What it does

Drafts, edits, summarises, translates, and reasons over text, and holds a conversation with you over multiple turns. The default starting point for AI in comms work.

The main tools

• ChatGPT (OpenAI). Strongest all-rounder, largest user base, most third-party integrations. Free, paid individual, team, and enterprise tiers. The enterprise tier does not use your inputs for training.
• Claude (Anthropic). Strongest at long-form writing, nuance, and following complex instructions. Includes Projects for keeping context persistent across conversations. Free, paid individual, team, and enterprise tiers.
• Gemini (Google). Strongest multimodal capability, meaning it handles images, video, and audio well. Best for teams already in Google Workspace.
• Microsoft Copilot. Strongest if your organisation runs Microsoft 365. Works directly inside Word, Outlook, Teams, and Excel, with appropriate permissions to your documents and calendars.
• Perplexity. A different shape: built around answer-with-citations, so it shows the sources it used. Strongest for research and current-information tasks where attribution matters.
• Open-source models (Mistral, Llama, DeepSeek, Qwen). Models you or your IT team can run inside your own infrastructure. Strongest where data sovereignty is critical and you have technical capacity.

How to choose between them

• If your data lives in Microsoft 365, start with Copilot before adding anything else. Integration usually beats raw capability for everyday work.
• If your data lives in Google Workspace, start with Gemini for the same reason.
• For long, nuanced writing with careful instruction-following, Claude tends to win.
• For quick all-purpose work and the largest ecosystem of integrations, ChatGPT tends to win.
• For research with citations, Perplexity, often used alongside one of the others rather than instead of.
• If your sector requires data sovereignty (some government, defence, legal), talk to IT about self-hosted options before committing to a commercial tool.

How to actually use them

The single biggest mistake new users make is briefing AI as if it were a search engine. A search engine takes keywords. AI takes context. A good brief has four elements: the audience, the purpose, the voice, and the constraints. Try the prompt below to see the difference between a thin brief and a full one.

Write a 200-word media response.
Audience: state government health journalists.
Purpose: respond to a journalist's question about our organisation's position on a new public health guideline.
Voice: measured, evidence-led, professional. Avoid jargon.
Constraints: do not invent statistics. Flag any factual claim I should verify. End with a quote from our CEO.
Topic: [paste the journalist's question and any background here]

Compare that to "write a media response about [topic]." The four-element brief takes 30 seconds longer and produces output you can actually use.

Common comms use cases

Drafting press releases, talking points, social copy, email responses, internal newsletters, blog posts. Summarising long documents into briefings. Generating Q and A from a policy paper. Translating between languages. Sense-checking your own writing for clarity. Ideating headlines and campaign concepts. Pressure-testing a draft by asking the AI to argue against it.

Risk to watch

Confidentiality. Anything you paste into a public chatbot account may be stored, used for training, or otherwise leave your control. Most enterprise versions address this; most personal accounts do not. Check what your account does with your inputs before pasting anything sensitive.

What it does

Generates images, video, audio, and graphic-design assets from text prompts or existing material.

The main tools

Image generation.

• Canva Magic Studio. Image generation, background removal, design suggestions, brand-kit integration. Best starting point for teams already in Canva.
• Adobe Firefly. Trained on licensed content, which matters for commercial-use clarity. Integrates with Photoshop, Illustrator, and Adobe Express.
• Midjourney. Highest aesthetic ceiling for many uses. Steeper learning curve, and weaker at text inside the image.
• Image generation inside ChatGPT (GPT Image). Strong on instruction-following and conversational refinement: describe a change in plain language and it adjusts. Easiest entry point.
• Google image generation (Gemini and Imagen). Strong photorealism and reliable text rendering, including non-English text. Good free tier and a fit for Google Workspace teams.
• Ideogram. Strongest at images that include readable text, like posters, signage, and social tiles with copy.

Video.

• Google Veo. Strong all-rounder for realistic marketing video, with native audio. A common first choice for short social and promotional clips.
• Runway. Image-to-video, text-to-video, and AI video editing with the most hands-on creative control.
• Kling. Strong on motion and multi-shot sequences, and cost-effective for high-volume iteration.
• Synthesia and HeyGen. AI avatars reading scripts, with voice cloning and translation. Common for training and explainer video.
• Descript. Audio and video editing by editing the transcript. The most forgiving starting point if you do not have a video editor.

Audio.

• ElevenLabs. Best-in-class voice generation and voice cloning.
• Resemble AI and Murf. Enterprise-friendly alternatives.
• Suno and Udio. AI-generated music.

Design and presentation.

• Canva Magic Studio, Adobe Express, and Microsoft Designer for everyday design.
• Beautiful.ai and Gamma for presentations with AI design assistance.

How to choose between them

• If your team already lives in Canva, start there before adding other image tools.
• If commercial-use clarity matters (publishing, advertising, regulator-facing work), Adobe Firefly's licensed training data is a meaningful advantage.
• If your images need readable text, Ideogram or Google's image generation are the practical choices.
• If you are doing video and do not have a video editor, Descript for transcript-based editing, or Google Veo and Runway for generated clips.
• If you need an AI voice for accessibility audio or short-form content, ElevenLabs is the default. Check disclosure obligations.
• For presentations under time pressure, Gamma or Beautiful.ai typically beat building decks from scratch.

How to actually use them

Image generators reward specificity in the same way conversational AI does. A vague prompt produces a generic image; a specific prompt produces something usable. Three things matter: the subject, the style, and the technical specification.

Subject: A community engagement scene in an Australian regional town. Three people in a sunlit community hall, mid-conversation, papers and notes on a wooden table.
Style: Documentary photography, natural light, neutral colour palette, shallow depth of field.
Technical: 16:9 aspect ratio. Photorealistic. No text in image.
Avoid: stock-photo composition, exaggerated smiles, blue tech overlays, futuristic elements.

The avoid list is the part most people skip and the part that does the most work. Image generators default to stock-photo conventions unless you tell them not to. Naming what you do not want is as important as naming what you do.

Common comms use cases

Social media tiles, presentation backgrounds, illustrations where stock photography fails, short video, voice-overs, accessibility audio, podcast trailers, infographic elements, animations.

Risk to watch

Copyright, brand consistency, and disclosure. AI-generated content of identifiable people raises additional risks. In Australia, AI-generated content that could mislead about whether it is human-produced may attract obligations under consumer law or sector codes.

What it does

Transcribes meetings, summarises decisions and action items, drafts emails and calendar invitations, and surfaces follow-ups.

The main tools

Meeting transcription and summarisation.

• Otter.ai. Strong transcription accuracy and integrations with Zoom, Teams, and Google Meet.
• Fireflies.ai. Particularly strong on action-item extraction and CRM integrations.
• Read.ai. Captures sentiment and engagement signals alongside transcription.
• Granola. A notepad workflow: you take notes during the meeting and the AI enriches them with context from the audio.
• tl;dv. Strong on multilingual transcription and short clip generation.
• Copilot in Teams and Gemini in Meet. Native to their platforms and included in the respective business plans.

Email and calendar AI.

• Copilot in Outlook and Gemini in Gmail, native to each client.
• Superhuman AI for speed-focused email, and Shortwave for Gmail.

Note-taking and knowledge.

• Notion AI, Mem, and Reflect (privacy-first) for AI-assisted notes.

How to choose between them

• If your meetings happen mostly in Teams or Google Meet, the native AI is the obvious starting point.
• For organisations that move across multiple meeting platforms, a third-party tool like Otter, Fireflies, or Granola is more flexible.
• For consent-sensitive contexts (client, HR, executive meetings), prefer tools that show a visible attendee badge so participants know recording is happening.
• For everyday email and calendar work, the AI inside your existing client usually beats a third-party tool.

How to actually use them

The biggest gain from meeting AI is not the transcript. It is the structured summary. Most tools produce a generic recap by default; the useful move is to ask for the summary in the structure you need.

Below is a transcript from a stakeholder meeting. Produce a one-page briefing with these four sections:
1. Decisions made (one line each, with the decision-maker named)
2. Action items (assignee, action, due date if mentioned)
3. Open questions (raised but not resolved)
4. Quotes worth capturing (verbatim, with the speaker named)
Do not invent decisions, actions, or quotes. If something was discussed but not decided, put it under open questions, not decisions.
Transcript: [paste transcript here]

The "do not invent" instruction is the key. Meeting AI is prone to inferring decisions that were not actually made, because that is what its training data expects to see in meeting summaries. Naming the constraint reduces it.

Common comms use cases

Capturing decisions from stakeholder meetings. Extracting quotes for internal newsletters. Drafting follow-up emails after media interviews. Generating action lists from board meetings.

Risk to watch

Consent. In most Australian jurisdictions, recording a meeting requires consent from participants. Inviting a meeting AI without consent is a Tuesday-morning decision with Friday-afternoon consequences. Standard practice is to flag the AI's presence at the start of the meeting and get explicit consent before recording begins.

What it does

Scans media, social, and online sources. Tracks mentions, sentiment, and emerging risk signals. Summarises landscapes and generates briefings.

The main tools

Media monitoring.

• Streem and Isentia. Australian, with strong coverage across print, online, and broadcast in the Australian and Asia-Pacific market.
• Meltwater and Cision. Global, wide-ranging coverage with AI summarisation. Cision is often paired with PR distribution tools.
• Mediaportal. Australian-focused, common in government and corporate.

Social listening.

• Brandwatch and Sprinklr for deep, enterprise-grade listening, including in regulated industries.
• Talkwalker, Sprout Social, and Hootsuite Insights for lighter-weight needs.
• Mentionlytics and Brand24 as mid-market alternatives.

AI-native research and search.

• Perplexity for landscape scans and competitor research, because it cites sources. Exa and You.com as newer AI-search options.

How to choose between them

• For Australian media coverage, Streem or Isentia tend to beat the global tools.
• For global media coverage, Meltwater or Cision.
• For social listening in regulated industries with enterprise procurement, Brandwatch or Sprinklr.
• For lighter mid-market needs, Mentionlytics, Brand24, or the AI features in Hootsuite and Sprout.
• For one-off landscape research and competitor scans, Perplexity is often enough on its own.

How to actually use them

Most monitoring tools produce daily digests by default, and the default digest is often noise. The useful move is to brief the AI to filter and summarise against your priorities, not against general coverage. The prompt below is for using Perplexity, ChatGPT, or Claude alongside your monitoring tool, not instead of it: the monitoring tool catches the coverage, the AI helps you make sense of it.

Below is today's media coverage about [your organisation or sector]. Produce a briefing that does three things:
1. Identify the three most significant stories from a reputational-risk point of view, and why each matters.
2. Flag any factual claims that should be verified before we respond.
3. Suggest two or three lines we could use if asked about the top story, written in our voice.
Voice samples: [paste two or three short examples of your spokesperson's voice]
Coverage: [paste coverage here]

This is AI as a thinking partner around monitoring data, not a replacement for the data. The AI has no access to fresh coverage unless you give it; the monitoring tool is what gives it.

Common comms use cases

Daily monitoring. Issues management. Crisis early-warning. Sentiment tracking around campaigns. Executive briefings. Competitor analysis.

Risk to watch

Bias amplification, especially in sentiment analysis. AI sentiment scores can be confidently wrong, particularly on Australian English, irony, and culturally specific references. Treat sentiment scores as directional, not precise. The other risk: paying for AI features in your monitoring contract that nobody on the team actually uses.

What it does

Connects multiple tools into automated sequences, runs repeatable workflows, and integrates AI into the systems you already use.

The main tools

Connectors and automation platforms.

• Zapier, the most established no-code connector; Make, more visual and more capable for complex flows.
• n8n, open source and self-hostable, strong where data sovereignty matters; Microsoft Power Automate, best if your work lives in Microsoft; IFTTT, the simplest of the connectors.

Custom AI assistants.

• ChatGPT custom GPTs, Claude Projects, and Gemini Gems: a reusable assistant holding your brand voice or evidence base.
• Microsoft Copilot Studio for building organisation-wide custom copilots.

AI agents.

• Agents that carry out multi-step tasks across tools are now real products, not just demos: ChatGPT Workspace Agents, Google Gemini Enterprise agents, the Claude Agent SDK, Microsoft Copilot Studio agents, and Salesforce Agentforce. For high-stakes comms the gating question is no longer capability; it is governance, data access, and human oversight. Proceed slowly and keep a human in the loop.

How to choose between them

• If your work is in Microsoft, start with Power Automate before adding Zapier or Make.
• For comms-specific automations, Zapier or Make are the most flexible.
• For repeatable AI tasks where the same brief gets used again and again, build a custom GPT, Claude Project, or Gemini Gem rather than re-prompting from scratch.
• For agents, watch the category but proceed slowly. The risk-to-value ratio is still high in regulated work.

How to actually use them

The most useful first automation for most comms practitioners is a custom GPT or Claude Project that holds your brand voice and core reference materials. A one-time setup that pays off every time you draft.

Project instructions (paste into the Project or custom-GPT setup):
You are the drafting assistant for [your organisation]. Every response follows these rules.
Voice: [describe in three to five sentences, or paste voice samples].
Audience: our primary external audience is [describe]; internal audience is [describe]. Adjust register accordingly.
Subject expertise: we work in [sector]. Topics we address: [list]. Topics we avoid: [list].
Constraints: do not invent statistics, dates, names, or quotes. Flag any factual claim that needs verification. Australian English spelling. Plain language. If asked for something outside your remit, say so rather than guessing.

Set this up once. Every conversation in that Project or GPT then carries these instructions automatically. The setup takes 15 minutes and saves you that brief every time you draft.

Common comms use cases

Sending media releases to a curated list with logged delivery. Auto-drafting follow-up emails after a meeting. Scheduled monitoring reports. Converting an approved press release into social posts across platforms.

Risk to watch

Invisible chains of automation. An automation no one is auditing is a system no one can fix or explain when it breaks. Document what your automations do, who set them up, and what they touch.