How to build an AI-powered client intake process that doesn't lose leads

Every professional services firm has the same dirty secret: leads fall through the cracks. Someone fills in your contact form at 9pm on a Tuesday, and by Thursday morning when someone finally reads it, they've already instructed a competitor. Or a referral emails you, gets a "thanks, we'll be in touch" reply three days later, and never hears from you again.

Client intake is where firms bleed revenue — quietly, invisibly, and constantly. The good news: this is one of the most practical, highest-ROI places to deploy AI right now. Not in some theoretical future. Today.

Let's build it.

Why intake is broken at most firms

At a typical 5–50 person firm — whether you're solicitors, accountants, IFAs, or recruiters — client intake usually looks something like this:

Enquiry arrives (email, web form, phone call, referral)
Someone (often a partner or office manager) manually triages it
A fee earner gets assigned — eventually
Someone sends an engagement letter or books a call — eventually
The client either waits patiently or goes elsewhere

The gaps between those steps are where you lose people. And the problem isn't laziness — it's that fee earners are busy doing billable work, and intake administration isn't anyone's dedicated job.

AI fixes this not by replacing human judgement, but by eliminating the dead time.

The four-part AI intake workflow

Here's a practical architecture you can implement with tools that exist today. No custom software development required.

1. Instant acknowledgement and qualification

When an enquiry hits your inbox or web form, an AI agent (built in something like Make, n8n, or Microsoft Power Automate) immediately:

Sends a personalised acknowledgement email within 60 seconds. Not a generic autoresponder — an AI-drafted message that references the specific enquiry. ("Thanks for getting in touch about the lease extension on your property in Brighton. Here's what happens next...")
Classifies the enquiry by service type, urgency, and estimated value using an LLM like Claude or GPT-4o. You give it your service categories and a few examples; it learns fast.
Flags conflicts or out-of-scope requests so a human can deal with those separately.

The client feels seen. That alone puts you ahead of 90% of firms.

2. Smart information gathering

Instead of waiting for a first meeting to discover you're missing half the information you need, the AI sends a tailored intake questionnaire within minutes of the initial enquiry.

The key word is tailored. If someone's asking about probate, they get probate questions. If it's a recruitment brief, they get role specification questions. You build a library of intake templates (one per service line) and the AI selects and sends the right one based on its classification in step one.

Tools like Typeform, Tally, or even a simple Google Form work fine here. The AI just picks which one to send and personalises the covering message.

3. Automatic routing and scheduling

Once the questionnaire comes back, the AI:

Summarises the responses into a brief for the fee earner (no one wants to read a raw form submission)
Routes it to the right team member based on rules you set: service area, capacity, seniority, location
Sends the client a Calendly or Microsoft Bookings link for the appropriate person — pre-filtered to available slots

The client goes from "I filled in a form" to "I have a meeting booked with the right person" without a single human touching the process. The fee earner opens their calendar, sees the meeting, and reads a one-page AI-generated brief. Done.

4. Pre-meeting preparation

This is where it gets genuinely powerful. Before the first meeting, the AI can:

Run a Companies House lookup if it's a business client (the API is free)
Pull the client's LinkedIn profile summary
Check your CRM for any prior relationship or touchpoints
Draft a preliminary engagement letter or scope of work based on the intake questionnaire responses
Prepare a fee estimate range based on your historical pricing for similar work

The fee earner walks into that first call more prepared than they've ever been — and the client notices.

What you need to build this

Let's be specific about the tech stack:

Automation platform: Make.com (formerly Integromat) is the easiest for non-technical teams. Power Automate if you're a Microsoft shop. n8n if you want self-hosted.
LLM: Claude (via API) for drafting and classification. It's better than GPT at following nuanced professional instructions, in my experience. Budget roughly £20–50/month for a small firm's intake volume.
Forms: Tally (free) or Typeform. Keep it simple.
Scheduling: Calendly or Cal.com. Microsoft Bookings if you're already in that ecosystem.
CRM: Whatever you already use. Most automation platforms integrate with HubSpot, Salesforce, Pipedrive, or even a spreadsheet.
Email: Your existing email. The automation sends via your domain so it looks like it came from your team.

Total cost for a small firm: £50–150/month in tools, plus a day or two of setup time.

The numbers that matter

Firms I've seen implement this kind of workflow typically report:

Response time drops from 24–72 hours to under 2 minutes
Conversion from enquiry to booked meeting improves by 30–50% (speed is the single biggest factor in winning new instructions)
Fee earners save 3–5 hours per week on intake admin
Data quality improves dramatically because the AI asks the right questions upfront

That last point matters more than people realise. How many first meetings have you sat through where you spent 30 minutes gathering basic information you could have collected in advance? That's not a good use of anyone's time — yours or the client's.

Where to be careful

A few things to watch:

Don't pretend the AI is a human. Your acknowledgement email can say "I'm Ada, the AI assistant at Smith & Co" or simply be transparent that it's an automated but personalised response. Clients don't mind efficiency. They mind deception.

Keep a human in the loop for edge cases. The AI should flag anything it's uncertain about rather than making assumptions. Conflicts checks, for example, should always have human sign-off.

Review your AI's output weekly for the first month. Check the classification accuracy, the quality of the drafted emails, the brief summaries. Tweak your prompts. After a few iterations, it'll be solid.

GDPR applies. You're processing personal data, so make sure your privacy notice covers AI-assisted intake. If you're using Claude or GPT via API, the data isn't used for training — but document this in your records of processing.

Start small, then expand

You don't need to build all four stages at once. Start with stage one — the instant, intelligent acknowledgement. That alone will make a measurable difference to your conversion rate within a week.

Then add the smart questionnaire. Then routing and scheduling. Then pre-meeting prep.

Each stage compounds. By the time you've got the full workflow running, you'll have an intake process that's faster, more thorough, and more professional than firms ten times your size.

And the best part? Once it's built, it works at midnight on a Sunday just as well as it works at 10am on a Monday. Your competitors are still checking their inboxes over Monday morning coffee. You've already booked the meeting.