Agents, agents… everywhere. That’s how I feel about AI at the moment. Every conversation about it will probably mention ‘agents’. They haven’t been this famous Agent Smith and co fought Neo in The Matrix, and they’re likely to become more so.
Agent Smith = Powerful
But what exactly are they and what they can do? This newsletter will run through everything you need to know about agents. Just what you needed in time for Easter.
This email summarises parts of our white paper, The Rise of AI Agents, which you can access for free.
An ominous background
I’d become relatively familiar with the concept of agents since reading Microsoft AI CEO Mustafa Suleyman’s excellent book The Coming Wave. I’d also been able to watch Mustafa in conversation with Amol Rajan in September last year. While he is optimistic over AI’s potential, the agent remains a concern - particularly when they can reach a stage of ‘Artificial Capable Intelligence’ via what he frames as ‘The Modern Turing Test.’
In August, former Google CEO Eric Schmidt had spoken at Stanford University and the whole thing was livestreamed. He didn’t think it was, and made several controversial statements, some of which bordered on alarming. The video has since been taken down and now exists in clips and transcripts, but one quote stuck with me.
In the next year, you're going to see very large context windows, agents and text action. When they are delivered at scale, it's going to have an impact on the world at a scale that no one understands yet. Much bigger than the horrific impact we've had by social media in my view.
It all sounds rather ominous, yet at the same time suggests a huge amount of change on the horizon. I don’t, for the record, really deem social media’s impact as ‘horrific’. It’s not all good, of course - but this quote strikes me as pessimistic vs the potential benefits.
My feeling last Autumn was that fully autonomous agents were a long way off due to the amount of verification they would need and privacy concerns. For example, is the market really going to be prepared to give access to numerous accounts, calendars and possibly payment details to semi-automate their life?
Certainly, some would be happy to take the punt - just as some were with Google Glass. But it would currently be hard to ascertain for many but the deepest of AI enthusiasts.
What are agents?
Before we get much further, I’ve only really mentioned that there I’ve seen a trend, and concerns from some big names over the trend. But what really are these AI agents the have the potential to run amock?
We can start with the Cambridge dictionary definition of an agent. Two are relevant:
A business that represents one group of people when dealing with another group.
The ability to take action or to choose what action to take.
If you add AI into these definitions, it evolves the meaning.
An artificial intelligence that acts on behalf of someone, with the ability to take action or choose what action to take.
Thus we enter into a semi autonomous world. One where certain tasks are decided and acted upon, automatically, by artificial intelligence.
The ‘agentic’
Staying close to my media roots, I read Nic Newmans’ annual Journalism, media, and technology trends and predictions 2025 for Reuters.
Publishing and media is scratching its head, if not sweating, about generative AI. It represents a fundamental challenge through the possibility of making the destination website less relevant. This definition of ‘Agentic’ was provided in the exec summary:
Meanwhile, Google announced a major upgrade to its AI suite Gemini in December. One of its features was a remarkable feature of being able to see your screen in real time and troubleshoot via Google Live AI Studio (which is free).
It’s been touted as ‘agentic’ - although I don’t think it operates with agency. It is, however, the shape of things to come. If it did operate with agency, then it could automate your workflow on desktop. Indeed, Anthropic released an agent building towards this as part of Claude in October.
I watched Nvidia CEO Jensen Huang’s keynote at CES a few weeks back, and it came up again:
The 3 giant steps of generative AI (so far)
We’ve seen the development of generative AI in really three key steps:
2022 - Image generation broke into the mainstream via Midjourney (released in the summer).
2023 - ChatGPT was launched in November 2022. Claude and Gemini were released the following spring. Thus 2023 became a year of general assistance, with custom GPTs announced at the end of it.
2024 - More complexed rendering tasks become available. Generative images reach near parity with photography, but generative video also becomes a major force. Sora and Veo 2 are made available in the US in December.
The slide behind Jensen Huang at CES shows ‘Agentic AI’ as a takeoff moment that becomes more about assistance. But also, it becomes about joining the dots of the generative applications created in the previous phase.
How will agents be useful?
AI agents could be managing your calendar, booking flights and doing your online shop for you. To some end, this is true already, as there are plenty of AI tools in our directory that can assist to this end - although, like Gemini, I hesitate to call them ‘agents.’
But we don’t need to hand over the keys to our personal data to start using agents. Just think of the vast amount of boring tasks that occur in our daily work. The frustrating grind of compatibility, data entry or simply not being able to find something.
I’ll use social media as an example. In many cases social media distribution is the promotion of deeper level content, which we have to do manually. Would this not be better as an automated process? Having been a Head of Social Media at a national newspaper, I give the answer to this question as a resounding yes. Here is a typical process:
Write an article
Summarise contents of blog in your tone of voice
Format and find relevant image
Post to social media
Each of these steps takes time. The basic process for a social media manager doing steps 1-4 is a minimum of 15 minutes, depending on how long you spend on step 2. If it is someone’s role to distribute hundreds of stories a day (such careers exist in publishing), then a huge amount of time is spent on fairly mundane aspects of the process.
In the near term, taking over this sort of workflow is what the move to ‘agentic’ really means. This will almost certainly not become fully autonomous this year, but sensible businesses will want to free themselves from the more tedious aspects of process management as the technology becomes available.
Automations via Zapier have long been known in digital marketing, and can connect all sorts of apps to streamline processes. Zapier is certainly a flexible tool, but it now has several new rivals. Some of these, like CrewAI, have declared themselves as AI agent platforms already.
Automation vs AI workflow vs Agent
Not everyone has taken too kindly to the sudden use of ‘agent’ across social media. Those who have been working more deeply with agents have sought to make clearer distinctions, a I’ve seen the following graphic share more than once.
In even closer summary (and in case the words above are too small):
Automation - Rule based tasks are executed, like social media scheduling.
AI workflow - An automation/programme that calls on an LLM. An example could be a blog summariser that publishes to LinkedIn.
Agent - An AI that performs tasks autonomously based on prompts.
There are very few, if any, tools out there that truly act as an autonomous agent, but they could be defined as agent-ish (rather than truly agentic). They require frequent human instruction and iteration, and most of what we’re seeing at the moment is in the AI workflow definition.
Enter Boardy, the conversational agent
If you’ve used LinkedIn in the last week, then it’s been hard to miss Boardy. Ths conversational AI agent apparently managed to close an $8m funding round on its own through negotiating with investors.
Boardy’s premise is that it will phone you up, and have a conversation with you about your business. It will then do some background analysis and connect you with other people in its network.
The second part of the above paragraph is hardly something new. Social media is a connecting tool, although you normally have to do the hard work yourself to find relevant business connections (or get someone else to do it). But Boardy can do this automatically.
And that it can speak, in a near parity human like way, is what is making people’s jaw drop. It seems to understand people.
In reality, its a kind of AI automation, which could crudely be put together using some existing connecting tools. You could create a basic conversational agent by connecting Eleven Labs’ conversational agents (which do the talking), to an LLM like ChatGPT (which processes the information and provides responses. Indeed, you can have a conversation with ChatGPT just by opening your phone app.
I tried it out. Yes it’s very cool, and potentially very useful. A kind of ‘ChatGPT’ moment for conversational AI.
What we’re seeing here is a push towards the agentic.
Agentic ethics
By this stage, we might be reaching a mood of concern not dissimilar to Eric Schmidt, when he mentioned agents in the same breath as the ‘horrific impact’ of social media.
I must say I got a pang of anxiety when I read the definition in the above graphic.
A programme designed to perform non-deterministic tasks autonomously.
It reminded me of Nick Bostrom’s sobering book Superintelligence, in which he outlines certain dangers and related scenarios of a super intelligent agent in astonishing detail.
Of course, we’re not at levels of super intelligence, or ‘AGI’ as Sam Altman keeps saying, but there is inevitable risk in passing agency over to a machine.
For some time, there will be hesitancy to move in that direction - both at an individual and business level.
Two recommended reads on the subject of AGI and agents.
Indeed, this is part of the reason why I feel the enterprise and corporate world will be relatively slow in agentic adoption. How can we trust them? How does affect confidentiality and privacy? Won’t something go wrong?
Similar questions were raised c.2008-2010 with the rise of social media, and the more chaotic form of communication which swept the old order away.
The end of SaaS?
If you can link various applications together in a workflow, and then use LLMs or agents to automate parts of those workflows, then what is the need for Software As A Service (SaaS).
The success of the SaaS business model has been one of the great heralds in the last two decades of technology development. Adobe, Atlassian and Salesforce are all SaaS nobility, and investors are obsessed with it.
But there are clear signals of disruption on the horizon. In December (what a month for AI that was), Microsoft CEO Satya Nadella spoke on the BG2 podcast and suggested the SaaS era was reaching its twilight.
They are [SaaS applications] essentially CRUD (Create Read Update Destroy) databases with a bunch of business logic. The business logic is all going to these agents, and the agents are going to be multi-repo CRUD. They are not going to discriminate what the backend is. They’re going to update multiple databases and all the logic will be in the AI tier so to speak.
What’s this essentially means is SaaS is an interface that allows a user to interact with a database. With AI agents, this interaction becomes less necessary, and thus SaaS becomes increasingly obsolete.
To add, if coding and automation becomes increasingly democratised, as is already happening, then it will become increasingly easy for companies to build their own agentic custom products and workflows.
Whether we like it or not, the agents are coming. They are going to transform software and task based work. They will very probably augment entire job roles as we know them over the longer term.
I’ve maintained that I don’t believe AI is coming for all the jobs. Disruption yes, transformation yes, existential existential issues probably not. Few ever talk about the vast amount of unknown new jobs created by AI. Agents will be the driver of this change.
