The future of communication: How generative AI is changing the way we connect
Exploring the six levels of AI-powered communication: From assistance to autonomy
Imagine: There will be a point in time in the future at which we no longer know whether the conversation we’re having online is with another human being or with an AI — nor will we care. And on the flip side: At this point in time (or sooner), companies will no longer know whether the inquiry they’re receiving has been sent from a human or from an AI. AI will not only be able to mimic human behavior and communication, but likely even influence it as well. Already this week’s announcements of OpenAI GPT-4 and Microsoft 365 Copilot seem to bring us one step closer to this vision.
But until we’re fully there, AI leaders need to:
Understand the potential of generative AI in different types of media (text, image, audio, video)
Assess when/ where/ to what degree to incorporate generative AI in their own products
Realize that, while quickly evolving, generative AI for business is still in its infancy
…and they can start shaping this journey today. New opportunities will open up as the technology continues to mature and as we find answers to pressing questions about its use. But how will generative AI likely proliferate our communication?
» Watch the latest episodes of “What’s the BUZZ?” on YouTube or listen to it wherever you get your podcasts. «
Automation levels of communication with generative AI
To put this into a clear structure, let’s borrow from the six levels of driving automation by SAE International. This established framework for autonomous driving is also frequently used to describe the levels of human decision making and involvement as the level of automation and autonomy increases — from no automation to full automation. Here is how this applies to the future of communication with generative AI and what to expect.
Level 0: Manual
You sit down and think about what outcome you want to achieve, what you want to say, and which words you think best convey your message. You then type the message yourself and send it. It’s all you: 100%. There’s no automation or AI involved.
What exists today: E-mail clients, chat/ collaboration platforms.
Level 1: Suggestion
Your communication tool offers suggested replies to instant messages or e-mails (e.g. “Thank you!”). Sometimes these suggested replies are quite accurate and helpful. Sometimes, they’re just off. And sometimes, you can even tell if someone has just sent you a suggested reply when it doesn’t seem like the natural/ human thing to say (e.g. “Likewise” in LinkedIn’s direct message system). However, you still decide whether or not you want to use one of the suggested replies and which one you choose.
Already today, AI can understand the sentiment and topic of the conversation, and propose common replies.
What exists today: Suggested replies in e-mail clients, chat/ collaboration platforms.
Level 2: Assistance
Your communication tool can summarize messages and return just the key information such as message topic, urgency and tone, and deadlines and action items. You choose to respond. But instead of writing the reply yourself, you ask your AI-powered communication tool to do it for you. You review the draft before you send it out.
What exists today: Applications incorporating generative AI (e.g. Otter.ai) or as developer APIs (e.g. OpenAI GPT-4).
What to look for: Generative AI integrated into your communications tools (soon!).
Level 3: Extension
You want to communicate using more than just text. Maybe send an audio message to make it sound like a natural conversation. Generative AI helps you to create a suggested reply and to turn this text into audio that reproduces your voice. This removes the need for you to get on the phone (which you dread anyways). You review the generated output and decide whether or not to send it.
What exists today: Applications incorporating generative AI for one type of output (e.g. text: ChatGPT, audio: Eleven Labs, image: MidJourney, video: D-ID) or as developer APIs.
What to look for: (1) Seamlessly combining the output of one application with the input of another (e.g. text + image: OpenAI GPT-4 + DALL-E 2), (2) integrating this combination of generative AI models into your communications tools.
Level 4: Preparation
You want to get more of your time back to focus tasks that excite you or that need your creativity. You delegate the preparation of messages and replies to your communication tool. Its AI has grown in scope and capabilities and can draft an entire conversation on your behalf. You define your tone and style preferences (aka “guardrails”). In addition to audio, you can now also send video messages that are personalized with an avatar that looks, talks just like you, and does in the context of the conversation. If the conversation does not yield the intended outcome, AI schedules a meeting or transfers the conversation to a “live agent”: You.
What exists today: Same as Level 3 + prompts to model text-based output based on a sample/ template of text, e.g. your tone and style
What to look for: Automated draft of longer replies while retaining your tone and style preferences, and AI-based text summarization
Level 5: Delegation
You don’t want to be bothered by simple questions and replies — whether you’re busy or asleep. You delegate even more tasks to your communication tool. Its AI creates synthetic audio and video that resembles your features. It also initiates a conversation with the recipient and responds in your usual tone and style — on your behalf. AI fully manages and conducts conversations autonomously on your behalf.
What exists today: Same as Level 4
What to look for: Automatic writing of elaborate replies incl. complex subjects and reasoning while retaining your tone and style preferences
Summary
Communicating in several of these levels is already possible today (or imminent) — thanks to AI. Increasing the use of AI in higher levels might technically be feasible. But it will depend upon the users (us!) whether, to what degree, and on which kinds of topics it will also be desirable to fully delegate our communication.
Far-fetched or far-reaching? I’d love to hear from you.
» Watch the latest episodes of “What’s the BUZZ?” on YouTube or listen to it wherever you get your podcasts. «
What’s next?
Appearances
March 28 - Join me for a virtual conference hosted by SwissCognitive together with 20 Global AI Ambassadors and experts on Redefining Business Performance with Generative AI.
Join us for the upcoming episodes of “What’s the BUZZ?”:
March 28 - Amit Arora, VP - Head of Product Cyber Risk Solutions, and I will talk about Assessing Cyber Risk Potential With AI. Join live and ask your questions on the show.
April 11 - Maya Mikhailov, Founder & AI Leader, will join the show to share how as an AI leader, you can Set Expectations With Your Leaders.
April 25 - Ramsay Brown, Founder The AI Responsibility Lab, will be on the show when we talk about how ethics is key in the age of generative AI and rapid innovation.
Follow me on LinkedIn for daily posts about how you can set up & scale your AI program in the enterprise. Activate notifications (🔔) and never miss an update.
Together, let’s turn hype into outcome. 👍🏻
—Andreas