts.
• Knowledge cutoff: LLMs are trained on data up to a certain date. They do not know about events, publications, or technologies released after their training cutoff. A model trained through 2024 cannot tell you about a library released in 2025.
• Lack of reasoning: LLMs simulate reasoning through pattern matching, but they do not perform logical deduction the way humans do. They can fail on problems that require multi-step reasoning, spatial understanding, or mathematical precision.
• Context limitations: Models can only process a limited amount of text at once. Very long documents may exceed this limit, causing the model to lose track of earlier content.
• Bias: LLMs reflect the biases present in their training data. They may produce text that reinforces stereotypes, favors certain perspectives, or underrepresents minority viewpoints.
• No accountability: An AI has no professional reputation, no career at stake, and no ethical obligations. It cannot take responsibility for errors. You can.

flowchart TD
    A[AI-Generated Output] --> B{Is it factually verifiable?}
    B -->|Yes| C[Verify against<br/>authoritative source]
    B -->|No, opinion/style| D[Review for<br/>appropriateness and tone]
    C --> E{Verified?}
    E -->|Yes| F[Safe to use]
    E -->|No| G[Correct or discard]
    D --> H{Appropriate?}
    H -->|Yes| F
    H -->|No| I[Revise]

AI Ethics in Communication

AI ethics communication encompasses the ethical principles and responsibilities involved in using AI tools to create content. As AI-generated content becomes more prevalent, you must navigate questions that did not exist a few years ago:

• Disclosure: Should you tell your audience that AI assisted in creating your document or presentation? In academic contexts, the answer is almost always yes. In professional contexts, norms are still evolving, but transparency builds trust.
• Attribution: If an AI generates a passage based on patterns learned from millions of authors, who deserves credit? Current best practice is to treat AI output as your own work — and accept full responsibility for it.
• Authenticity: When you use AI to draft a personal message — a recommendation letter, a performance review, a condolence note — does the message still carry your genuine voice and intent? Authenticity matters in communications where personal connection is the point.
• Equity: Not everyone has equal access to AI tools. Using AI to produce polished work may create unfair advantages in contexts where the quality of writing is being evaluated — such as job applications or academic assignments.
• Intellectual property: AI models are trained on existing content. The legal and ethical implications of this training process are still being debated in courts and professional organizations.

There is no single rulebook for AI ethics in communication. But a reliable guiding principle is: Would you be comfortable if your audience knew exactly how this content was produced? If the answer is yes, you are probably on solid ethical ground. If the answer is no, reconsider your approach.

Fact-Checking AI Output

Fact-checking AI output is the practice of independently verifying claims, data, and references produced by an AI system before using them in your communication. This is not optional. It is a professional obligation.

A practical fact-checking workflow for AI-generated content includes the following steps:

1. Identify verifiable claims: Separate factual assertions (statistics, dates, technical specifications) from opinions and stylistic content
2. Check primary sources: For each factual claim, locate the original source — a research paper, official documentation, a reputable database
3. Verify references: If the AI cites a specific paper, book, or URL, confirm that the source actually exists and says what the AI claims it says
4. Test technical content: If the AI generates code, commands, configuration examples, or mathematical formulas, run them. Do not trust that they work based on appearance alone
5. Cross-reference with domain knowledge: Use your own expertise to evaluate whether the AI's claims are reasonable. If something sounds surprising, investigate further

AI Output Type	Fact-Check Method	Risk Level if Unchecked
Statistics and data	Verify against primary source	High — incorrect data undermines credibility
Citations and references	Confirm source exists and content matches	High — fabricated citations destroy trust
Technical procedures	Test in a safe environment	Critical — incorrect procedures cause failures
Definitions and explanations	Cross-reference with authoritative texts	Medium — inaccuracies mislead readers
Stylistic suggestions	Review for tone and audience fit	Low — subjective, easily corrected

AI Workflow Integration

AI workflow integration is the practice of incorporating AI tools into your existing communication processes in a systematic and sustainable way. Rather than using AI sporadically when you think of it, integration means building AI into your standard operating procedures.

A well-integrated AI workflow for a technical communicator might look like this:

1. Planning phase: Use AI to brainstorm outlines, identify potential audience questions, and research background material
2. Drafting phase: Use AI to generate first drafts from detailed prompts, then edit substantially
3. Review phase: Use AI to check for grammar, consistency, and readability scores, then have a human reviewer evaluate substance
4. Publishing phase: Use AI to generate summaries, abstracts, and social media posts from the finished document
5. Maintenance phase: Use AI to flag sections of documentation that may be outdated based on new product releases or API changes

The goal is not to maximize the amount of AI involvement. The goal is to use AI where it adds genuine value — speed, consistency, coverage — and preserve human involvement where it is indispensable — judgment, creativity, accountability.

Welcome to Your AI Toolkit

Building an AI workflow does not have to be complicated. Start by identifying the one communication task that takes you the most time each week. Try using an AI tool to assist with just that one task for two weeks. Evaluate whether it saved time and maintained quality. If it did, keep it and add another. If it did not, adjust your prompts or try a different tool. Incremental adoption beats wholesale transformation every time.

The Future of AI Writing

The future of AI writing is not a single destination but an ongoing evolution. Several trends are already visible and will shape how engineers communicate in the coming years:

• Deeper integration: AI writing assistance will be embedded directly into every tool you use — IDEs, email clients, project management platforms, documentation systems — rather than existing as standalone applications
• Personalization: AI models will learn your writing style, your team's conventions, and your organization's terminology, producing drafts that require less editing over time
• Real-time collaboration: AI will participate in live collaborative writing sessions, suggesting edits as multiple authors work simultaneously
• Agentic workflows: AI systems will move beyond generating text on demand to proactively managing communication tasks — drafting status updates from project data, flagging documentation that has drifted from the codebase, or preparing meeting summaries before you ask
• Regulatory frameworks: Governments and professional organizations will establish clearer rules about disclosure, liability, and acceptable use of AI-generated content

What will not change is the need for human communicators who can think critically, understand their audience, and take responsibility for their messages. AI will make the mechanics of writing faster and easier. The strategy of communication — deciding what to say, to whom, and why — will remain a distinctly human skill.

The engineers who thrive will not be those who resist AI or those who defer to it entirely. They will be the ones who use AI as a tool while continuing to develop the judgment, empathy, and critical thinking that no model can replicate.

Key Takeaways

• AI in communication accelerates the application of communication principles — it does not replace them
• Large language models generate statistically plausible text, not verified truth; every output requires human review
• Prompt engineering applies the same communication skills you already have — clarity, audience awareness, structure — to instructing an AI
• Human-AI collaboration follows a cycle: human sets the goal, AI generates a draft, human reviews and refines, human approves the final product
• Multimodal AI extends assistance beyond text to images, audio, video, and presentations
• AI limitations include hallucination, knowledge cutoffs, bias, and an inability to take responsibility for errors
• Fact-checking AI output is a professional obligation, not an optional step
• AI ethics in communication centers on transparency: if you would be uncomfortable disclosing your process, reconsider it
• Workflow integration should be incremental — start with one task, evaluate results, and expand gradually
• The future belongs to engineers who use AI as a tool while developing the human judgment that AI cannot replicate

---

Chapter 12 Quiz

Next Chapter: Feedback and Improvement — Learn how to give, receive, and act on feedback to continuously improve your communication skills.