Your clients came to WPUP to have their sites maintained, and that work is still there. This is about what WPUP can sell on top of it. Here are two services we already deliver, with the examples behind them and a way for WPUP to offer both under its own name.
A few months ago the ticket queue was a good deal busier than it is today. The clients are still there and their sites still need looking after. What has slowed is the number of new sites coming in, and a maintenance service only grows when new sites arrive.
New services change that in two ways. The clients WPUP already has can buy more than upkeep, and WPUP has something new to open a conversation with when it talks to a business that is not a client yet.
Most of your clients have no developer of their own, which is why they need WPUP for maintenance. The same clients also have work they repeat by hand every week and search visibility they are losing to AI answers. Nobody is offering to fix either for them. WPUP can.
Google AI Overviews, ChatGPT, Perplexity and Gemini now answer the informational questions that used to send visitors to a site. The pages that survive are the ones the engines cite. We write and rebuild content for that, measure which prompts cite the client against named competitors, and report what the engines get wrong about them. It is a monthly program with a baseline, a six-week read and a binding decision at twelve weeks.
Impressions on the pages our articles link to against the un-linked pages on the same site, latest 28-day reading. The gap has held in three consecutive readings, and the control fell while the linked set rose, so season does not explain it.
Of 6,704 Copilot citations land on articles, not on the commercial route pages. Articles also take 89% of organic clicks while 102 route pages share the rest. Articles earn the citation, service pages collect the visit.
Of 525 engine answers cited the client at the recheck. 17 of 18 pieces moved from 0 to 3 citations at baseline to between 2 and 16 of 30 checks each. First sample of three, so a method shown, not yet a graded result.
AI answers and Google AI summaries recorded and scored with their sources, so every recheck is a comparison and not an opinion. The same instrument produced the LPS report.
Stand-alone guide and data sites built for one venture rank on their own and send inquiries back to it. The one property a competitor cannot buy on a subscription. Opt-in, never assumed.
The WPUP-branded baseline report is ready to share. The short read: LPS is cited in AI answers on local commercial terms, which almost no local gym is, but that authority points at nothing. The combine pages dead-end, the online coaching page sits at position 23 with most of its impressions coming from the United States, and the one measurable local demand (body-composition testing) is owned by a clinic. The program is a rescue month on the pages that sell, then a testing cluster, then a hockey cluster.
We will never promise leads from AI answers, a traffic number, a ChatGPT citation or a ranking. The report shows ratios and states, and the twelve-week decision is written into the plan.
Every business has work that is the same every time: answering the same inquiry, copying a booking from an email into a system, typing up a call, checking which payments came in, building the same report every Monday. We build systems that do that work, connected to the tools the client already uses, with a person approving anything that goes out to a customer.
Clients rarely ask for this by name. They describe a routine that eats their week, and that is where the work is.
Most small businesses pay monthly, often per user, for a stack of tools: one for scheduling, one for chat, one for reports, one for transcription, one to pass data between the others. A system built for the client does those jobs in one place, belongs to them, and costs what the data costs. Our own search and AI tracking replaced a subscription tool and runs at about $13 a month for four markets.
Anything a person does the same way every time, and anything too large for a person to go through at all. We measured 172,971 websites across two countries at about 2,000 an hour and found the 13,048 WordPress sites in one of them. One person checking a site a minute would need about 360 working days. We have also recorded and scored more than 5,000 AI answers with their sources, and matched 4,540 delivery trips to 4,332 orders for one client.
A transfer company with one shared inbox and a small office team. Booking questions, change requests and quote requests arrived mixed in with mail from booking platforms, toll operators and banks, and every reply meant looking the booking up first.
It works around the clock. Every message that arrives is read and labelled in the mailbox, at night and at weekends too, in an inbox that took 753 messages in 60 days. The team opens it in the morning to mail that is already sorted, with replies waiting to be checked, and the website chat answers a customer at midnight the same way it does at noon.
One message in that first run failed and was fixed the same week.
We built these for ourselves first. Each row is one job, how it was done by hand, what the system does now, and what a person still decides.
Open the last contract, save a copy, replace the client's name, address and contact, recalculate the totals and the tax, fix the dates, hope nothing from the previous client is left in.
Pick the client and the type of contract, enter the price. It fills in both parties from the client record, calculates the totals and tax, sets the dates, keeps only the clauses that apply to that type of contract and adds the signature block. Out comes a Word file ready to sign.
Reads it and signs it.
Find the client, look up the next invoice number, type each line, work out the sales tax, export the PDF, check whether hosting or support for that month was already billed.
One sentence is enough: "2.25 hours of development for this client". It takes the next number, calculates the totals and sales tax and produces the PDF. It first shows a trial run with the exact numbers, and it refuses to continue if the client name could mean two different clients.
Approves the trial run and sends the invoice.
Open the bank statement, go down it line by line next to the list of open invoices, mark each one paid.
Reads the bank statement and matches every incoming payment to its invoice. One run recorded 16 payments, including a single transfer that covered ten invoices and two part payments, and left the remaining balance on each. The next run rechecked every payment since February and found two that had never been recorded.
Reads the list of who still owes what.
Nobody notices until the money does not arrive.
Compared the invoice list with the sent mail and found 23 invoices that had never gone out. It wrote the emails to the right contact at each client, taken from earlier invoice mail, attached the PDFs, and added a reminder where an older invoice was still unpaid.
Presses send on each email.
Every month the bookkeeper asks for the receipt behind each card payment. Someone searches the mailbox supplier by supplier and forwards them one at a time.
Reads the bank statement, books each expense into a category by supplier and skips what is already booked (14 new, 42 duplicates in one run). It finds the receipt for each payment in the mailbox and forwards it to the bookkeeping software: 21 receipts in one month, plus a list of the nine payments that have no receipt anywhere.
Chases the nine.
Take notes during the call, or rely on memory, then write the tasks up afterwards. Half of them are lost.
Reads the transcript of every recorded call and files each agreed action as a task under the right client. Each task is graded by how clear the recording was, and an unclear one comes back as a question to ask the client, not as a task somebody acts on.
Confirms the unclear ones.
Read everything, work out what is a request, remember to do it. Email took 18 hours a quarter, measured with a time tracker.
In 60 days the inbox took 549 messages, 196 of them from machines. It drops the machine mail, files one request per conversation under the right client and drafts the reply.
Works from a list of requests, not from an inbox.
Nobody could say which shop visits cost more than they earned. We joined 4,540 GPS trips from the delivery fleet to 4,332 orders from the order system, so the owner sees cost against revenue per shop and per route.
Search and AI visibility figures are pulled from four sources and assembled into a report on the first of the month. Checking the same things by hand measured 14.8 hours a quarter. The data costs about $13 a month for a four-market program.
An image generator misspelled product names and redrew the logo. Now a real photo is restaged and the name and price are set as live text, with every set checked by script before it posts. Seven products for $1.11.
Ideas to start a conversation with, written the way a client would meet them: something happens, the system acts, a person gets the result. Each is built from parts we already run and connects to what the client uses today, whether that is QuickBooks, Stripe, HubSpot, Mailchimp or a booking system.
A member's card fails on renewal, or a membership is 30 days from ending.
Writes to the member with a payment link, tries again after three days, and puts anyone still unpaid on a short list for the office.
Renewals that used to lapse quietly are caught, and the office calls five people, not five hundred.
Someone fills in the contact form at 9 p.m..
Answers within a minute from the real price list and timetable, offers the free slots, and writes the lead into the booking system.
The inquiry is answered while the person still has the page open, and staff start the day with bookings, not a backlog.
The phone rings after hours, at lunch, or while everyone is with a customer.
Answers the call in a natural voice, knows the opening hours and the services, checks the team's Google or Outlook calendar, books a free time, and texts the caller a confirmation. Anything it cannot handle goes to a person with a written summary of the call.
No call goes to voicemail, the calendar fills itself, and it replaces a paid answering service.
A candidate or a client emails about a role or a job.
Finds the record in the firm's own system, drafts the reply with the right details, and logs the conversation against that record.
First replies go out the same hour, and nothing lives only in one person's inbox.
A new video or recorded session is uploaded.
Transcribes it, adds captions, writes the summary and chapter list, and drafts an article and an email from it.
One recording becomes a lesson page, a search-friendly article and a newsletter, without a second day of work.
Monday, 7 a.m..
Pulls last week's sales, payments, ad spend and website inquiries from each source and writes one email in plain language: what came in, what failed, what changed and what needs a decision.
The owner reads one page and knows where the business stands.
A quote request arrives by form or email.
Prices it from the firm's own rate card, produces the quote as a PDF, and when the client accepts, turns it into the contract and the first invoice.
A quote goes out in minutes, and the contract and invoice carry the same figures with nothing retyped.
Health, legal or financial data is involved and it cannot be sent to an outside AI service.
Runs the AI model on the client's own machine or server, so documents are read, sorted and summarized without leaving the building.
The same help with paperwork, and privacy rules such as HIPAA in the US or PHIPA in Ontario are far easier to meet when the data stays in house.
We do not quote hours saved before measuring them on the client's own work. The first deliverable of any automation is a before and after, so the client sees the saving in their own figures.
You own the relationship and the invoice. Everything outward is WPUP: the reports, the plugin headers, the staging URLs, the voice in Help Scout. The LPS report is already WPUP-branded. Discovery calls run as WPUP, either with Rinalds on the call as WPUP's technical lead, or with you running the questions below and sending notes.
A 20-minute call, then a written list of what repeats, what it costs today and what the fix is. Fixed price, 4 to 8 hours of our time. Also where we say no: if checking a machine's work takes longer than doing the task, it stays manual.
Each item quoted as a fixed price from the audit. The before and after measurement is part of the build.
Anything that runs on a schedule carries a small retainer: model and data costs, a review loop, and maintenance when a model or an API retires. All four lines quoted up front.
Our published list, the partner rate to WPUP, and the band partners resell in. You invoice your client in CAD, we invoice WPUP in USD as now. Setup halves at six months and is waived at twelve on the top two tiers.
| Tier | Monthly capacity | List, USD | To WPUP, USD | WPUP resale band |
|---|---|---|---|---|
| Watch | 40 prompts tracked, report only | 650 | 520 | 650 to 900 |
| Foothold | 4 points of work a month | 1,400 | 1,120 | 1,400 to 2,000 |
| Cluster | 8 points, weekly cadence, 60 prompts | 2,600 | 2,080 | 2,600 to 3,600 |
| Authority | 12 points | 4,600 | 3,680 | 4,600 to 6,400 |
| Setup | one-off, baseline and rebuilds | 2,800 | 2,240 | 2,800 to 4,000 |
| Baseline report | stand-alone, like the LPS report | 1,200 | 960 | 1,200 to 1,700 |
LPS maps to Cluster: a fix month first, then two coverage months, with the twelve-week decision built in. A second language adds 60%.
No proposal before question five. The answers are the audit.
Which two clients would you start with?