Human insight. Machine precision.

Setting up a new real-estate investor meant creating every record by hand: the client, their companies, their properties and their team, all linked together. One typo could file a property under the wrong company and cause a costly bookkeeping mistake months later.

The client fills in one form. It creates their records, their Drive folders, a Chat space for the team, an onboarding task in ClickUp, and a follow-up asking them to book their call.

Every client is set up the same way, from one checked source. Properties can no longer be filed under the wrong company, and each onboarding saves the team 15–25 minutes.

A team of about twenty keeps everything about its clients in Notion: their companies, properties, accounts, call notes and history. Clients change hands all the time. A new client is assigned to someone, or a teammate covers a client while the person who normally handles it is on leave or out sick. Each time, the new person had to dig through Notion page by page, opening the client, then each company, each property and the call notes, before they could do any work. That took thirty minutes to an hour, or meant interrupting a colleague who might not be there to ask.

An assistant the team messages in Google Chat, in their own words: “tell me all about this client.” It replies straight away so the chat doesn’t time out, then posts the full picture in the same thread. It reads Notion live, following the links from a client to their companies, properties and call notes, the same path a person would click through, and it reads the firm’s how-to guides for “how do we…” questions. Every fact links to the page it came from. If it can’t find something it says so instead of guessing, and if a name could mean two clients it asks which. Only the company’s own Google accounts can use it, and the firm’s own books stay off-limits.

Picking up a client now takes a few minutes instead of up to an hour, and nobody has to track down the person who’s away. The handoff no longer depends on who is in that day. The answer can be checked in one click, because every line links to its source. It answered 24 real questions in its first four weeks, and 21 came by direct message, so it was made direct-message only. It answers questions and doesn’t change records, on purpose.

Writing up calls was the easy part. The hard part was making sure each summary went to the right client, and never into another client’s workspace. If the system wasn’t sure, it had to stop rather than guess.

After each call, the system looks at the meeting title, who attended, who spoke and what was said, and uses all of it to work out which client the call belongs to. Nothing is sent until it’s confident.

About 15 seconds after a call ends, the client gets a recap email, the team gets a briefing in Chat, the notes are filed in Notion and the tasks are assigned in ClickUp. If the system isn’t sure which client it was, it sends nothing and asks a person to confirm.

Unpaid invoices slid to 30, 60 or 90 days because chasing them meant writing an awkward email, remembering to follow up a week later and keeping track of who promised what. So it got put off, and money the business had already earned didn’t come in.

A day after an invoice is due, a friendly reminder goes out, then a firmer one a week later, then an offer to split the payment. AI writes each reminder to suit the client, but the amount, due date and payment link are copied straight from the invoice. Before every email it checks again, so a paid, voided or paused invoice is never chased, and a client’s reply pauses the chase and tells the owner what they said.

Every invoice ended where it should: paid ones closed with a receipt, replies paused the chase, and unpaid ones went to the owner after three reminders with a Slack message asking them to call.

The business wanted an event funnel that could take sign-ups, text people about the event and take payment. Building the pages is the quick part. In the US, a business has to be registered with the phone carriers before it can text customers, and unregistered texts show no error: they just quietly fail to arrive for some people. The domain, the email setup and what happens to failed messages can break a launch just as quietly. And once it was running, someone was still texting reminders for the community’s two recurring events by hand, two to three hours every week.

Built in the order it has to happen. First the approvals: carrier registration for texting (A2P 10DLC) and the email settings that keep messages out of spam. Then the business’s own domain, so pages, links and replies all run on it. Then the CRM with a four-stage pipeline, the funnel with a thank-you page for each ticket type, and a short form that sorts sign-ups before anyone gets on a call. Finally, twenty-one automations. One group runs the event. The other removes anyone whose messages bounce, who complains or who opts out, so they drop out of every follow-up at once. After launch, the event reminders run on their own: members are tagged, and every Wednesday the account checks whether it’s the third week of the month and texts them in batches of 25, so the number stays in good standing.

The launch went out and it converted: fifty sign-ups in the two weeks around the opening, seventeen of them paying clients. The texting registration was approved first time, in about three weeks. Twenty kinds of message failure are caught and handled automatically, so the contact list keeps itself clean instead of slowly going bad. The weekly reminders send themselves, and because nobody picks recipients by hand, the wrong person can’t get the text.

Every client had their own spreadsheet in their own folder, around twenty-five tabs of trackers, checklists and personal notes, with no two set up the same way. Finding one client’s bank details meant knowing which folder, file and tab to open, and only the longest-serving staff knew. New starters couldn’t work on their own, clients were asked again for access they’d already given, and document requests went to the client instead of the assistant who already had them.

One shared workspace, with a home page and a client database. Each client has one record, and their companies, properties, bank accounts, vendors, team and meeting history are all linked to it, so you can start from a property and find its company and client, or the other way round. Company names match the bookkeeping software and the Drive folders. For each account, it records whether the team has the main login, limited access or none. The client’s own staff are listed separately, with who handles what. Thirty-plus files were moved over in two weeks: by hand at first, then with a custom tool.

One place to look, with every answer attached to the right client. Twenty people use it, and new starters no longer learn how each spreadsheet was organised, because every client is set up the same way. There’s no time-saved figure here: the real cost was interruptions and repeat requests, and those can’t be timed with a stopwatch.