Test Your AI Answering System the Way Clients Actually Use It: One Hand, One Phone

Today's tip: run your own intake on a phone, one-handed, on an average connection

Not on your desktop. Not on office wifi with two monitors and a keyboard. Pick up a normal phone, hold it in one hand the way a person holds a phone while walking out of an emergency room or standing in a parking lot next to a wrecked car, and run the entire AI Answering experience end to end as if you were the prospect.

Then write down every place you hesitated.

That's the whole exercise. It takes fifteen minutes and it will tell you more about your intake conversion than a month of reviewing call recordings, because the people who review intake systems almost always review them from a desk — and the people who use intake systems almost never do.

Why this matters more than it sounds like it should

Scorpion's 2025 legal-consumer report found that 72% of prospects would move on if a firm did not respond within 24 hours. That number gets quoted constantly as an argument for faster response times, and it is. But there's a second reading of it that firms miss.

A prospect who abandons your intake flow halfway through — because a button was ambiguous, because nothing happened for four seconds after they tapped, because they were asked to type a paragraph on a cracked screen — has functionally become a non-response. You never got their name. You never got a callback number. There is nobody to respond to. From your CRM's point of view, that contact never happened. From the prospect's point of view, your firm was the one that didn't show up, and they are now inside that 72% moving on to the next firm on the results page.

Every prospect who drops out of a mobile intake flow becomes a non-response you can't measure, can't follow up on, and will never know you lost.

This is the quiet version of the missed-call problem. Firms invest heavily in never letting a phone ring out — and then lose the same case to a form field that was hard to tap.

The four failure patterns to look for

When you run the test, you're hunting for four specific things. Name them out loud as you find them, because vague notes like "it felt clunky" don't get fixed.

1. Measuring calls without outcomes

This is the reporting failure that hides all the others. If your dashboard tells you how many calls came in but not what happened to each one — retained, referred out, wrong practice area, hung up at second nineteen — you don't have intake data, you have a tally. Volume without outcomes means you can't tell a good week from a bad one, and you certainly can't tell which step of the flow is leaking.

Ask the question directly: for every call that came in last week, can I see the disposition? If the honest answer is "sort of," start there.

2. Unclear buttons

On a desktop, an ambiguous label is a mild annoyance — you hover, you read the surrounding context, you figure it out. On a phone held one-handed, an ambiguous label is a stop. Does "Continue" mean continue the call or continue to a form? Does "Submit" send this to a human or into a queue? Is that grey text a button or a caption?

The standard is simple: a stressed person should know what every tappable element does without reading anything else on the screen.

3. Delayed feedback

Tap, then nothing. No spinner, no state change, no confirmation. On a good connection this is a half-second of uncertainty. On an average one — which is what you're testing with, deliberately — it's three or four seconds of a person wondering whether it worked, then tapping again, then getting a duplicate submission or an error.

Every interaction needs an immediate acknowledgment that something is happening, even if the actual work takes longer. Silence reads as broken.

4. Unnecessary typing

Typing on a phone is the single most expensive thing you can ask a prospect to do, and most intake flows ask for far more of it than they need. Open-ended "describe your situation" boxes, address fields with no autocomplete, dates entered as free text, anything that requires switching keyboards to reach a number or a symbol.

Every field should justify itself. If it doesn't change what happens next — whether you take the case, who calls them back, how fast — it doesn't belong in the first interaction. Get it on the follow-up call.

What you actually need to capture

Cutting friction doesn't mean cutting information. It means being ruthless about which information has to be collected in that first contact versus which can wait. The non-negotiable set is short:

• Name — so the callback isn't cold.
• Callback number — the single most important field in the entire flow. If you get nothing else, get this.
• Email — your second channel when the phone goes unanswered.
• Reason for calling — enough to route the matter to the right attorney and rule out conflicts or wrong-practice-area calls.
• Consent-appropriate notes — whatever your jurisdiction and your recording and privacy obligations require, captured correctly the first time rather than reconstructed later.

That's it. Everything else is a second-conversation question. A flow that captures those five things reliably in ninety seconds beats a flow that captures fifteen fields from the 40% of people who make it to the end.

Why desktop reviews miss all of this

The reason this friction survives is structural, not careless. The people who evaluate intake systems — firm owners, office managers, marketing directors, the vendor's own demo team — evaluate them on large screens, with fast connections, using a mouse, already knowing what every button does because they helped design the flow. Under those conditions the system works fine. It genuinely does.

Meanwhile the actual user is outside, on cellular, one-handed, distracted, often upset, and has never seen your interface before. Those are two completely different products. Only one of them generates revenue.

Running the test yourself on a phone is the cheapest way to collapse that gap. No analytics tool will tell you that a button label is confusing. You have to feel the hesitation.

Your next step

Block fifteen minutes this week. Phone only, one hand, step outside so you're on cellular. Run the full flow as a prospect. Write down every hesitation, then sort them into the four buckets: outcome tracking, button clarity, feedback timing, typing burden. Fix the typing burden first — it's usually the largest single drop-off and the easiest to cut.

Then repeat it monthly. Intake flows drift as vendors ship updates and staff add "just one more field."

If what you find is that the underlying system can't support fast, low-friction capture no matter how you rearrange it, that's a different conversation — one about the infrastructure underneath. You can see how we approach it on the Bosseo AI Answering page.