Everyone gets an AI agent

In the near future, every content consumer, creator, and newsroom will have an AI agent that works for them. This will change the way we find and interact with information, and therefore how we publish and monetize it.

An agent is a software program that acts on behalf of a person or a party. The most important thing to know about any agent is who it serves. A browser serves you when it renders web pages. A reservation system acts on behalf of an airline. Search and social media algorithms serve the companies that built them. Algorithmic agents sit between consumers and content creators and make these companies gatekeepers of information. We should never allow a few parties to amass this much power on the internet again.

It’s possible that generative AI overthrows current gatekeepers and becomes the new one. But I believe there is a narrow path for a reset, where we level the playing field. In this future, every individual and every newsroom has an AI agent that only serves them. Power is much more evenly distributed, and news organizations and the audience connect more directly and more deeply. Or as we like to say at Mozilla: “This technology could fall into the right hands.” This is how I imagine it starts in 2024.

How we interact

Early adopters will see personal AI agents emerge in 2024. My agent’s top priority will be to protect my time and wellbeing. Much of the internet provokes and manipulates. In the future, most content will be written by bots. My agent will work with your newsroom agent to help me cut through the noise and discover, filter, read, analyze, fact-check, refute, summarize, and recommend — all while I sleep. My agent will ask your newsroom agent clarifying questions — perhaps about a quoted source’s credentials and biases.

My agent will help me discuss and refine my positions. It has the final say in what I see, when I see it, and in what format. “Story page” begins to lose meaning. But when I want to read a long-form piece the way the human writer intended, I will finally have the time.

How we publish and serve

By the end of 2024, newsrooms will begin using agents to tailor their reporting and information serving to the needs of each reader. When news consumers have quick questions, they’ll ask their trusted news agent — not a search engine. One of your agent’s most important jobs is to reach out to my agent many times per day with important updates and breaking news. My agent decides when to interrupt me.

In 2024, most newsrooms will begin experimenting with the content itself. They’ll start with making AI-parseable feeds and “writing for AI” the same way they wrote for SEO. Things will get really interesting when newsrooms deliberately try to break the current news story model. You’ll experiment with unstructured content that allows the agent to determine the presentation. Over time, newsrooms will move away from publishing specific formats on deadline and towards becoming modular information services. The underlying content will always be human auditable, both by your newsroom and by the public.

How we monetize

Today’s dominant online revenue model has consumers trading their attention for free content. This model can’t work if my AI agent’s priority is to defend my time. The hardest thing about personal AI agents might be getting content creators paid.

AI will accelerate the current trend of premium content fleeing behind paywalls. Agents will work well with paid subscriptions, and most of the news innovation in agents will come from companies with subscription models. I will sign my agent into my subscription where it interacts with your news feed and your news agent.

Ad-monetized content will struggle even more when the agent can provide a good-enough answer without anyone having to ever visit the original story page. The donation model (NPR, ProPublica) will translate to agents. We’ll revisit micropayments and content bundles yet again and may finally solve them.

How we keep it open

If we want to avoid crowning a new AI gatekeeper, we need choice and competition. This is where open source excels, as we saw in the browser wars. Open source AI is just getting started and hasn’t been a viable alternative for small teams to build upon. That is quickly changing as we head into 2024.

Much of what we’ve seen so far is not truly open source, but it’s permissive enough to get us started. Amazon recently began offering Llama 2 hosting on AWS. Demand for alternatives also surged after OpenAI’s governance crisis underscored the risk of betting your AI future on one startup.

Scary, but better than the alternative

I have palpable AI concerns that I didn’t raise here. But I still believe the best option is a future where individuals and newsrooms have access to the same powerful technologies that big companies have. Let’s give ourselves a chance in 2024.