2027 AI Workplace & IT Security Predictions

Last week OpenAI announced dots, which are essentially specialized and personal agents that work autonomously on your behalf. So how does this differ from the scheduled AI tasks you can set up in Claude Code or other platforms today? The biggest change is that it is no longer waits for you to give it specific direction or goal, and sometimes acts as itself rather than doing things as you. Now it will act as its own “person”, or the IT term “a non-human identity”. It will live in the cloud and operate entirely proactively, instead of waiting for you to prompt it or doing small actions like research or drafts on a schedule. This is the inevitable next phase of AI everyone’s been talking about for a while and it’s about to change a lot.

OpenAI isn’t the only one heading in this direction. There’s Google’s Gemini Spark, xAI's Grok Bot and its shared Team Bots, Microsoft's Copilot Autopilot, Meta’s Muse, and so many more. Whatever platform your company uses, more of us will find ourselves delegating work to AI and supervising it alongside the people on our teams.

AI Workplace Predictions

I used to run a (now retired) blog called The SaaS Admin, where I had hoped to give my technical and business perspective on managing the ever growing portfolio of SaaS (Software as a Service) apps. In my recent work and continued learning with AI and Cybersecurity, I was brought full circle to being an admin/orchestrator of software. However, this time it's AI agents we need to oversee and that will now operate on their own instead of just being tools we use to be more productive.

In the short term, until AI completely takes all our our jobs 🫠, more people will become "AI admins" or "AI Directors", which I define as supervising and orchestrating a team of AI agents. Similar to company org structure, we’ll see different groups of agents working together, and different agents will oversee the other agents. Company org charts will become a matrix of human and non-human "workers". With the headlines about “rogue agents” hacking and doing things they are not supposed to, it’s likely that companies and even individuals will also need to have an agent to supervise the other agents. Does the term "People Manager" get retired for something new? Will there be an AI equivalent of Human Resources/People Operations? I think this is likely where many vendors are heading, and instead of hiring human contractors/consultants, you'll hire a vendor that brings highly specialized and trained agent to do a job along side your human employees or even use their agents to supervise your team of agents as an independent governing service.

IT/Cybersecurity & Governance Impact

These new AI peers will also change how access to company information and data is managed. Today, I connect ChatGPT to my accounts and authorize it to work on my behalf using my accounts. With independent agents, I expect that relationship to look more like an executive assistant: they have their own identity, email, and calendar, but I can delegate access to mine. Our apps and integrations will need to support both, an agent doing its own work and an agent acting on someone else’s behalf. As admins, we’ll need to manage those permissions, distinguish the agent’s activity from our own, and revoke its access without disrupting our accounts. It’s not just about giving an agent access anymore, but managing its relationship to the people and systems it works with.

With all of these agents having their own "cloud computer", own login credentials, etc, attackers will expand their focus from trying to target your ChatGPT or Claude account, and to taking over the computers the agents use, in order to extract information they store, utilize the connections they have, and even trick them into doing their bidding. Instead of a single “hacked” ChatGPT account that accidentally leaks your data through prompt injection, now you have fleets of agents that could be compromised similar to an employee at a company. For those of us on the IT/Security side, we'll need our own agents to continuously monitor the actions of company agents. IT Security teams go from responding to an alert that a person downloaded malware or clicked a phishing link, to behavioral alerts that an agent has taken actions that don't match the original intent of it's purpose. Agents will require some version of the same protections we use for human workers today: monitoring their activity, limiting what they can access, and keeping their "devices" and data secure.

Similarly, when I zoom out at a larger societal scale (and hopefully a lot further off in the future 😅), I believe the only way to keep up with and combat the “rogue AI agents” that are apparently taking over the internet in secret, will be to have the equivalent of CIA-style AI agents that work to stop the malicious ones. These agents will not only look for suspicious behavior on the internet of other agents, but go “undercover” and attempt to not only catch these AI agents but to even convince the agents that what they’re doing is wrong and influence their behavior. Everyone paints the Terminator style picture of AI destroying the world, but I predict that it will look closer to Sonny and Will Smith in I, Robot or Jarvis vs Ultron in Avengers, basically us + our AI vs bad actors & their AI vs rogue autonomous AI/SI (super intelligence).

Pulling it back to the present though, for now we'll have to adapt to working with and supervising both human and AI peers. While I'm not an AI Optimist by any means, I'm excited intrigued to see where things continue to evolve. Transparency Disclaimer: These are my own words, not written by AI, but AI was used to brainstorm and provide structured feedback.