Carrying 25 people's work with 1 operator comes from 5 moves, not a better AI tool. Document every job. Group the jobs by type. Write each client's rules and context in 1 place. Connect that context to the tools you already use. Add 1 approval step before anything ships. Then the operator runs trained agents instead of handoffs.
The Bigger It Got, the Worse It Got
I scaled a 25-person marketing agency the hard way. And the bigger it got, the less profitable and less desirable it became.
Every agency owner is told the same story. Grow the team and the profit follows. I grew the team. The profit did not follow.
Here is the realization that changed how I saw it. Headcount is not capacity. Every new person also adds handoffs, meetings, briefs and reviews. The work around the work grows faster than the work.
So I stopped asking "Who do we hire next?" I started asking a different question. "What does each job actually do, step by step?" That question is where the map starts.
If you run an agency, you can feel this before you can name it. Do this. Look at your last hire. Ask if your margin went up or down in the months after.
The Problem Was Never the People
Most agency teams are good. That is the confusing part. Good people work hard, and the business still gets worse as it grows.
The problem is where the knowledge lives. In most agencies, it lives in people's heads. How a client likes their reports. Which claims legal will kill. Who approves what. When someone is out, the work slows. When someone leaves, the knowledge leaves too.
AI does not fix this on its own. It makes it more obvious. Every tool starts blank. Teams spend their time re-briefing ChatGPT, Claude and Gemini on clients they already know. The AI saves time on the draft and loses it on the setup.
That is not anyone's fault. Nobody had ever written the jobs down. You cannot hand a job to an agent that you have never described. Do this. Pick 1 role on your team. Try to write its weekly tasks on 1 page. Notice how hard it is.
Step 1: Document Every Job
This is the step everyone wants to skip. It is the step that makes the rest work.
I documented every single job. Not job titles. Jobs. The actual repeatable units of work that moved a client forward. For each one, I wrote the same 6 things.
- The trigger. What starts this job?
- The inputs. What does the person need in hand?
- The steps. What do they do, in order?
- The output. What does done look like?
- The approver. Who signs off before it ships?
- The standard. What makes it good, not just finished?
Write it in plain words. If a new hire could not follow it, it is not done. Agents are the same. They follow what is written, not what is meant.
The lesson. You cannot rebuild what you have not mapped. Do this. Pick the 1 job your team does most often. Write those 6 lines for it this week.
Step 2: Group the Jobs
Once every job is on paper, a pattern shows up fast. Most jobs are not unique. They fall into a small set of types.
Grouping matters because you do not build 1 agent per job. You build 1 way of working per group. Here is how I think about the groups.
| Job group | What the agent does | What the operator keeps |
|---|---|---|
| Recurring production | Drafts posts, emails, ads and pages from the client file | Final review and send |
| Reporting | Pulls data, writes the summary, flags changes | Reads the flags, decides what to tell the client |
| Client communication | Drafts updates, recaps and replies | Tone check, relationship calls |
| Research and analysis | Gathers sources, compares options, sums up | Judgment on what matters |
| Planning | Proposes next steps from goals and past decisions | Picks the plan, owns the priority |
Notice the right column. The operator keeps judgment, relationships and the final yes. The agents take the volume. That split is the whole idea.
Do this. Sort your documented jobs into these groups. Add a group if your agency needs another. Keep the list short.
Step 3: Write the Rules and the Context
This is where most AI projects quietly fail. The agent can do the job. It just does not know the client.
So every client gets 1 file. Their strategy. Their rules. Their priorities. Their approvals. The decisions already made, so nobody reopens them. Their voice, and the things they never want said.
This is what we now call Portable Delivery Intelligence. Each client's context lives in 1 place and connects to the AI tools the team already uses. The context travels with the work. Nobody re-briefs the AI or re-explains the client.
The rule I hold to. If it matters, it is written. If it is not written, the agent will not follow it, and the operator will end up fixing it. Do this. Open your biggest client's folder. Check if a stranger could learn their top 3 rules in 5 minutes.
Step 4: Connect the Tools
You do not need new software to do this. You need the tools you have to see the same context.
Your team already works in ChatGPT, Claude or Gemini. They already use a project tool, a shared drive, a chat app. The mistake is building context inside 1 tool where only 1 person can see it. Then the next person starts from zero.
Connect the client file to where the work happens. The agent drafting the report reads the same rules as the agent drafting the email. When the client changes a priority, you update 1 place, and every agent sees it.
This is also what makes the system portable. Swap a tool and the context comes with you. Lose a person and the knowledge stays. Do this. List every tool your team touched for 1 client last week. Mark which ones can see the client's current goals.
Step 5: Add the Approval Step
Agents without review are a liability. Review without rules is a rewrite. You need both.
So every piece of client work passes 1 approval step before it ships. The operator checks it against the client's written rules. Small misses get fixed. Big misses go back with the rule that broke. Then that rule goes into the client file, so the same miss does not repeat.
This is what 1 operator actually does all day. They do not draft from scratch. They run trained agents, review the output, make the calls that need a human, and keep the client files sharp. The work that used to move through a chain of handoffs now moves through 1 set of hands.
In our internal mapping, 1 operator plus the system now carries the work a 25-person marketing team used to carry. The same delivery system has supported $7M in client revenue.
Most agencies cap out at 4 to 8 accounts per account manager. Our per-operator target with the system is 18 to 25. You can see the proof here, and how we think about agency capacity.
You stop charging for seats and you start charging for deliverables. Do this. Add 1 approval step to your next AI deliverable. Log every send-back and the rule behind it.
FAQ
Can 1 operator really carry a whole team's work?
In our internal mapping, 1 operator plus the system carries the work a 25-person marketing team used to carry. It works because every job is documented, every client's rules live in 1 place, and the operator reviews instead of drafting.
Where should an agency start?
Start by documenting your most common job with its trigger, inputs, steps, output, approver and standard. Then write the client file it depends on.
Do we need to replace our AI tools?
No. Keep ChatGPT, Claude or Gemini, whichever your team uses. Connect them to the same client context so every tool works from the same rules.
What does Portable Delivery Intelligence add?
It keeps each client's strategy, rules, priorities and approvals in 1 place and connects them to your tools. Nobody re-briefs the AI or re-explains the client.
Bring 1 client account. We map where your team's time leaks and how many more accounts each person could carry. You keep the map either way. Book your map here.
