The Next-Gen Merch Agency

I co-founded Imprint Engine in 2012. I later co-founded MerchAI, an AI research project focused on the merchandise industry. I spend a lot of time thinking about where this business goes next.

I see a lot of opportunity. I also think what customers pay an agency for is going to change.

This is bigger than which tool to buy or which task to automate. If intelligence, coordination and execution get cheaper, what will customers still pay us for?

That is the question I want us to work through.

We’ve been here before

I remember a world before the internet. Information took effort to find. Then came cell phones, constant connectivity and access to almost anything we wanted to know.

We went from hunting for information to drowning in it.

That changed what we expected. It changed the businesses that helped us find and organize information. Now AI is starting to do the same for intelligence and the work it can do.

Think about the progression: answers, reasoning, action.

First, you ask a question and get an answer. Then you ask it to compare options, look at the risks and work through a plan. Give it the information, tools and permission it needs, and it can start doing the work.

The bigger change comes when these tools can work together. An answer helps you do a task. An agent can do parts of the task for you.

Imagine waking up to decisions

Imagine a client tells you they are onboarding 5,000 people globally. They want it to feel meaningful. They give you the budget and ask what you would do.

Today, that means a lot of work. Read the brand guidelines. Look at past orders. Find products. Contact suppliers. Compare prices. Check inventory and figure out delivery.

Now imagine that, by the time you read the message, agents have already started. They have looked at the brand guidelines, past orders, engagement data and budget. They have checked products, suppliers, inventory, pricing and delivery.

You arrive to find options. You can see the assumptions and what you would give up in each one.

Not every part of this works reliably today. But as more of it becomes possible, think about what changes in your day.

You wake up to decisions.

Which option best serves the customer’s purpose? Which assumptions need challenging? Where are the risks? What should change before anyone makes a commitment?

Hold that image for a moment. What does it make valuable?

What will AI never replace?

When people ask what AI will never replace, familiar answers come up: relationships, creativity, judgment and empathy.

I understand those answers. But before we build a business around them, let’s take them apart. The next time you say AI will never do something, ask: what would it need in order to do it tomorrow?

I use five categories:

• Knowledge: What would it need to know?
• Context: What would it need to understand about this customer and situation?
• Reasoning: What would it need to compare or think through?
• Tools: Which systems would it need to access?
• Agency: What permission would it need to act, and within what boundaries?

Break the task down. What is missing? What happens if AI gets those capabilities?

Take relationships and judgment apart

Consider relationships. What creates a strong customer relationship? Memory matters. So do follow-through, responsiveness and trust.

A customer wants you to remember what they told you. Keep your promises. Tell them about problems early. Respond like you understand what they need.

Which of those things depend on information we already have? Where do we fall short because it is scattered across the business? Could a system help us remember a promise and keep it?

Trust still has to be earned. But be specific about what earns it.

Now take judgment apart. We define success, weigh trade-offs, assess risk, consider scenarios and learn from outcomes. AI can help with parts of that. It can work through options we would not have time to consider ourselves.

Saying “we have judgment” isn’t enough. What are we trying to do? What are we willing to give up? What will we take responsibility for?

You own what happens next

A system can make a recommendation. Someone still has to decide what matters.

We make promises in this business. There is a deadline. There is a person opening the box. There is a customer with a budget, a reputation and expectations that might not all be in the brief.

We need to decide which promises we can keep and what we will do when circumstances change.

We still make the promise to the customer. Better tools give us more options, but we own what happens next.

For those 5,000 new hires, the cheapest option might feel wrong. The most customized option might arrive late. The fastest option might miss what the customer wanted.

Decide what matters. Choose. Then stand behind the promise.

Use the time to take on a bigger problem

As intelligence, coordination and execution get cheaper, value moves upstream. Cheap means abundant. It does not mean bad.

We can spend less time doing and more time choosing.

Return to those 5,000 new hires. What should someone feel on their first day? What went wrong with the previous program? What needs to be different across regions? How should this year’s experience reflect changes inside the company?

You need time with the customer to ask those questions. You need to understand why they are placing the order.

If agents give us time back, we can get involved earlier and take on more of that problem.

I think about the evolution this way: products, projects, outcomes. We sell products. We manage projects. We can increasingly take responsibility for designing outcomes.

That means taking responsibility for a bigger customer problem. It means being clear about what success looks like.

The question becomes: if AI gives you time back, what are you going to do with it?

The shape of what comes next

Here is what I think is coming. I do not know exactly when.

1. Quotes in seconds. If a quote can take seconds, customers will not want to wait days. It still needs to answer the right question and be accurate.
2. Automated pricing. Systems do more of the pricing work. We still set the rules and handle orders that do not fit them.
3. Product search recedes. We spend less time hunting through catalogs and more time choosing from options built around what the customer needs.
4. Suppliers sell to agents. Agents may ask suppliers for information, compare answers and coordinate orders for buyers. Supplier information needs to work for machines, too.
5. Revenue and headcount decouple. More orders may no longer mean adding people at the same rate. We can use that time to take on a bigger problem.
6. Context becomes the asset. The tools may be similar. What we know about our customers and what went wrong last time is not.

Context becomes an asset

If we all have access to similar tools, what makes ours different? What we know.

Remember the customer’s preferences, promises, past decisions and what went wrong before. Remember why. Knowing what they ordered helps. Knowing why they rejected something else can help even more next time.

Write that down so your team and your tools can use it. Make better recommendations. Stop learning the same lesson over and over.

Keep company work in company accounts. Know what information you are sharing and who can see it. Get someone who understands security involved before giving a tool more access.

You need the context. You also need to look after it.

Start with experience, then build a habit

At Imprint Engine, agents review our kitting and fulfillment workload so leaders can see what needs attention each morning. They also research incoming leads before we hand them to a salesperson. We keep improving them as we learn.

Start with a small task you know well. Try the tools yourself. See where the result helps and where it falls short. Look for assumptions it got wrong or information it did not have.

You need that experience to understand what is changing.

Find people in each department who want to try things. Give them time and support to help their teammates. Get them together regularly to share what is working.

Read what you send. Make sure it says what you mean. Sensitive conversations with your team need your own thought and care.

Keep coming back to the customer. What has gone wrong before? What needs to be different this time? Use the time you get back to take responsibility for a bigger problem.

Reinvent the value

The biggest threat is defending value that is already gone.

We can keep saying a machine will never do something. Or we can ask what it would need, watch what changes and decide what to do next.

The future belongs to the ones willing to reinvent their value.

If intelligence becomes abundant, what will you choose to become responsible for?