Landscape overview

AI agents are now the rage and we are now hearing companies announce agents focused on different verticals (a popular term being used is AI teammates/ AI coworkers). Recent announcements from large companies like Google, Asana, and myriad startups around AI teammates, shows the growth in the AI agent ecosystem.

As the number of AI teammates grows and eclipses the number of human teammates, a lot of changes are required around culture, governance, processes, traceability, and change control. Only when these changes are implemented can we realize the maximum value from the technology. Disruptive changes (gig-economy as an example) lead to changes in behaviors and processes, and AI agents will change more than just technology in the Enterprise.

We wrote about the changes needed to Govern exponential AI technology in an NACD (National Association of Corporate Directors) article in early February this year. Here is a link to the article.

Even in the Cloud era, we are still seeing companies struggle to realize value (and talk of moving back on-prem) because they see it as a technology change and are not looking at it more holistically.

What are AI Agents?

AI Agents are autonomous systems that can perform complex multi-step goals without human intervention. Agents break the goals into actions. They then perform these actions, using multiple tools and datasets even as they recalibrate based on environmental feedback (failure) to achieve the goals.

Example

Here is an example "Drug discovery" agentic workflow. For a question that is asked the agent can first get the right research papers and analyze them using LLMs. Based on the same it can fetch the datasets, analyze the data by installing the right python packages like SciPy, NumPy etc (and if it does not install properly, get permission to "pip install" the right versions) and based on the task, perform any forecasting that is necessary by using the right models. It can tune use LLMs to give the results in Natural language.

Notice the level of autonomous-ness and the level of changes done by the agents. For any critical tasks in production environments.

Changes needed

Operator economy => Builder Economy

“We’re going to see going to see 10-person billion-dollar companies pretty soon....and there is a betting pool for when we will see a 1-person billion dollar company" - Sam Altman

The Zscaler CEO Jay Chaudhry talked of 2 kinds of employees' builders and operators. Builders are curious learners who often have strong views, are on a mission, comfortable with chaos, passionate about their work and its impact. They have a high risk appetite.

Operators, on the other hand like well defined structured processes, value status and compensation, even as they work to build, manage and maintain teams and their morale.

The impending AI agent revolution, with processes being continuously validated and updated lens well for the Builders.

Automated Enterprise => Autonomous enterprise

Enterprises will change from using AI as a tool to help with automating menial tasks to working and treating AI as a tool to help amplify human creativity. Robotic Process Automation (RPA), chatbots were examples of automated Enterprise.

Autonomous Enterprise is about more empowerment to AI tools, even as we have checks and balances where humans can overlook and then quickly trace every activity to identify changes and root causes.

Static governance => Dynamic Governance

We discussed how governance is failing exponential technology in our article. In an AI agent world where enterprises are more autonomous, enterprise processes have to keep up with the change in the enterprise landscape.

Current Enterprise Governance models are top down and often removed from reality. The governance structures and processes need to be continuously changed based on the reality in the organization. This will help us move from human centric processes to an autonomous enterprise centric world view and processes, where humans are still in control.

Maker => Designer/ checker

In an autonomous world, the role of a human will shift from being a maker to being a designer and checker. The autonomous enterprise will replace menial work that makes us feel less human to more fulfilling work. Work will be about continuously changing, understanding others' needs and imagining possibilities versus actually putting the nuts and bolts to get things done.

Enterprises will be value humans not just being a cog in the wheel, imparting and having existing knowledge but even more towards their cognitive capital. Having customer empathy, having customer networks and key IP will be the priorities for the autonomous enterprise. BTW here is an interesting link to the history of future of work.

What does Agentic AI mean for cybersecurity?

In an Autonomous enterprise being able to control changes done and to be able to trace the root of an action will be critical. Being able to monitor the components and quickly sandbox, suspicious or malfunctioning components will be critical. Oversight (checker) and being able to pinpoint accountability will be critical in such environments.

As Precize we have been interesting technology in this space which we will unveil soon. We are working on Governance, traceability and other challenges that are there in the automated enterprise and will be greatly magnified in the autonomous enterprise. Want to keep up to date use this link.

Timelines

As engineers often look at only the engineering aspects when considering the time lines of adoption technology and often forget the organizational inertia and politics. My view is these technologies will take at least 10 years to become widely adopted across vertical and geographies, however the early use cases and quick wins in industries that most direly need this will be seen in the next 3-5 years. I would guess customer support, marketing, sales and even software will be affected by this.

Questions

If you have comments, questions feel free to reach out to me. I started this focusing on Cybersecurity but realized I need to provide a background of this thought.

A more detailed cybersecurity related post will follow.

References

‍

Share This