What an awesome article on agents, hats off Petra.
Takeaways
Principles beat rules
Long checklists of specific rules make agents brittle and robotic. Durable principles (e.g., “Be helpful, not defensive” or “Lead with empathy”) transfer better to new situations and keep outputs more natural.
Agents need to learn how to learn: Raw feedback often leads the agent to add brittle exceptions/rules. You need a separate “meta-skill” that teaches it to extract generalizable principles from corrections instead of just patching cases.
Feedback must fit into existing team workflows: Make participation near-zero effort (e.g., emoji reactions in Slack). The learning loop should run where the team already works, or adoption drops. Treat evolving agent skills like code (PRs, reviews, version control) for safety.
Whether it’s existing consulting firms, new ones that emerge, FDEs from agent vendors, or new internal agent engineering roles, the amount of work that is going to be created to implement agents in enterprises will exceed anything we imagine today.
The complexity of implementing agents in any existing organizations is very real. When I talk to large enterprises, as you move from a chat paradigm to agents that participate in meaningful workflows, there are a number of things they need to do.
First, you have to get agents to be able to talk to your data securely across your systems. In many cases, enterprises have decades of legacy infrastructure that contain the valuable context for AI agents. That’s going to take a ton of work to go modernize and move to systems that work well with agents.
Then, you need to ensure that you’ve implemented agents with the right access controls and entitlements, the right scopes to be safely used, and have ways of monitoring, logging, and securing the work that they do.
Next, you need to actually document the processes in the organization in a way that agents can utilize for doing the work. You also need to figure out what the new workflow looks like when agents and people are working together on a process, and who steps in where. Just replicating the old workflow will mute the gains. Oh and you likely need to create evals for your top new end-state processes.
Finally, you have to keep up with a rapidly changing set of best practices and architectural shifts happening in the agent space. While it’s fun for people to change their personal productivity tools on a dime, it’s 100X harder to do this in a business process. The speed of change is a blessing and a curse right now for anyone trying to keep a stable system design.
All of this means that individuals and companies that develop expertise on the above set of components (and more) are going to be needed to help organizations actually implement agents at scale. This is also the rationale for vertical AI agents right now that can go in deep on a business domain and help bring automation to it.
This is a huge opportunity right now whether you’re doing this internally or as an external business provider.
@fortworthchris@nateliason@AlphaSchoolATX $1M GP by graduation is a big swing. What excites me more is that our smartest people are starting to point at education like it's a problem worth solving.
how much permission should you give an agent in an enterprise system?
Too much -> backdoor for humans to see things they shouldn't.
Too little -> hits the same walls as everyone else and fails silently.
where are you drawing that line?
@TTrimoreau any time spent working on anything that is not the most important thing is time wasted
hard part is figuring out what that is
Tim Ferris' 7 Steps for Prioritizing Effective work has been my go to framework for this
@BatsouElef https://t.co/anAve7cjMG
murder mystery, each suspect is an AI chatbot that will lie and try to push you off track
pictures / documents can be attached to their chat to contradict their lies, forcing them to tell the truth
the most exciting time to be alive is also the hardest time to focus.
AI hit me like a wrecking ball, monkey mind madness is back in charge
I went back to what worked before, once again it's helping me find signal
@tferriss's 7 Steps for Prioritizing Effective Work:
this song created by our AI overloads is incredible
listening to it feels like being on the titanic but instead of passengers playing the violin, the iceberg is serenading us
and yet i’m playing it on repeat haha