How three AI agents can automate your customer acquisition

Three connected AI agents that find companies with a real need right now, build personalized material, and handle the outreach. A walkthrough using a baker who wants to sell anniversary cakes to companies.

In my consulting projects I spend a lot of time on sales optimization with AI, and one thing keeps showing up: the most advanced agent rarely wins. What works is taking existing tools and redesigning the sales process around them.

From broad marketing to proactive customer identification

A concrete example shows how the traditional sales process can be turned on its head. Picture a baker who wants to sell anniversary cakes to companies. The traditional approach:

• Run Facebook and Instagram campaigns
• Send newsletters
• Update the website with the new offer
• Hope a few percent of the people who see the ads click through

This is reactive and inefficient. You spread the message wide and hope the right customers find you.

The new approach: find the customers who need you now

What if you instead start by identifying the companies that actually need anniversary cakes right now?

In the baker's case, upcoming company anniversaries are public information. You can look up company age in the CVR register and check websites and social media for announcements about anniversaries or events.

Doing that by hand would eat your week. That's where AI comes in.

The agent system

I've built a system of three connected AI agents that automate the whole process.

1. The research agent

This agent continuously scans the local area for potential customers. It:

• Goes through CVR data to find companies with upcoming anniversaries
• Monitors company websites for event announcements
• Spots employee anniversaries and other occasions worth celebrating
• Filters and prioritizes leads by size, timing, and probability

Say the agent discovers that Novo Nordisk turns 100 and has announced a big celebration six months out. That's a perfect lead. They will need cakes for both internal and external events.

2. The design agent

Once a relevant lead is identified, the design agent takes over. It:

• Collects the company's brand material (logo, colors, design guidelines)
• Works out the character and scale of the anniversary
• Generates personalized images of cakes with the company's branding
• Builds a tailored landing page with the relevant product suggestions
• Prepares the presentation material

The personalization is the point. Instead of a generic offer, Novo Nordisk sees exactly what their anniversary cakes could look like, with their own logo, in their own colors.

3. The outreach agent

The outreach agent handles the contact itself:

• Finds the right contact people in the organization
• Writes personalized messages
• Sends emails with the tailored material
• Follows up
• Coordinates meetings

The right message at the right time, sent to a company we know has a need.

From theory to practice

I'll be honest: it isn't as simple as it sounds in theory. But the process is straightforward to understand and implement.

In my current projects we still run the process fairly manually. We've built the three agents and are testing them separately to:

• Find the failure points
• Tune each agent's performance
• Improve output quality
• Learn where human review is still needed (human-in-the-loop)

Once each agent works on its own, we can connect them into one nearly automated pipeline with only a few human checkpoints.

Why this approach works

Timing does most of the work. You contact customers exactly when they need the product, not when you happen to feel like advertising.

Deep personalization used to be reserved for large accounts because it was too expensive to do by hand. Now every single outreach can be tailored, regardless of the customer's size.

The leads are warm. Instead of cold-calling random companies, you contact companies with a documented need.

And the system runs around the clock and finds opportunities on its own, while your sales team works the qualified leads.

The difference from two years ago

The level of personalization is what separates this from anything we saw one or two years ago. Back then you could segment your audience into a few broad categories and send a different message to each. Now you can tailor the message, the visuals, and the timing to each individual company, at scale.

Will this work for your business?

Ask yourself one question: can you identify when your potential customers need your product?

If yes, you can save serious hours and get far warmer leads with this approach. Some examples:

• Accountants contacting companies before filing deadlines
• IT security firms reaching out after published data breaches
• Recruiters contacting companies that have announced growth plans
• Catering companies finding businesses with upcoming events

The short version

AI in sales rarely requires building the most advanced technology. It requires redesigning your sales process around:

• Proactively identifying customers with real needs
• Personalizing the message and the material
• Timing the outreach right
• Automating the repetitive steps

Put those together in an agent system and a small company can get sales results that used to require a large sales department.