There is a quiet shift happening among EOS® Implementers. The good ones are noticing that the companies they coach are not asking how to run a better L10® anymore. They are asking how to integrate Claude or ChatGPT into the L10®. And the implementers do not always have an answer.
This is not a threat to the implementer role. It is an expansion of it.
The Professional EOS® Implementer® already does the hardest part of what AI integration requires. The implementer makes the leadership team write the V/TO™ clearly. The implementer makes the team set Rocks with real Done states. The implementer holds the L10® to the time box. The implementer coaches the team into IDS discipline. Every one of those disciplines is exactly what agentic AI needs in order to work at all.
So the path from EOS® Implementer to AI Implementer is shorter than it looks. The framework holds. The new layer is on top of the old one, not instead of it.
What an EOS® Implementer is asked for now
Leadership teams that have run EOS® for two or more quarters are increasingly asking their implementer for:
- A point of view on which AI seats to add first.
- A way to keep the L10® disciplined when half the team wants to add ChatGPT lookups during the meeting.
- A V/TO™ that is sharp enough to be a system prompt.
- A Scorecard that can be defended at the definition level so an agent can push the number.
- A read on whether their internal "AI Director" hire is the right move or whether they should keep AI inside the existing seats.
- Help thinking through the agent seats on the Accountability Chart.
None of these are technical questions. All of them are framework and discipline questions. The implementer is exactly the right person.
What the implementer does not need to do: write code. Pick a model. Configure a tool. Implement an API. Those belong to a separate role, often an internal Chief of Staff, a technical Integrator, or an outside agent-build partner.
The split is clean. Implementer coaches. Builder builds. Same as it has always been between the implementer and the company's IT vendor.
The five things implementers can learn fast
If you are an EOS® Implementer and you want to credibly coach AI integration, learn these five things.
One, the difference between an assistant and an agent. An assistant waits for a prompt. An agent runs on a schedule, has a job, and escalates. This is the single most important framing. Coach it before you coach anything else.
Two, what a system prompt is. Not how to write one. What it is. It is the standing instruction set the model reads. It contains identity, job, values, constraints, escalation. The V/TO™ lives at the top. You can read one and tell whether it is sharp or fuzzy. That is enough to coach.
Three, what kinds of work agents are good at. Reading structured data. Summarizing. Drafting in a known voice. Routine triage. Scheduled monitoring. Cluster analysis. What they are not good at: novel judgment, high-trust human relationships, anything that ends in a contract or a public statement.
Four, what an SOP looks like for an agent. Six fields: identity, job, scorecard, data sources, escalation rule, off-limits actions. You do not need to write these for your clients. You need to be able to read one and tell whether it is rigorous or sloppy.
Five, the vendor landscape at a 60-second level. Anthropic makes Claude. OpenAI makes ChatGPT and the OpenAI API. Both have enterprise tiers with no training on customer data. Both are credible. The choice is rarely the bottleneck. Discipline is.
That is the whole curriculum. Five things. Each one teachable in a focused afternoon by someone who has been running agents for a quarter or more.
How an EOS® session changes
A quarterly with AI integration on the table has new exhibits but the same structure.
The implementer still opens with a check-in. Still reviews Rocks. Still does the V/TO™ section. Still does the Issues List. Still concludes.
What is new is the leadership team brings two new artifacts to the quarterly. The Agent Layer Map, showing each agent seat on the Accountability Chart, its accountability partner, and last quarter's scorecard. The Agent Issues List, showing the SOP gaps, model drift, vendor changes, and integration problems that surfaced during the quarter.
The implementer reads those the way the implementer reads any other prep doc. The implementer's job is to ask sharp questions. "This agent missed its scorecard for three weeks. Has the SOP been updated. Is the accountability partner answering for it." Same kind of question the implementer would ask about a human director.
The agent layer becomes part of the EOS® session. It does not become the session.
What this means for EOS® Worldwide and the implementer community
The honest truth is the community is moving at different speeds.
Some implementers have already integrated agents into their own practice. They use Claude or ChatGPT to read prior quarterlies, propose Rock candidates for review, format the V/TO™, draft the post-session memo. They coach clients on the same.
Some implementers are skeptical and rightly so. EOS® is a system that works because it is disciplined and simple. Adding AI without discipline ruins the simplicity. A cautious implementer is not wrong, they are guarding the framework.
The middle path is the right one. Take the principles of EOS® Implementer training (Get Real, Get Simple, Get Sharp) and apply them to the AI layer. Get Real means agents are not magic. Get Simple means one seat one agent. Get Sharp means definitions and SOPs that survive a quarterly review.
The implementers who do this work will quietly become the most valuable advisors their clients have. Because the companies are about to have a hybrid workforce, and almost no one knows how to coach a hybrid workforce yet.
FAQ
Should an implementer require clients to use Claude or ChatGPT? No. Implementers stay vendor-neutral. The discipline is the asset, not the tool.
Should an implementer build agents for clients? Generally no, unless the implementer has explicit technical depth. Recommend a partner who builds. Stay in the coaching seat.
How should implementers price AI integration coaching? Same way they price EOS® coaching. Per session, per quarter, per ongoing engagement. The AI work is additive context, not a separate product.
Will EOS® Worldwide publish official guidance on AI? Outside the scope of this article. Track the official EOS Worldwide channels for official positions. This series is an independent practitioner perspective.
EOS®, Entrepreneurial Operating System®, V/TO™, Level 10 Meeting®, L10®, Rocks™, Scorecard, Issues List, Accountability Chart, EOS® Implementer®, and Professional EOS® Implementer® are concepts and trademarks of EOS Worldwide, LLC. This article is an independent practitioner perspective and is not affiliated with or endorsed by EOS Worldwide.