There is a lot of noise right now about what agencies should become in the AI era. Most of it asks you to tear up your org chart and start over. We disagree. Your org chart is not the problem. It is the map.
The Chasm Is Real. The Bridge Is Already Built.
Every agency leader we talk to feels the same tension. AI is clearly going to change how agency work gets done, but the path from here to there looks like a cliff: new structures, new roles, new operating models, invented from scratch while the current business still has to perform.
Here is the reframe that changes everything. You do not have to re-invent the idea of your organization to cross the AI Forward chasm. The structural integrity of your operating business is the accelerator, not the obstacle.
Your departments already encode how value flows through the business. Your job descriptions already define roles, scope, and outcomes. Your onboarding already teaches new team members how your agency works. Your approval chains already establish who has authority to decide what.
That is not legacy baggage. That is a deployment blueprint. You spent years building it. Use it.
Role-Based Deployment: Agents Join the Org Chart, They Do Not Replace It
The AI Forward motion is role-based and department-aligned. Agents are identified and activated the same way you would hire into the organization: against a role, inside a department, with a job description, an onboarding plan, and defined authorities.
In practice, agents deploy in two classes.
First-class agents: role-level partners
One per function or department: operations, business development, client services, creative, finance. A first-class agent carries the context of its role the way a senior hire would: the accounts, the history, the standards, the judgment calls that define the seat. It is onboarded against the same job description a human in that seat would receive, and it operates within the same authorities.
NBD sub-agents: Narrow But Deep specialists
The specialists underneath. Each one does a single thing exceptionally well: research, CRM hygiene, deployment checks, content drafting, reporting. They report to a first-class agent the way specialists report to a department lead. They are narrow by design. Depth is the point.
This is an org chart you already know how to read, because it is your org chart. Same departments. Same reporting lines. Same accountability structure. The difference is capacity.
How You Identify the First Agents
Skip the AI strategy offsite. The identification work is already done; it is sitting in your HR files.
Start with the job description.
Every JD in your agency is a candidate agent specification. Role, responsibilities, outcomes, interfaces. The roles with high-volume, well-documented, repeatable work surface first.
Use your onboarding as the training plan.
Whatever you teach a new hire in their first 30 days is exactly what an agent needs to know to operate in that seat. If your onboarding is thin, fixing it pays twice: better humans, better agents.
Inherit the authorities already in place.
Your approval matrix already says who can send what to a client, who can change a deliverable, who can spend money. Agents inherit those same boundaries on day one. Nothing reaches a client without crossing the same threshold a junior employee's work would cross.
No new governance framework to invent. No committee. The controls your business already trusts become the controls your agents run under.
The Outcome: Capacity Is the Guarantee
Let's name what this produces, plainly.
Increased human capacity is not a projection. It is a guarantee. When agents carry the high-volume, well-documented work inside each seat, every person in the lead operates with multiples of their former output. Run the math at the agency level and it becomes obvious: a 100-person agency can operate as a 200-person agency, at 10 to 15 percent of the cost of the second hundred. Not by working harder. By deploying capacity that does not sleep, does not context-switch, and onboards in days instead of months.
Capacity is the engine. The metrics it drives are the ones your business already lives by:
Speed
Work that took a week ships in a day. Turnaround becomes a competitive weapon.
Quality
Standards are enforced on every unit of work, not just the ones a senior reviewer had time to catch.
Growth
Capacity you used to buy with payroll becomes capacity you deploy at will, so you can say yes to work you used to turn away.
Margin
The cost structure of delivery changes permanently.
Notice what is not on that list: AI project counts, token usage, number of pilots launched, tools adopted. Those are activity metrics, and activity is not the business. An AI Forward agency measures exactly what it measured before. The numbers just move.
Human in the Lead, Not the Loop
"Human in the loop" puts a person at a checkpoint, reviewing output as it flows past. That is a quality gate, and it is not enough.
AI Forward deployments keep a human in the lead. The human owns the outcome, directs the work, sets the standard, and makes the judgment calls. Agents multiply that person's capacity; they do not dilute that person's accountability. Especially in early deployments, this distinction is the difference between an agency that compounds learning and an agency that ships unreviewed work until something breaks.
The lead is a role, not a rubber stamp. Leads catch the output that is technically correct but commercially wrong. Leads turn failures into permanent system improvements. Leads are how the agency's taste survives the transition.
Two Surfaces: How You Operate, and What You Deliver
Becoming AI Forward is not one transformation. It is two, and they reinforce each other.
How you operate. Your own GTM, your pipeline, your delivery operations, your reporting, your internal coordination. This is where you start, because you control every variable and your own business is the proving ground. An agency that runs an AI Forward GTM internally walks into every client conversation with receipts.
What you deliver. The same motion, productized for clients. The capacity you build internally becomes the capability you sell. Your clients do not want an AI slide deck. They want to see the operating model working, inside a business that looks like theirs.
Each and all can become AI Forward. The agencies that move first on both surfaces will not look like software companies. They will look like the best version of themselves.
Pick a Place to Start. Don't Look Back.
You do not need a transformation program. You need a first deployment.
Pick one department. Pull the job descriptions. Identify the first-class agent for the seat and the two or three NBD sub-agents underneath it. Onboard them the way you would onboard a hire. Put a human in the lead. Run it for thirty days against the metrics that department already carries.
Same business. Same metrics. Just better: in speed, in quality, in capacity, in growth, and in margin.
That is the whole thesis. Not a new company. Yours, rebuilt seat by seat, while it keeps running.
One Last Thing: Agents Run on Context
Everything above depends on a single ingredient we have not discussed: context. An agent without your agency's context is a generalist; an agent with it is a member of your firm. How that context gets assembled, and how it compounds into an asset your agency owns, is its own discipline. We call it the CLM, and we have the roadmap.
That is the next post.
Michael Murray
Michael Murray is the Managing Partner of Abeba Co, an AI accelerator that partners with agencies making the AI Forward transition: same business, same metrics, multiplied capacity. If you are picking your place to start, visit abeba.co.