WE LET AI LOOSE ON A CLIENT'S CRM. HERE'S WHAT CHANGED.

Most business owners are still using AI as a writing assistant.

The thing that genuinely changed how I think about this happened last week, inside one of our client's accounts. Not because of some new tool. Because of what AI actually did.

After 24+ months of building AI systems for 60+ businesses, and 171,000+ real customer conversations later, here's the shift most owners haven't fully internalised yet:

AI is no longer just helping with the work. It's starting to do the work.

THE REAL EXAMPLE

We were helping a client this week who'd just opened installer coverage in a new area near Wollongong.

She knew people had inquired about that area in the past.

The problem was that the records weren't clean.

Some contacts had the suburb saved properly on the contact record.

Some had mentioned the area in an old email thread.

Some had typed it into a live chat conversation 6 months ago.

The information was there.

It just wasn't searchable in any useful way.

Old way to solve this: someone manually digs through the CRM for half a day.
Reads old emails. Skims chat history.
Tries to remember who said what.

What we did instead: we used the "Ask AI" feature in the Earlibird app.

We asked it to find any contact who had inquired about the new coverage area.
The AI swept across more than 1,400 contact records, including full conversation history, emails, SMS, and chat transcripts.
In about 30 seconds it came back with 5 confirmed matches.

Then it expanded to 6 after picking up additional context buried in another conversation thread.

The 6 contacts it surfaced:

• One tagged as a hot lead with intent to purchase, sitting in an active live chat conversation that morning

• Two who had specifically requested a free measuring quote (high purchase intent)

• Three others who had asked about the area in prior conversations

THE OUTREACH

Then we asked the AI to draft outreach.
Email and SMS, in her voice.

This client's writing style is direct and blunt.
No fluff, no padding, no corporate softening.
It's basically the opposite of how AI usually writes by default.

The AI read her prior conversation history, picked up the tone, and produced drafts that actually sounded like her.
Direct. Clear. Specific to each contact's prior conversation context.
The only manual edit we had to make was removing the em-dashes (the giveaway that the text came from an AI).
That single rule is now baked into our voice guidelines so it doesn't happen by default anymore.

THE DETAIL THAT MADE ME STOP

One of the 6 contacts didn't have a phone number saved on the contact record.
He'd originally come in through a Facebook DM, and the standard flow had captured his email but not his phone.
Normally, the SMS would just not send.
The AI would skip the SMS step and just send the email.

Instead, it searched conversation history.
Found the mobile number sitting in an old email signature from months earlier.
Added it to the contact record. Sent the SMS.

That's not a feature we built explicitly.
It's just what the AI did when it had access to the full context.
That's the moment that stuck with me.

WHAT IT FOUND IN OUR OWN DATA

We then ran the same "Ask AI" feature against our own database, looking for hidden trends.

What it surfaced:

• "Qualified but didn't book" is our single biggest addressable gap. Not "didn't qualify." Qualified, then didn't book. Different fix, much higher leverage.

• A natural re-engagement window in early April, identified by reading hundreds of out-of-office reply timestamps across our nurture emails.

• Multiple-decision-maker scenarios are a silent barrier in our sales conversations, which gave us a specific change to make in the discovery flow.

• Our nurture sequence and our sales sequence look too similar and need to diverge.
  

None of that came from a dashboard.
It came from the AI reading the actual conversations, emails, and replies, and surfacing patterns nobody on the team had spotted.

THE SHIFT TO INTERNALISE

In a training I ran recently, I said work is no longer just something you do.
It's something you design.
You design the system, you design the flow, you design what the AI handles and what the team handles.

This is what that looks like in practice.
AI isn't replacing the sales team.
It's doing the part of their job that they didn't have time for anyway: finding what matters in the data they already had, drafting outreach grounded in real context, and surfacing trends nobody was looking at.

The businesses getting real value from AI right now aren't the ones with the most subscriptions or the latest tools.
They're the ones who've put their information in one place, cleaned up the basics, and started asking AI to use that information to help them make better decisions.

THREE THINGS TO DO THIS WEEK

1. CAPTURE MORE INFORMATION IN ONE PLACE.

Phone calls, emails, SMS, DMs, live chat, contact form submissions.
If it's spread across 8 different tools, the AI can't help.
Pick one location and start centralising.

2. CLEAN UP THE BASICS.

Tags, pipelines, notes on contact records.
Doesn't have to be perfect overnight.
Messy data makes AI dramatically less useful.

3. STOP ASKING "DO WE HAVE AI?"

Start asking "how can AI use our information to help us make better decisions, faster?"
That's the better question.

WHAT NOT TO BELIEVE

There's a lot of noise in the AI space right now.
"I built 50 AI agents and now I'm running a $25 million business while making $5 million a week."

Most of that is marketing at best and outright lies at worst.
The practical applications are far more grounded:
AI handling the structured work in your business so your team can focus on the high-judgement work.
Faster follow-up, better decisions, less wasted time, more value from information you already have.

If you want help thinking through how this applies in your business, book a call and I'll walk you through it personally.

Book an AI Review Call → earlibird.ai/schedule-ai-review-call