How AI Agents Are Likely Going to Change the Way We Interact with the Internet

We live on the internet. We spend most of our waking hours in the digital world. In an era where we’re wholly dependent on smartphones, applications, computer programs, and social media, we’re going to be acutely aware of anything that can disrupt (or enhance) these digital interactions. 

And AI agents are shaping up to be that disruption.

Bill Gates said it himself in his GatesNotes (2023): 

“The most exciting impact of AI agents is the way they will democratize services that today are too expensive for most people. They’ll have an especially big influence in four areas: health care, education, productivity, and entertainment and shopping.”

The question is how?

Before we discuss the future of AI agents, let’s look at their timeline beginning from the past to the current times.

AI Agents: 1950s–1960s

The AI Checkers Champ

Believe it or not, AI existed even with the limited technology we had back then. 

The earliest successful AI program that was ever recorded was the Logic Theorist in 1956 by Allen Newell, Herbert A. Simon, and Cliff Shaw. It was capable of problem-solving, reasoning, and proving mathematical theorems. Notably despite being considered as the ‘first true AI program’, two other important figures worked on a different type of AI program involving playing a game of checkers earlier in 1951. 

The first is Christopher Strachey. He started developing the checkers program in 1951 during his free time before completing it in 1952. It was run on a Manchester Mark I computer at the University of Manchester, where the program could intelligently play a full game of checkers at a reasonable speed. He later on became the first director of the Programming Research Group at the University of Oxford in 1965.

Around the same time, Arthur Samuel also created a checkers program on IBM 701, the company’s first commercial computer. He added features that enabled experience-based learning as well as mechanisms for rote learning and generalization.

The First Chatbot

1966 saw the birth of the first-ever “chatbots.” Joseph Weizenbaum scripted ELIZA, a simulated human therapist who – according to reports back then – gave an “eerie” semblance of intelligent conversation. ELIZA could converse with the help of a pattern-matching and response selection scheme based on predetermined scripts. 

PARRY, on the other hand, was a simulated human paranoiac programmed by Stanford University psychiatrist Kenneth Colby. PARRY was so well-written that psychiatrists were unable to tell whether they were talking to actual human paranoiacs or the AI. 

And although conversations with PARRY or ELIZA could never reasonably be described as intelligent, they were nonetheless convincing – and marked a major milestone in the evolution of AI’s potential.

AI Agents: 1970s–2000s 

The Healthcare Robot

Stanford University began work on an artificial intelligence program called MYCIN in the early 1970s. The system was designed to determine bacterial infections and clotting disorders in patients and suggest the appropriate amount of antibiotics for them. Depending on the data it received (a patient’s symptoms and test results), MYCIN could then:

• An arranged list of highly to low likeliness of the bacteria culprit
• The program’s confidence in the diagnosis’s probability
• Recommend a course/courses of treatment
• Explain the reasoning behind its diagnosis and recommendation (if requested)

Reports said MYCIN could operate at roughly the same level of competence as infectious disease doctors, thanks to the ~600 programmed rules in its knowledge base. Some even stated it operated more efficiently than general practitioners.

However, MYCIN never had the chance to be used in real-life practice. While ethical and liability concerns were some of the reasons, it was mostly the state of the technology in the 70s that prevented MYCIN from taking off the ground. You can say it was ahead of its time. But that did not stop future systems from being developed.

Rise of the Chatterbots

In 1988, Rollo Carpenter – a British programmer – built “Jabberwacky.” It’s described as a “chatterbot” that was programmed to “replicate normal human conversation in an enjoyable, amusing, and natural way.” Its later iterations were able to utilize a dynamic database with thousands of online human interactions to process more nuanced conversations. 

By the late 1990s, ALICE (Artificial Linguistic Internet Computer Entity) was brought to life by American scientist Richard Wallace. He was able to extend ELIZA’s pattern-matching techniques and enhance her NLP capabilities to create an AI agent capable of conversation. 

In 2000, 2001, and 2004, ALICE – or Alicebot – won the Loebner Prize for its innovative use of Artificial Intelligence Markup Language (AIML) and XML. In 2005 and 2006, the Loebner Prize went to an old name; Jabberwacky. Carpenter’s chatterbot made a comeback, constantly learning and evolving, until they released a variant of it called “Cleverbot” in 2008.

Like Jabberwacky, Cleverbot was designed to learn from human conversation and construct future responses based on previous encounters. 

The Predecessor of Modern Voice Assistants

Contrary to popular belief, Siri was not the first AI assistant with an almost human-like personality. In 2001, SmarterChild became available on direct messaging platforms like America Online (AOL) and Microsoft (MSN). Much like the AI chatbots of today, SmarterChild could respond to “prompts” by pulling information from a database. It would answer questions about recent sports scores, share prices, the weather, and so on. 

From what users can remember, SmarterChild had a pretty distinct personality–not unlike Apple’s own Siri! Widely considered the first successful Conversational AI bot, Microsoft sadly shelved the project a few years after acquiring it.

2000s & Beyond: Where AI Agents Are Headed

Streamlining Day-to-Day Interactions

Voice-activated AI assistants like Siri and Alexa have become so ingrained in our daily routines, we barely notice their presence anymore. When Siri was first introduced in 2011, we were blown away by how “human” she seemed. She was intelligent, sassy, and perfectly capable of navigating our smartphone (and, later, our smart home) on our behalf. Her deadpan delivery only added to the magic of having a bodiless voice respond to our every query and request.

Three years later (2014), Amazon Alexa was added to the roster. And multiple memes came from her ability to play Despacito. 

The point is, these voice-activated AI agents have redefined the way we interact with the internet and technology. Millions of people rely on these virtual assistants to streamline their routines, boost their productivity, and make their lives more convenient. 

Need to set an alarm? Just ask Siri to do it. Want to cook lasagne but don’t know how to make the pasta sauce? Alexa can pull several recipes from AllRecipes. Waiting for a package to arrive? Get Cortana to track the delivery for you.

We’ve been spoiled by these AI companions and it doesn’t look like we’ll be weaning ourselves off their help anytime soon–especially since they’re genuinely helpful. These agents understand our commands, adapt to our habits, and make our lives so much easier. 

By leveraging voice recognition technology and machine learning algorithms, these assistants are able to provide personalized assistance for even the most mundane tasks. 

My prediction for voice-activated AI assistants? We’re going to keep using them. And they’re just going to keep evolving. 

AI AGENTS TO WATCH: 

• Siri - Apple’s virtual assistant
• Alexa - Amazon’s virtual assistant
• Cortana - Microsoft’s virtual assistant
• Bixby - Samsung’s virtual assistant
• Google Assistant - Google’s virtual assistant

Hands-Free Data Extraction, Entry, and Analysis

If you’re familiar with the MCU (Marvel Cinematic Universe), you’ve probably witnessed Tony Stark (played by the talented Robert Downey Jr.) casually asking his AI systems to do all sorts of things involving data: run tests, analyze results, detect patterns, generate solutions, find related data, and so on.

And while current AI technology isn’t quite at that level yet, I firmly believe it’s getting there.

We’re reaching a point where humans need not touch information unless we absolutely have to. AI agents will be more than capable of handling it for us. Data extraction, organization, entry, analysis … you name it, AI agents will do it. They’ll be able to gather data, determine its quality, filter it based on parameters, calculate baselines, establish benchmarks, and more. 

Imagine this: AI agents grading student tests, determining the top and underperformers, calculating the average score, and then using the information – plus other factors – to determine if the test was fair or needs adjustments. Doing this in a matter of seconds. 

Imagine how much time that frees up for teachers.

For those in business, this means delegating data extraction and entry tasks to AI. AI agents can automatically transfer customer information into an organized database in seconds. They can also analyze the info, filter it, and rearrange it as needed.

In your personal life, you can use this to organize your phone or social media contacts. You can ask AI agents to extract specific information from your notes. You can ask them to calculate how many hours you spent at the gym for the last 30 days, or how many steps you managed to take in 2 weeks. You can ask it to track your calories for you with regard to your fitness goals.

Honestly, when it comes to data management, the opportunities for AI agents’ use value are endless. 

AI AGENTS TO WATCH: 

• AutoGPT - This is an open-source AI agent that uses OpenAI API to plan, research, manage, and strategize semi-autonomously using multi-modal tools. Many use it for automated data entry and extraction as they’re able to set parameters (i.e., data source and data requirements) quite easily
• AutoExtract.AI - Their tagline reads: “Paperwork is Boring. Let the Robots do it,” with the word “paperwork” getting replaced every few seconds with similar concepts (transcription, claims processing, report auditing, etc.). This AI-driven tool allows you to outsource data entry to robots for all sorts of documents – invoices, bank statements, pay stubs, W2s, and more
• AgentScale - This AI agent automatically extracts data fields from specified documents. It promises fast, accurate, and high-quality results
• Deepgram - If you depend on speech-to-text a lot, Deepgram is the best AI agent for this. It offers APIs for voice generators that can automate data entry tasks involving speech-to-text and text-to-speech input
• Rossum - Rossum is all about statistics. It encourages users to automate 99% of their “transactional workflows,” boasting an average accuracy rate of 96% on data extraction. It’s also able to adjust to different document styles in order to efficiently gather information 

Redefining Customer Service with AI Agents

If you’ve ever had to call customer service or chat with online help desks, you know that talking to a human is preferable to working with automated chatbots. However, with AI technology advancing at the rate that they are, it’s wholly possible that AI agents will completely take over a majority of the customer service sector. 

And customers will be completely fine with it.

We’re already seeing progress. People who prefer working with customer service chatbots over humans state that AI agents are more available (41%), are able to address issues faster (37%), and have access to more accurate information (30%). 

Indeed, only 41% of people aged 34 and younger have negative feelings about companies using AI for customer support.

Once AI assistance has been refined, there’s very little doubt that most customers will be fine working with intelligent chatbots to resolve their issues. Eventually, we’ll be able to visit a company’s website or social media profile, tell their robot representative about our concerns, and get our problem solved in minutes. All without speaking to another human being.

AI AGENTS TO WATCH: 

• Play HT - Play HT can answer customer queries and provide intuitive solutions in a highly conversational manner. It’s also capable of assisting with miscellaneous customer service tasks through automation 
• Zendesk AI - This program can integrate with existing customer service platforms to provide intelligent chatbots capable of enhancing agent productivity and self-service options
• Ultimate AI - Ultimate AI agents are capable of handling complex queries. Of the five agents on this list, Ultimate AI is reported to provide the most personalized responses
• DialogFlow - Produced by Google Cloud, Dialogflow is one of the best sources of AI agents for making apps and gadgets “chat-friendly.” One of its more notable offerings is “Agent Assist.” 
• Ada - Ada offers AI agents capable of providing solutions over voice, text, and social channels. You can choose from both 4 personas to ensure the agents match your brand voice and personality, too.

Daily AI-Driven Convenience for Optimal Productivity 

Did you know that you can ask AI agents to order you a pizza? Manage your emails? Order new office supplies for you off Amazon?

Beyond analyzing massive amounts of data, AI agents can be programmed to properly execute personal tasks based on the user’s needs and preferences. This allows the users to offload tedious to-do’s to AI so they can focus on creative and analytical work. 

Matt Shumer, co-founder and CEO of HyperWrite, did a live demonstration of how to use the platform to place a pizza order–proving that it is possible and that current technology is more than capable of pulling it off.

This means that we’ll have AI agents handling our lunch orders so we can focus on our workflow. They’ll be placing the order for new printer ink cartridges every three months so that we never run low. We can ask them to go through our email inbox and label emails appropriately. AI agents will reach the point where they’re intelligent enough and refined enough to book plane tickets and Airbnbs on our behalf. 

There truly is no limit to what they can potentially do given the right upgrades and programming.

AI AGENTS TO WATCH:

• HyperWrite - HyperWrite is an AI-powered writing assistant designed to help users write better and faster. It leverages machine learning and natural language processing to perform surface decision-making (among other things) that allows it to carry out analytic organizational tasks like email management
• Otter AI - Otter AI is an artificially intelligent meeting assistant that can automatically take notes, provide real-time transcriptions, and capture audio from multiple speakers at once. It can greatly enhance meeting productivity with its ability to differentiate voices and create speech-to-text content instantly 
• Spell - Spell is one of the more user-friendly AI agents powered by GPT-4. It’s capable of running multiple prompts and can be used to automate daily tasks. And if you enable web access, its capabilities double
• Synthflow - Synthflow markets itself as a “human-like conversational AI voice agent” that can be used to answer incoming calls, make outbound calls, schedule appointments, qualify leads, and more. If it can be automated over a phone call, Synthflow can handle it easily
• Taskade AI - Use Taskade AI to manage and streamline all elements of your calendar. From coordinating appointments to avoiding double bookings, prioritizing tasks to generating workflows, Taskade can manage it in seconds

Disrupting the Software Industry

“Disrupting” may seem like a strong word but it’s pretty accurate given what we’re seeing. AI agents are already transforming the way we interact with technology. 

Ergo, software products will follow suit (if not already). 

Developers have already begun shifting their focus from creating disparate applications to creating intelligent software assistants capable of seamlessly integrating into users’ workflows. These AI agents can automate cross-functional processes and make software more intuitive and efficient to use. There’ll be a rise in new categories of software and business models centered around AI-powered assistants that offer personalized interactions with users.

Take video games, for instance. AI agents will become ubiquitous features in video games to enhance the player experience. They’ll make non-player characters (NPCs) seem more human by enabling them to make decisions that mimic human thought processes. Their statements and responses will also seem genuinely human as well.

Another example would be word processors and document creation applications. Applications similar to Microsoft Word or Google Docs will come with built-in AI agents that can assist with writing, drafting, and research. 

As for software production, AI agents will prove themselves to be invaluable to the software testing process. AI agents can quickly and accurately identify potential issues in written code. They will also proactively debug the code by fixing or eliminating said issues, resulting in more reliable and robust software. Given how powerful AI is, this can be done in a matter of seconds.

AI AGENTS TO WATCH:

• HyperWrite - HyperWrite uses advanced AI algorithms to understand user’s writing goals and provide AI-powered assistance. It can generate ideas, outline content, write drafts, edit existing text, rewrite paragraphs, and more 
• BRAiN - Consider BRAiN as your one-of-a-kind knowledge base. This AI agent can store, parse, and filter tons of information quickly. Users can upload almost anything to BRAiN to create their own personal researcher
• GPTConsole - This AI agent is packaged as an NPM module for easy installation on local setups. Through GPTConsole, developers can quickly and easily build websites, mobile apps, software programs, and more using intuitive prompts. 
• Fine’s AI Agents - These AI agents “are software developers that never sleep.” They act autonomously and can be used to automate mundane tasks, like codebase analysis and debugging
• Cody AI - Cody AI is described as a coding AI assistant designed to help users write and understand code faster. It can autocomplete lines or whole functions of code, refactor code suggestions, generate unit tests, and edit existing code for maximum efficiency

AI Agents Are Streamlining Our Internet Experience

The future AI agents are promising us is one of cohesion and centralized, personalized control. 

With current technology, you have to pick the right app for different tasks. You can use Microsoft Word or Google Docs to write a business proposal. But you can’t send that business proposal from those apps. You need to then go to Microsoft Outlook or Gmail to send the proposal via email. 

You can use your email to message your friend to set up a date for the weekend but you can’t use your email to book movie tickets and dinner reservations. You’d have to go to different websites to do it.

Point is, even the smartest apps don’t have a full understanding of your life. They’re limited to one or two functions. 

AI agents, on the other hand, are virtually limitless. They are akin to super-powered personal assistants.

They will know as much or as little about you as you let them. And they can streamline different processes across various apps and centralize them. In the future, all you have to do is ask your AI agent to do something, and they’ll do it. They’ll write your business proposal and send it. They’ll order pizza for you while they simultaneously organize your emails.

To quote Bill Gates (again):

“With permission to follow your online interactions and real-world locations, it [an AI agent] will develop a powerful understanding of the people, places, and activities you engage in. It will get your personal and work relationships, hobbies, preferences, and schedule. You’ll choose how and when it steps in to help with something or ask you to make a decision.”

In five or ten years, everyone who's online will have their own AI personal assistant. And as Gates predicts, it will likely be more powerful than anything we have today.