Type the client in once. Every tool that needs it follows.

Somebody at your shop typed the same client into a fifth tool this week. CRM, board, Friday timesheet, invoice, notes doc.

Nothing connects them, so a person does. A new client is five setups, a scope change is five edits, and one of the five stays wrong until the client points it out. It gets worse with every client you sign.

I build the connection, inside the tools you already pay for. No new platform, no new hire.

Get the free audit

William Jin, software engineering student at Waterloo. Built the policy search 17,000 federal staff use. I write the code myself.

This is most agencies. It's probably yours.

Agencies aim for 70 to 80% billable and call 85 to 95% realization healthy. So a fifth or more of every week isn't billed, and 5 to 15% of the hours that are logged never reach an invoice.

That is where the day a week goes.

The hours are the smaller cost. The person doing the retyping is usually your best account manager, and it is the part of the job good people quit over. A client typed for the fifth time is typed a little worse, by someone who stopped reading closely around the third. The hour you spend fixing it is an hour that didn't go to a pitch, a referral, or home.

Pick the rows that happen at your shop.

By hand, todayOnce it's built

A new client is five setups. The CRM, the board, the timesheet, the invoice, the doc.One setup. The other four are created from it.

A scope change is five edits. One gets missed until the client points it out.One edit. The rest follow, so there is nothing to miss.

Month end. Ad accounts into a spreadsheet, fixed by hand, pasted into the deck. One client gets last month's number. 94% of audited spreadsheets have at least one error, and this is a spreadsheet.The report builds itself on the first, in your format, from this month's numbers.

The invoice comes from a timesheet someone filled in on Friday from memory. The hours don't match the SOW, and someone eats the difference.The hours come from the board. The invoice is built from them.

The proposal is last quarter's with find and replace, and one old client name survives.Built from the scoping call notes. No old names.

The zap between the CRM and the board worked, until someone renamed a field. It failed for three weeks before anyone noticed.When something breaks, it says so, the same day, to a person.

Who I've helped

Two builds, both still in use. What the team did by hand, what they do now, what I built, and the numbers.

Automated policy search for federal staff. 6 to 7 hours a week back per person.

Government of Canada · Policy chatbot · 2026

17,000

staff who ask instead of reading

6 to 7 hrs

a week per person that used to go to reading directives

220

policy documents searched in one go, some of them hundreds of pages

The story

A federal department runs on 220 policy directives. Some are a page, some run to hundreds of pages, and a few are old PDFs that barely open.

When someone needs the rule, they find the right directive and read through it by hand. 17,000 staff work under those directives, and the people who look things up most were losing hours a week to it.

What they do now

They type the question and get the answer, with the passage it came from.

The people who used to do the most looking up get 6 to 7 hours a week back. Still in use today.

What I built

Two human steps, ask and read. The search between them runs itself.

a person does thisruns on its own

Ask
A question in plain words.
Search
The right passages are found across all 220 directives.
Answer
An answer in plain words, with the passage it came from.
Read
Check the source if you want to. Rate the answer if it's wrong.

Runs in the department's own accounts. Still in use today.

Automated citation tracking for 7,000 papers and the search through 5,000 experiment files.

University of Alberta · Research lab · 2025

200 hrs

of searching by hand that nobody does now, at two minutes a paper

7,000

papers tracked without anyone searching

5,000

experiment files that answer a question in plain words

The story

A research lab needed to know who was citing 7,000 papers, and where. Each one was a search by hand, about two minutes a paper, so 200 hours of someone's time.

The lab's own experiment files, 5,000 of them, sat on one computer and were found by clicking through folders.

What they do now

The lookups run on their own, and what comes out is a map of who cites what, by region, that the lab uses in its talks.

The experiment files are in one place, and a researcher can ask in plain words: what did this run show, how does today compare with last week. Still in use today.

What I built

A person at each end. Everything between the folder and the answer runs itself.

a person does thisruns on its own

Hand over
The list of 7,000 papers and the folder of experiment files. That is the whole input.
Look up
Every paper is looked up and every citing author pulled, the way a person would, without the person.
Load
The citations and the 5,000 experiment files go into one place.
Map
Who cites what, drawn by region.
Ask
Put the map in the talk. Ask the files a question in plain words and get the answer.

Built at the lab, left with the lab. Still in use today.

Start with the free audit →

Why two builds outside your industry count

Neither client is an agency. Both had the same problem you have, at a bigger scale:

Same problem, bigger numbers

17,000 federal staff found the rule by reading 220 directives by hand. A lab searched 7,000 papers one at a time, 200 hours of someone's time.

Yours is one client typed into five tools. Smaller, which is why it is a smaller build.

Still running, without me

Both are in use today, in the client's own accounts, with no retainer. Yours will be too.

Nobody learned a system

Staff type a question. Researchers ask the files. Your team keeps every tool it has, and the typing between them stops.

What gets built first at an agency

The account side, mostly. The jobs that repeat every month and eat billable hours. Bring one to the audit. I have room for a few more this month before I'm heads down on builds.

Client onboardingone intake feeds the CRM, the board, the contract and the kickoff

Scope changesone edit, and every tool that carries the scope follows

Proposals and SOWsassembled from the scoping call notes

Monthly client reportspulled from the ad accounts and analytics, in your deck, on the first

Hours to invoicewithout the Friday timesheet in between

The unbilled checkwhich projects logged hours this month and have nothing invoiced, run every week instead of never

Chasing approvalsthe reminder goes out without an account manager remembering

How it works

Three steps. You see each piece running before you pay for it, and you keep everything at the end.

Audit

A 30-minute call. You share your screen and walk me through one normal week: where the client gets typed, where the report comes from, where the invoice starts. I don't need access to anything.

I map the chain and count the hours per job. You get one page back: what to automate first, what it saves a week, what it costs.

Free, and yours to keep if you stop here.

Build

Scope and milestones in writing before any code. The first milestone is the job with the most hours in it, so something useful is running early.

I build inside the tools you already pay for, in your accounts. Your data never leaves them. Every week you get a screen share of what runs now.

You pay after each milestone runs. Nothing hourly, no retainer.

Handover

You get the code, documentation written for whoever runs it, and a walkthrough on a call. Then my access is removed.

It keeps running without me and without a retainer. If it breaks because of something I got wrong, I fix it. If a tool changes on your side, I'm one email away and you decide whether that's a new milestone.

Three rules I work by

Accountability

Milestones agreed up front, paid after each one runs and does what we said.

Weekly check-ins, so nothing drifts for a month.

Scalability

Built for the whole team. When the team grows or a tool changes, the automation keeps working and is easy to extend.

You get the code, so nothing depends on me.

Utility

The build has to be useful, or it doesn't ship.

If the piece we scoped turns out not to be, or we hit a blocker, we rescope on the spot and you keep what's already built.

Start with the free audit →

Why not an automation agency, a consultant, or one more hire?

Each is a fair option. Here is what each one costs you.

An automation agency

Their smallest project is bigger than yours, and the retainer runs whether or not anything ships. The person who sold it hands you to a project manager, who hands you to a developer you never meet, and the follow-ups go quiet.

Here you pay per milestone, and the person on the call is the person who builds it, checks in and answers the email.

A consultant

You get a recommendation. You still need someone to build it, and then a second vendor to manage.

The audit is the recommendation, free. The build is the next step with the same person.

One more hire

Onboarding first, then a person does the typing faster. The typing is still there, by hand, with about one error per hundred fields, and it grows with the client list.

Once the job runs itself, the people you have get their day back, and the work they keep is the work that needs a person.

Questions you probably have

Short answers. Ask me the long ones on the call.

I don't know what I'd automate yet. Is it too early to talk?

No. That's what the audit is for. Bring the job that annoys you most and we start there.

What does it cost?

The audit is free. The build is quoted per milestone, in writing, on the page you get back from the audit. You pay after each milestone runs. Nothing hourly, no retainer.

Where does our client data go?

It stays in your accounts. I build inside them, copy nothing to my own systems, and my access is removed at handover.

Do we need new software?

Usually not. Most builds run on what you already pay for. If something new is better, I'll say so and you decide.

We have an IT person, or a compliance consultant. Do they get a say?

Yes. They see everything as it's built, and the documentation is written for them.

What if it breaks after you leave?

You have the code and the documentation, so anyone technical can fix it. If it's something I got wrong, I fix it.

How long does the first build take?

Depends on the job. The one page from the audit has a date on it, and the first milestone is the job with the most hours in it.

About

I'm William. I'm a software engineering student at the University of Waterloo, and Wren Automation is the business I'm building.

My work so far has been internships: data cleaning, pipelines that connect systems, and CRMs. The two builds above came out of that work.

I want to get very good at this. The fastest way is to do it for companies that have the problem, and do it well enough that they tell other companies.

Why me

I'm trained as a software engineer and I write the code myself, so it holds up after the demo. Most automation agencies hand you something beautifully packaged, with bugs and shortcuts that show up in month three.

I'm building a business on this work, and I want every build to be so good you refer me to every owner you know. I'll relax the financial terms to get that result, and I'll say so on the call.

Thanks for taking the time to read this far.

Get the free audit

Tell me what eats your week. I'll reply within a business day with a few times for the call.

Got it, thanks.

Look for an email from william@wrenautomation.com within a business day. If it isn't there, check spam or email me directly. Talk soon.