From the frontier
The Post-AI Agency
How agencies survive and thrive in the age of AI.
By Chris Willow
Former agency owner. 10 years of seeing what works across 1,000s of agencies.
Skill is abundant. Responsibility is scarce.
Agencies used to sell skilled labor by the hour.
Now clients can get a lot of the work done with AI. They can move faster. They can also make expensive mistakes faster.
So our job has changed. We decide what to do, what to ignore, and what is good enough to ship. Then we put our name behind it.
Clients pay you for a result they can trust. How many hours it took is your problem.
Bill by the hour and AI makes the invoice smaller. Sell a fixed result and AI makes delivery cheaper. The client gets the result they paid for. You keep the upside from building a better system.
If AI can do most of the work, what are clients actually paying us for?
Lessons from $500M in agency services.[2]
The top 1,000 Wayfront workspaces have sold more than $500M in agency services.
The patterns below come from what we’ve seen in the product and heard in customer interviews.
The best agencies make more money from the clients they already have.
They sell a clear result, deliver it well, and make the next project easy to buy.
Specialize, then charge more.
Agencies that do everything are easy to compare. That usually turns into a price fight.
Pick one kind of client, one problem, or one result you want to become known for. Do it often enough to build a better process and stronger proof.
When the right client thinks, “These are the people for this,” we can charge more.
What do we want to be the obvious agency for?
Productize one clear result.
Productizing means selling the same kind of result again and again.
Start with one thing clients already ask you to do. Give it a clear price, a clear scope, and a clear finish line.
Keep a person on the parts that need taste or judgment. AI can do the prep. We choose what actually gets sent to the client.
- Sell a finished result. The client knows what they get and why it matters.
- Set a hard boundary. Say what’s included, what the client must send, how many changes they get, and when the work is done.
- Name the hard calls. Show where experience still matters.
- Save what you learn. Put the brief, choices, and feedback back into the client record.
Could a client understand what we sell, what it costs, and where it ends without talking to us?
Post the price before the discovery call.
Don’t make clients book a call to learn what you sell or what it costs. Put the offer, the price, the process, and previous client results on the homepage.
When they book, they already know what the deal is. The discovery call is the final vibe check: do they trust us, and do we want to work with them?
One agency owner we interviewed lists 3 plans at $299, $499, and $749. He gives each lead a free competitor report. By the follow-up call, he says the data has usually done the selling. A “no” now costs him a 30-minute call instead of hours spent on a custom proposal.[3]
What could we show before asking for 30 minutes of someone’s time?
Before the callOffer, price, process, proof
On the callTrust, fit, edge cases
Give the agency one memory.
One person on your team uses Claude. Another uses ChatGPT. A third keeps everything in Notion. None of those tools knows what the others learned.
The best agencies keep sales and delivery in the same client record: how the client found us, what they bought, what they paid, what we promised, what they approved, who they referred, and what they should buy next.
That record should do more than help us deliver. It should show the next useful project before the opportunity gets lost in someone’s inbox.
Then our team and AI can use the same facts. Each fact still needs an owner, a source, and an expiry date. Old or wrong info poisons the next job.
One agency we interviewed spent more than $80,000 on custom dashboards and ended up with 3 separate ordering systems. One dashboard took about 8 months to build. The new systems gave the team even more work to run.[6]
I wrote a separate guide to AI client management if you want the practical version.
What will the next project know because we did this one?
Put every step in the right lane.
Use automation for steps that never change. Use AI for messy reading and writing. Use people for calls that can hurt the client.
- Automation moves things. Payments, forms, routing, deadlines, reminders, status changes, and delivery.
- AI does the first pass. Research, summaries, comparisons, sorting, drafts, and account notes.
- People make the call. Positioning, price changes, sensitive messages, final claims, approvals, and client trust.
After each job, save what we learned. The next person should never have to guess why we did it that way.
One video agency we interviewed sets a hard limit: up to 2 hours of editing per client per day, and up to 4 clients per editor. White-label work made up about 40% of the business when we spoke.[5]
The AI for agencies guide shows how to map one workflow from start to finish.
Which steps need rules, which need AI, and which still need us?
Keep people on the expensive mistakes.
AI can give us an answer. We decide whether it should touch the client’s money, name, or customers.
When a wrong answer can do real damage, write down the source, reviewer, approval, and final call. NIST says human and AI roles should be clear and someone should own the result.[7]
Automate the steps around the call. Keep the call human.
Some work won’t fit a neat box. Standardize what you can. Charge for the mess.
The agency AI policy is a plain checklist you can copy.
Where could a wrong answer cost the client money or trust?
A paid first project
The first project can make the next one obvious.
The agency sells a one-time Google Business Profile project. The client gets a finished result. The agency now has the client’s baseline, approved changes, and initial results. That makes monthly management easier to recommend and deliver.
The owner reported that 95%[1] of those buyers moved into the recurring service. Demand from freelancers, consultants, and smaller agencies later created another opportunity: a reseller offer.
Another agency owner used a £49 landing-page review to prove demand and collect testimonials. He later raised the price to $349. A related done-for-you service made up about 30% of his revenue. An automatic follow-up after 15 days often got the client talking again.[4]
The first project lowers the risk
A small paid project lets the client see how we work before they make a bigger commitment.
Save what you learn
Save the brief, decisions, approvals, and feedback. The next project doesn’t start from zero.
Turn one client into a sales channel
Partners can resell the same service instead of hiring their own delivery team.
This is the only way an agency gets negative net revenue churn.
Take only the clients you had at the start of the period. Their next engagements, larger scopes, recurring work, and added white-label volume must be worth more than what you lose from cancellations and smaller scopes.
New clients and referrals don’t count toward this number. Your existing book grows before sales adds a single new client.
What should this client buy from us next?
Steal the good parts of SaaS.
SaaS shows value every time a customer logs in. They can start, see progress, and know what happens next.
Agencies still make clients wait for a call, an email, or a monthly report. That feels broken.
Give clients one place to buy, send info, see progress, approve work, get files, and buy the next project. They wait less. Our team gets more time for the parts that need a person.
- Show value while it happens.A monthly report is a receipt. Let clients see the work move.
- Let clients move themselves forward.Let them buy, send info, approve work, and get files without booking a call.
- Show the next project.Clients should see what to buy next before the current project ends.
Where are we making the client wait for us?
Copy this playbook.
Start with one service. You can do the first pass in a week.
- Pick the service. Choose one result clients already buy from us. Pick the kind of client we want more of.
- Write the offer. Put the result, price, scope, client inputs, changes, due date, and finish line on one page.
- Map the work. Give each step to automation, AI, or a person. Keep people on the calls that can cost the client money or trust.
- Build one client record. Put the sale, brief, work, approvals, feedback, and next project in the same place.
- Show the next project. Give the client a useful next step before the current work goes cold. Ask who else could buy it.
Keep score with 5 numbers: margin, delivery time, revision count, revenue per client, and repeat sales.
Make every project improve the next one.
Every project should leave behind the brief, choices, approvals, results, and feedback that make the next project better.
Clients get a clear next step. Our team stops rebuilding the same context. We can earn more without hiring at the same rate.
We built Wayfront for this. It keeps buying, intake, delivery, messages, approvals, renewals, repeat sales, referrals, and reselling in the same client record.
Claude and ChatGPT help produce the work. Wayfront keeps the client context, promises, approvals, history, and next opportunity in one place.
Earn more from the clients you already have.Sell the result. Keep the upside. Make the next project easy to buy.
Build it in Wayfront
References and methodology
Sources for the numbers and examples in this article.