How Lawfully Hit an 85% AI Resolution Rate in Under Two Months After Switching from Intercom

1. About Lawfully

Lawfully started as a way to help immigrants track their own cases and help law firms manage them more efficiently.

Today, the Lawfully Case Tracker app serves 2.7 million B2C users, alongside a growing B2B platform for law firms, with plans to expand beyond immigration into other areas of law.

Rebecca Park, B2C Lead at Lawfully, oversees sales, marketing, customer service, and product for the consumer side of the business, including a marketplace built into the app.

2. The Challenge: Intercom's Fin Wasn't Resolving Anything

Before Channel Talk, Lawfully ran support on Intercom with just two CS agents handling everything, maxing out at around 200 inquiries a week between them.

Costs climbed with volume, not despite it.

Lawfully was paying thousands a month, but the AI layer wasn't earning its keep.

Fin's resolution rate hovered around 1%, and the CS agents ended up turning it off themselves, it simply wasn't holding up.

The stakes made the gap worse, not better. Lawfully's users are often anxiously tracking green card or visa cases, a segment where USCIS gives little visibility and Lawfully is often the only place customers can see any real detail.

With an AI that couldn't reliably help, and couldn't be trusted to know when to step back, the only lever left was hiring more people, indefinitely, to keep pace with growth.

3. Evaluating Alternatives and Channel Talk

Rebecca's team didn't take the first alternative that came along. They compared several AI-support tools, Intercom's Fin, Crisp, and Gorgias among them, before landing on Channel Talk.

Given the stakes of immigration support, an unreliable AI isn't just a bad experience; it's a liability.

So instead of switching outright, Lawfully ran Channel Talk and Intercom side by side for a full month before fully committing.

The bar for success was specific: the AI didn't need to solve everything; it needed to know when not to.

AI Agent ALF needed to be able to understand when an inquiry should be handed to a human agent or not.

The proof came fast; within days of testing, the team could already see ALF handling the volume and knowing when to hand off.

"Not even in a week, in three days you can see the result of ALF.

If you don't like how AI is taking care of it, you can turn it off. But you can see the result in three days." — Rebecca Park

4. Setting Up AI Agent ALF

Two things anchored Lawfully's configuration:

1. Catching emotional or high-complexity conversations early

2. Staying firmly out of legal advice

ALF is set up to recognize when a customer is frustrated, or a question is beyond what it should attempt, and hands the conversation to a human agent without being asked.

"ALF can catch if the customer is frustrated or not. And if ALF thinks, 'I think this one I cannot solve right away, or this person is way too frustrated,' then it's going to automatically shift it to a human agent. I was surprised." — Rebecca Park

Given the liability of immigration advice, Lawfully was explicit that ALF should never give legal guidance; two months in, there have been zero incidents on that front.

An example of ALF handling customer inquiries carefully:

Lawfully also chose to run ALF without pre-set decision-tree menus, just an open chat box.

Counterintuitively, that drove more conversations, not fewer, since customers were willing to type out real questions instead of hunting through canned options and ALF absorbed the extra volume without added headcount.

The full migration took two to three weeks, helped along by the hands-on support from the Channel Talk team.

This type of support Rebecca contrasted directly with her prior experience on Intercom's enterprise plan.

"We were on the Enterprise plan with Intercom but they did not support us the way Channel Talk has." — Rebecca Park

5. The Results

By the numbers, this month alone:

• 85% resolved by AI Agent ALF

• 43% handled by ALF

• 50%+ reduction in CS costs

The number of inquiries handled by ALF is roughly what one full-time human agent used to resolve in a week under the old setup, now happening automatically, alongside everything the remaining human agent handles.

Rebecca didn't expect it to work this well.

"I thought the maximum resolution rate would be 30%, that's a realistic number. But I checked the inquiries myself and it's truly 85%." — Rebecca Park

And on cost savings:

"For the cost reduction, we have more than 50% cost saving after Channel Talk." — Rebecca Park

With ALF handling the bulk of routine inquiries, both remaining team members have expanded what their work looks like.

One agent now spends his time on marketing and community efforts instead of ticket volume, a career expansion made possible by ALF taking on his old workload.

The other has shifted from just working through a queue to actively managing her own performance, reviewing her resolution-time analytics and setting her own improvement goals.

6. What's Next

Lawfully is extending ALF to email, its next-highest-volume channel after chat, and exploring Channel Talk's marketing features to turn the chatbot into a lead and sales tool as well as a support one.

Rebecca's advice to other teams in high-stakes industries considering AI is direct:

"AI quality is getting better and better. I'd understand if you complained about AI last year, I've been there too.

But I feel like this year is the starting point where AI is making a real impact. Just test it out for two or three days, and you'll see the result." — Rebecca Park