New to AI? Get a free, personalized AI plan for your small…

Home New to AI

Tools

New to AI: Discovery AI Readiness Scorecard Build vs Buy vs Subscribe Advisor First 90 Days AI Plan Generator AI Opportunity Finder Process Cost Baseline Calculator AI Subscription Sprawl Auditor

AI Budget Guardrail Builder AI Cost Estimator Model Picker ROI & Payback Simulator ROI Use Case Builder True TCO Calculator

AI Risk Assessment Generator AI Usage Policy Generator Vendor AI Due Diligence Questionnaire Generator

Batch API Suitability Checker Context Window Budget Planner Cost-Aware Routing Designer AI Cost Estimator Model Picker Prompt Caching Savings Analyzer Rate Limit & Capacity Planner Token Counter

CLAUDE.md Generator Few-Shot Example Set Builder MCP Config Wizard System Prompt Generator Prompt Optimizer RAG Chunking & Retrieval Planner ReAct Prompt Converter Skill File Generator Structured Output Schema Builder Subagent & Slash Command Generator Tool / Function Schema Generator Workflow to Agent Blueprint

CLAUDE.md Auditor Eval Set Generator Eval Runner Golden Dataset Manager Grounding / Hallucination Checker LLM-Judge Calibrator Prompt Auditor Prompt Injection & Red-Team Tester Prompt Regression Diff Skill Auditor Agent Trace Analyzer

Data Handling & Redaction Advisor AI Incident Response Runbook Generator

All 47 tools → Features Use cases Live demo Impact analytics Pricing Blog FAQ About

New to AI? Start here.

Answer a few questions about how your business works. Costiva writes a detailed, free report on what AI could do for you, what to try first, and what it could save — then helps you track what actually works.

Create a free account to start Already have an account? Sign in

Free · about 15 minutes · built for businesses with under 100 people

Is this for me?

You run a business with fewer than 100 people.

You’ve heard about AI, or tried ChatGPT or Claude, but don’t know where it fits in your business.

You want practical steps, not hype.

You want to know whether it’s actually worth it.

How it works

Describe your business

In your own words: what you do, who your customers are, where the time goes.

Answer a few follow-up questions

A handful of questions picked just for you, to fill the gaps that matter.

Get your report

A detailed plan with your top 5 ideas, what they could save, and how to start.

Track what you try

Record a “before” number, check in now and then, and see what changed.

What’s in your report

Your plan at a glance

A one-page summary: the big picture, the three things to do first, what it could save you, and the one caution that matters most.

Your top 5: what to try first

Step-by-step instructions for the five ideas we ranked highest, with what they could save you.

Everything AI could do for you

The full list of realistic ways AI could help, across every part of your business, each rated and ranked.

How to tell if it’s working

One to three simple things to count for each top-5 idea, and how to take a “before” number this week.

Your first 90 days

A 30/60/90-day plan: what to do each month, when to check in, and what to decide at day 90.

Staying safe: guardrails and an AI use policy

The risks that matter for the information you handle, simple safe-use rules, and a one-page AI policy you can give your staff.

Bringing your team along

How to introduce AI to your team, who to involve, a simple two-week learning plan, and honest answers to common worries.

AI in plain English

What AI is good and bad at for a business like yours, why it sometimes makes things up, how to check its work, and a short glossary.

Next steps in Costiva

The Costiva tools that fit your plan best, and why.

How we worked it out

The method, the scoring and the honest limits of every estimate — published here.

See a sample report

Sample: a 14-person plumbing and heating company. The business is made up; the scoring and math are exactly what your report uses.

Sample report

Your plan at a glance

A one-page summary: the big picture, the three things to do first, what it could save you, and the one caution that matters most.

AI fits your business best in the office, not the field: replying to customers, writing quotes and chasing payments. These are the jobs that keep you working evenings. None of them needs new software to start — your email, accounting software and a free AI assistant are enough. Expect useful drafts, not perfect ones; a person checks everything that goes to a customer.

Do these three first

Friendly payment reminders — It is almost all set-up and no ongoing effort, and late invoices were one of your biggest annoyances.
Draft replies to common customer questions — Your office answers the same questions all day; drafts save time from the first week.
Turn job notes into a polished quote — It goes straight at your late-night quoting, and you stay in control of every line and price.

Based on your own answers, these starters could free up roughly 3.5–12.5 hours a week, worth about $6,950–$36,000 CAD a year — an estimate, not a promise.

The most important caution: Never let an AI tool give safety advice about gas, carbon monoxide or flooding. Those calls always go to a person.

Your top 5: what to try first

Step-by-step instructions for the five ideas we ranked highest, with what they could save you.

#1 Quick win Bookkeeping and finance

Friendly payment reminders

Your accounting software can send polite, automatic reminders when an invoice is overdue, using wording an AI assistant helps you write.

In your business: Three days after an invoice is due, the customer gets a friendly nudge with a payment link — no more evening phone calls.

Could free up: about 0.5–2.5 hours a week — roughly $1,150–$7,200 CAD a year. Estimate from your answer about “Chasing invoices and payments”, assuming it takes a lot (30–50%) of that job off your plate.

How to try it

Open your accounting software and look for the setting called reminders or payment reminders.
Ask an AI assistant to write a short, friendly reminder in your voice (see the example below).
Paste the wording into the reminder template and add your payment link.
Set the first reminder for 3 days after the due date and a second one at 14 days.
Turn it on for new invoices only for the first two weeks.
Check each Friday which reminders went out and whether customers paid.

What you could type into an AI assistant

Write a short, friendly payment reminder for a family-run plumbing company in Ontario. The customer's invoice is [number] for [service]. Keep it warm and polite, under 80 words, and end with a thank-you.

Time to set up: About an hour
Who should own it: Whoever does the bookkeeping
You’ll know it’s working when: You stop making evening phone calls about unpaid invoices and fewer invoices are over 30 days late.

What could go wrong: A reminder could go to a customer who already paid by cheque if records are behind.

In your own report, an “I’m trying this” button here starts a simple tracker.

#2 Quick win Customer service

Draft replies to common customer questions

An AI assistant drafts answers to the questions you get every day — prices for a service call, what a furnace tune-up includes, how soon you can come. Your office staff check each draft and send it.

In your business: A homeowner emails asking whether you service tankless water heaters and how soon you could come. The office gets a ready-to-check reply in seconds instead of writing it from scratch.

Could free up: about 1.5–5 hours a week — roughly $2,900–$14,500 CAD a year. Estimate from your answer about “Answering the same customer questions”, assuming it takes a lot (30–50%) of that job off your plate.

How to try it

List the 10 questions your office answers most often.
Write your best short answer to each one in a document.
Open an AI assistant and paste in your answers as background.
When a customer email arrives, paste it in and ask for a draft reply that uses your answers.
Read and fix the draft, then send it from your normal email.
Add any new question and answer to your document each week.

What you could type into an AI assistant

You help the office of a plumbing and heating company reply to customers. Use only the answers below. If the question is about a gas smell or flooding, reply that they should call us right away at [phone]. Here are our answers: [paste your answers]. Draft a friendly reply to this email: [paste customer email].

Time to set up: An afternoon
Who should own it: Office lead
You’ll know it’s working when: Most customer emails get a reply the same morning and the office spends less time typing.

A person checks every result: Replies go to customers and may mention availability or service details that must be right.

What could go wrong: It could promise a time slot you do not have if nobody checks it.

In your own report, an “I’m trying this” button here starts a simple tracker.

#3 Quick win Sales

Turn job notes into a polished quote

You paste in your rough notes from a site visit, and an AI assistant turns them into a clear, professional quote description with the scope of work spelled out.

In your business: After looking at a basement that needs a new water heater and repiping, you jot ten words on your phone. The assistant turns them into a tidy scope of work you can paste into your quoting software.

Could free up: about 1.5–5 hours a week — roughly $2,900–$14,500 CAD a year. Estimate from your answer about “Writing quotes or estimates”, assuming it takes a lot (30–50%) of that job off your plate.

How to try it

Pick two of your best past quotes as examples.
After a site visit, write or say your rough notes into your phone.
Paste your notes and the examples into an AI assistant.
Ask it for a clear scope of work in the same style.
Check every line and add your prices in your quoting software yourself.
Send the quote as you normally would.

What you could type into an AI assistant

Here are two example quote descriptions from our plumbing company: [paste examples]. Using the same style, turn these rough site notes into a clear scope of work a homeowner can understand. Do not add any work that is not in my notes, and do not include prices. Notes: [paste notes].

Time to set up: About an hour
Who should own it: The owner or estimator
You’ll know it’s working when: Quotes go out within a day of the visit and you stop writing them late at night.

A person checks every result: A quote is a promise to the customer; the scope and prices must be checked by you.

What could go wrong: It might describe work you did not plan to include, so read every line.

In your own report, an “I’m trying this” button here starts a simple tracker.

#4 Quick win Marketing

Write seasonal posts and reminders

An AI assistant drafts short posts and email reminders for the busy seasons, like booking a furnace tune-up before winter.

In your business: In September you ask for three short posts reminding homeowners to book a tune-up before the first cold snap.

How to try it

Choose your next busy season, like fall tune-ups.
Ask an AI assistant for five short post ideas for local homeowners.
Pick two and add a local detail, like your town or a recent job.
Schedule them on the social media accounts you already use.
Send the best one as an email to past customers.
Note how many bookings mention the post or email.

What you could type into an AI assistant

Write three short social media posts for a plumbing and heating company in [town] reminding homeowners to book a furnace tune-up before winter. Friendly and practical, under 50 words each, no exaggerated claims.

Time to set up: About 30 minutes
Who should own it: Office lead
You’ll know it’s working when: You post regularly in busy seasons without it eating your weekend.

What could go wrong: Posts can sound generic unless you add your own local details.

In your own report, an “I’m trying this” button here starts a simple tracker.

#5 Quick win Operations and field work

Job summaries from technician voice notes

Technicians record a quick voice note after each job, and an AI tool turns it into a short written summary for the job file.

In your business: After replacing a pressure valve, a technician talks for thirty seconds in the van; the office gets a clean summary for the customer's file.

How to try it

Pick one technician who is happy to try it.
Have them record a 30-second voice note after each job: what they found, what they did, anything to follow up.
Use an AI tool to turn the recording into a short written summary.
Have the office paste the summary into the job file.
After two weeks, ask the technician and office what to change.
Roll it out to the rest of the team if it helped.

What you could type into an AI assistant

Turn this technician's voice note into a short job summary with three headings: What we found, What we did, Follow-up needed. Keep part numbers exactly as said. Note: [paste transcript].

Time to set up: About an hour
Who should own it: Lead technician
You’ll know it’s working when: Job files are complete without the office chasing technicians for details.

What could go wrong: Noisy recordings can produce wrong details, like a part number.

In your own report, an “I’m trying this” button here starts a simple tracker.

Everything AI could do for you

The full list of realistic ways AI could help, across every part of your business, each rated and ranked.

Area of the business

Kind of idea

Sort by

Showing all 15 ideas.

Quick win Bookkeeping and finance Top 5 Score 100

Friendly payment reminders

Your accounting software can send polite, automatic reminders when an invoice is overdue, using wording an AI assistant helps you write.

Example, ratings and risks

In your business: Three days after an invoice is due, the customer gets a friendly nudge with a payment link — no more evening phone calls.

How much it would help: A lot
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A feature in software you already use

Could free up about 0.5–2.5 hours a week — roughly $1,150–$7,200 CAD a year.

Information it needs: Your invoices, already in your accounting software.

What could go wrong: A reminder could go to a customer who already paid by cheque if records are behind.

Quick win Customer service Top 5 Score 95

Draft replies to common customer questions

Example, ratings and risks

How much it would help: A lot
How much work it takes to set up: An afternoon or less
What could go wrong: Needs checking
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Could free up about 1.5–5 hours a week — roughly $2,900–$14,500 CAD a year.

Information it needs: Your usual answers to common questions, written down once.

A person checks every result: Replies go to customers and may mention availability or service details that must be right.

What could go wrong: It could promise a time slot you do not have if nobody checks it.

Quick win Sales Top 5 Score 90

Turn job notes into a polished quote

You paste in your rough notes from a site visit, and an AI assistant turns them into a clear, professional quote description with the scope of work spelled out.

Example, ratings and risks

How much it would help: A lot
How much work it takes to set up: An afternoon or less
What could go wrong: Needs checking
Is the information it needs ready?: Some tidying first
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Could free up about 1.5–5 hours a week — roughly $2,900–$14,500 CAD a year.

Information it needs: Your site-visit notes and one or two of your best past quotes as examples.

A person checks every result: A quote is a promise to the customer; the scope and prices must be checked by you.

What could go wrong: It might describe work you did not plan to include, so read every line.

Quick win Marketing Top 5 Score 83

Write seasonal posts and reminders

An AI assistant drafts short posts and email reminders for the busy seasons, like booking a furnace tune-up before winter.

Example, ratings and risks

In your business: In September you ask for three short posts reminding homeowners to book a tune-up before the first cold snap.

How much it would help: Noticeably
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Information it needs: A sentence about your offer and the season.

What could go wrong: Posts can sound generic unless you add your own local details.

Quick win Operations and field work Top 5 Score 83

Job summaries from technician voice notes

Technicians record a quick voice note after each job, and an AI tool turns it into a short written summary for the job file.

Example, ratings and risks

In your business: After replacing a pressure valve, a technician talks for thirty seconds in the van; the office gets a clean summary for the customer's file.

How much it would help: Noticeably
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Information it needs: A phone and a few seconds after each job.

What could go wrong: Noisy recordings can produce wrong details, like a part number.

Quick win Hiring and HR Score 83

Write job ads and interview questions

An AI assistant helps you write a clear job ad and a short list of interview questions for a new technician or office role.

Example, ratings and risks

In your business: You need an apprentice before winter; in ten minutes you have a job ad and eight interview questions to adjust.

How much it would help: Noticeably
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Information it needs: A few lines about the role.

A person checks every result: Hiring decisions must stay with people, and ads must follow employment rules.

What could go wrong: Wording could accidentally put off good candidates; read it with fresh eyes.

Quick win Sales Score 75

Follow up on quotes that went quiet

Your email or booking software can send a friendly follow-up when a customer has not answered a quote after a week, using a message AI helps you write once.

Example, ratings and risks

In your business: A customer got a quote for a new furnace and went silent. Seven days later they get a short, friendly check-in written in your voice.

How much it would help: Noticeably
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within weeks
Kind of tool: A feature in software you already use

Could free up about 1–3 hours a week — roughly $1,450–$8,650 CAD a year.

Information it needs: A list of open quotes and the customer's email address.

What could go wrong: Follow-ups could annoy customers if they go out too often.

Quick win Admin and paperwork Score 75

Summarize long emails and contracts

Paste in a long email thread or a supplier agreement and get a short summary of what it says and what you need to do.

Example, ratings and risks

In your business: A builder sends a 12-page subcontract; you get a one-page summary of the payment terms and deadlines to look at closely.

How much it would help: Noticeably
How much work it takes to set up: An afternoon or less
What could go wrong: Needs checking
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Could free up about 1–3 hours a week — roughly $1,450–$8,650 CAD a year.

Information it needs: The document or email.

A person checks every result: A summary can miss a key clause; read the original before you sign anything.

What could go wrong: It may leave out an important detail, so it never replaces reading the contract.

Project Scheduling Score 68

Smarter scheduling suggestions

Some booking tools now suggest the best technician and time slot for a job based on location and skills. Your office still makes the final call.

Example, ratings and risks

In your business: A no-heat call comes in from the north end; the tool suggests the technician already working two streets away.

How much it would help: A lot
How much work it takes to set up: A few days
What could go wrong: Needs checking
Is the information it needs ready?: Some tidying first
How soon you would see a difference: Within weeks
Kind of tool: A feature in software you already use

Could free up about 1.5–6 hours a week — roughly $2,900–$17,500 CAD a year.

Information it needs: Technician skills, working hours and job addresses in your booking tool.

A person checks every result: Scheduling mistakes mean a customer waiting at home for nobody.

What could go wrong: Suggestions are only as good as the details in the booking tool.

Quick win Operations and field work Score 68

Plain-language maintenance guides for customers

AI drafts short how-to guides customers can read before calling, like how to relight a pilot light safely or reset a breaker.

Example, ratings and risks

In your business: Customers who call about a cold house get a link to a simple checklist first, with a clear warning to call if they smell gas.

How much it would help: A little
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within days
Kind of tool: A general AI assistant (a chat tool)

Could free up about 0.5–1.5 hours a week — roughly $500–$4,300 CAD a year.

Information it needs: The questions customers ask most.

What could go wrong: A guide that is not checked by a licensed technician could give unsafe advice.

Quick win Hiring and HR Score 58

Safety checklists and onboarding guide

AI helps you turn what you already know into a simple onboarding guide and safety checklists for new staff.

Example, ratings and risks

In your business: A new apprentice gets a two-page first-week guide and a van safety checklist you wrote in an hour.

How much it would help: A little
How much work it takes to set up: An afternoon or less
What could go wrong: Little harm if it slips
Is the information it needs ready?: Yes, already in one place
How soon you would see a difference: Within weeks
Kind of tool: A general AI assistant (a chat tool)

Information it needs: Your existing rules and tips.

What could go wrong: Safety content must be checked against the regulations that apply to you.

Project Inventory and purchasing Score 55

Parts list from a job description

Describe the job and an AI assistant suggests a starting parts list to check against your stock.

Example, ratings and risks

In your business: For a water heater swap, it lists the fittings, venting and valves you usually need, so nothing gets forgotten.

How much it would help: Noticeably
How much work it takes to set up: A few days
What could go wrong: Needs checking
Is the information it needs ready?: Some tidying first
How soon you would see a difference: Within weeks
Kind of tool: A general AI assistant (a chat tool)

Could free up about 0.5–1.5 hours a week — roughly $500–$4,300 CAD a year.

Information it needs: Your common parts and a few past job lists.

A person checks every result: A wrong part means a second trip, so a technician must check the list.

What could go wrong: It may suggest parts you do not stock or that do not fit local code.

Project Bookkeeping and finance Score 50

Read and file supplier invoices

Some accounting tools can read a supplier invoice from a photo or email and fill in the details for you to approve.

Example, ratings and risks

In your business: You snap the supply house invoice on the counter; it shows up in your books, ready to approve.

How much it would help: Noticeably
How much work it takes to set up: A few days
What could go wrong: Needs checking
Is the information it needs ready?: Some tidying first
How soon you would see a difference: Within weeks
Kind of tool: A feature in software you already use

Information it needs: Supplier invoices by email or photo.

A person checks every result: Amounts and tax must be checked before they hit your books.

What could go wrong: A misread number could throw off your books if nobody checks it.

Bigger bet Customer service Score 45

A website chat helper that books service calls

A chat window on your website answers simple questions and collects details for a service call, then hands it to your office to confirm.

Example, ratings and risks

In your business: At 10 pm, a homeowner with no hot water describes the problem in the chat. In the morning, your office finds the details ready to call back.

How much it would help: A lot
How much work it takes to set up: Weeks, maybe outside help
What could go wrong: Mistakes could really hurt
Is the information it needs ready?: Some tidying first
How soon you would see a difference: After months
Kind of tool: An app made for this one job

Could free up about 1.5–5 hours a week — roughly $2,900–$14,500 CAD a year.

Information it needs: Your services, service area and booking rules.

A person checks every result: A person must confirm every booking and anything the chat told the customer.

What could go wrong: It could give wrong advice on an emergency, like a gas smell, unless it is set up to send people straight to a phone call.

Bigger bet Operations and field work Score 25

Predict which furnaces need a fall tune-up

A custom tool could look at your past jobs to guess which customers' furnaces are due for service, so you can contact them first.

Example, ratings and risks

In your business: In August, a list shows the 60 customers whose furnaces are over ten years old and have not been serviced in two years.

How much it would help: Noticeably
How much work it takes to set up: Weeks, maybe outside help
What could go wrong: Needs checking
Is the information it needs ready?: Not yet
How soon you would see a difference: After months
Kind of tool: Something built for you

Information it needs: Clean records of past jobs with equipment ages.

What could go wrong: With messy records, the list will be wrong and waste calls.

How to tell if it’s working

One to three simple things to count for each top-5 idea, and how to take a “before” number this week.

1. Friendly payment reminders

Invoices more than 30 days late — Count of unpaid invoices that are more than 30 days past their due date, checked on Friday. Your “before” number: Run the aged receivables report in your accounting software this Friday and count the invoices in the over-30-days column. · Check in: every week · Aim for about 25-50% less after 90 days
Time spent chasing payments — Minutes a week anyone spends calling or emailing customers about payment. Your “before” number: Keep a tally on a sticky note by the phone for one week. · Check in: every week · Aim for about 50%+ less after 90 days

2. Draft replies to common customer questions

Hours spent replying to customer emails — Total hours the office spends writing replies to customer emails in a week. Your “before” number: For one week, each person notes roughly how long they spent on email replies each day. · Check in: every week · Aim for about 25-50% less after 90 days
Emails answered the same day — Out of customer emails received before noon, how many got a reply the same day, as a percentage. Your “before” number: Look at last week's inbox and count both numbers. · Check in: every week · Aim for about 10-25% more after 90 days

3. Turn job notes into a polished quote

Days from visit to quote sent — Average number of days between a site visit and the quote going out. Your “before” number: Look at the last 10 quotes and work out the gap for each. · Check in: every two weeks · Aim for about 25-50% less after 90 days
Evenings spent on quotes — Number of evenings a week you work on quotes after hours. Your “before” number: Mark it on a calendar this week. · Check in: every week · Aim for about 50%+ less after 90 days

4. Write seasonal posts and reminders

Tune-up bookings in the season — Number of tune-up bookings in the four weeks after the posts go out. Your “before” number: Count tune-up bookings from the same four weeks last year in your booking tool. · Check in: every month · Aim for about 10-25% more after 90 days

5. Job summaries from technician voice notes

Job files missing details — Number of completed jobs in a week where the office had to chase a technician for details. Your “before” number: The office keeps a tally for one week. · Check in: every week · Aim for about 25-50% less after 90 days

Your first 90 days

A 30/60/90-day plan: what to do each month, when to check in, and what to decide at day 90.

First 30 days

Two quick wins that free up your evenings.

Turn on payment reminders in your accounting software.
Write down your 10 most common customer questions and answers.
Start drafting email replies with an AI assistant and check every one.
Record your before numbers for both.

Check-point: Compare late invoices and email time with your before numbers.

Days 31 to 60

Quotes and your busy-season marketing.

Use an AI assistant to turn site notes into quote descriptions.
Draft and schedule fall tune-up posts and an email to past customers.
Hand the customer-reply habit to the office lead.

Check-point: Check days from visit to quote and how many evenings you worked on quotes.

Days 61 to 90

Field notes and deciding what to keep.

Try voice-note job summaries with one technician.
Review every tracked item with your numbers.
Look into smarter scheduling in your booking tool for next winter.

Check-point: Decide what stays, what grows and what stops.

At day 90, decide

Keep: Keep anything that saves time every week and that the team now uses without reminders.
Expand: Expand what clearly helped — for example, using reply drafts for text messages too.
Drop: Drop anything that needed more checking than it saved, or that the team avoided.

Staying safe: guardrails and an AI use policy

The risks that matter for the information you handle, simple safe-use rules, and a one-page AI policy you can give your staff.

The risks that matter for you

AI tools can state wrong things confidently, like a wrong service time or price.Rule: A person reads every message before it goes to a customer.
Customer names, addresses and payment details could end up stored in an AI tool.Rule: Never paste payment details into an AI assistant, and remove customer names and addresses where you can.
Safety advice from AI could be wrong in an emergency like a gas smell.Rule: Anything about gas, carbon monoxide or flooding always says to call you or emergency services right away.
Staff may rely on AI for things it should not decide, like hiring.Rule: AI drafts; people decide.

Northfield's AI use rules

We use AI tools to save time on writing and paperwork. These rules keep our customers and our company safe.

Do use AI to draft emails, posts and summaries.
Do read and fix everything before it leaves the company.
Don't paste payment card or bank details into any AI tool.
Don't paste customer health information or anything about children.
Don't use AI to decide who to hire or how much to charge.
Do use only the AI tools the owner has approved.
Do tell the office if an AI tool gives a strange or wrong answer.
Don't use AI for gas, carbon monoxide or flooding advice — call the customer.

Questions? Ask the owner before you try something new.

Privacy

Canada's privacy law for businesses (PIPEDA) may apply to the customer information you handle. In general it expects you to tell customers how their information is used and to keep it safe.
If you work with customers in Quebec, Quebec's Law 25 may apply too, with extra rules about sending personal information outside the province.

Not legal advice. This is general guidance, not legal advice. Check with a lawyer or privacy professional about your situation.

Bringing your team along

How to introduce AI to your team, who to involve, a simple two-week learning plan, and honest answers to common worries.

Introducing AI to your team

Start with one small, boring job everyone dislikes, like payment reminders.
Say clearly that AI is there to take away paperwork, not jobs.
Show a real example from your own business in a team meeting.
Let one keen person try it first and share what they learned.

Who to involve

Office lead — They handle the most writing and will notice the time saved first.
A senior technician — Field staff trust a peer more than a new tool.

A simple two-week learning plan

Days 1–2: Owner and office lead try an AI assistant on five real customer emails.
Days 3–5: Write down your common questions and answers together.
Week 2, day 1: Short team meeting: show what worked and the safety rules.
Week 2, days 2–5: Office drafts all replies with AI and checks each one; note any problems.
End of week 2: Fifteen-minute review: keep, change or stop.

Common worries, honest answers

Will AI replace my job?: No. It drafts the routine writing so people can spend more time with customers and on the tools.
What if it gets something wrong?: It will, sometimes. That's why a person checks everything before it goes out.
Is our customer information safe?: Only if we follow our rules: no payment details, and only approved tools.

AI in plain English

What AI is good and bad at for a business like yours, why it sometimes makes things up, how to check its work, and a short glossary.

Today's AI tools are programs you talk to in plain language. They are very good at writing and summarizing, and they learn nothing about your business unless you tell them.

What it’s good at

Drafting customer emails from your notes.
Turning messy site notes into clear descriptions.
Summarizing long documents.
Writing first drafts of posts, ads and guides.

What it’s bad at

Knowing your real schedule, prices or stock unless you give them.
Safety-critical advice, like what to do about a gas smell.
Math you have not checked.
Knowing local codes and rules for certain.

Why it sometimes makes things up

These tools predict likely words. When they do not know something, they can still produce a confident answer that sounds right but is wrong. That is why every customer-facing message needs a human check.

How to check its work

Read every draft before it is sent.
Check names, dates, times and part numbers against your records.
Ask it where an answer came from, and verify anything important.
When in doubt, leave it out.

Words in this report

AI assistant: A chat tool, like ChatGPT, Claude, Gemini or Copilot, that you type questions or instructions into.
Prompt: What you type into an AI assistant: your question or instructions.
Hallucination: When an AI tool states something false as if it were true.
Integration: A connection that lets two pieces of software share information automatically.

Next steps in Costiva

The Costiva tools that fit your plan best, and why.

Process Cost Baseline Calculator

Measure what quoting and invoice chasing cost you today, so you can prove the change later.

Open Process Cost Baseline Calculator

AI Usage Policy Generator

Turn the AI use rules in this report into a full staff policy.

Open AI Usage Policy Generator

ROI & Payback Simulator

See when your top ideas would pay for themselves.

Open ROI & Payback Simulator

How we worked this out

The AI wrote the descriptions and rated each idea on fixed scales. Costiva’s own code — not the AI — turned those ratings into scores, picked your top 5 and worked out every hours and money range from your answers. The AI never saw a money figure.

Savings are estimates from your own answers, shown as ranges. They are not a promise.
This report is guidance, not professional, legal or financial advice.
Product names are examples, not endorsements. Costiva is not paid to mention any of them.

Read the full method and scoring weights

Method version 2026.10-1.

Why create an account?

Your answers are saved, so you can stop and come back any time.

Your report stays in your account.

Track your results over time.

Use your business profile to get better results from the other Costiva tools.

Questions, answered plainly

Today’s AI tools are computer programs you can talk to in plain language. They are good at writing, summarizing and sorting information. They also make mistakes and can state wrong things confidently, so a person should always check their work. Your report explains this in more detail, with examples from your industry.

No. You answer questions about your business in your own words, and the report is written for people who have never used AI. Every step it suggests can be done without a technical person.

About 15 minutes to answer the questions, then a few minutes while your report is written. Your answers save as you go, so you can stop and come back any time.

Your answers are saved in your Costiva account. To write your report, we send your answers to Anthropic, the company that makes the Claude AI, through its service for businesses. Anthropic’s terms for business customers say it does not use that information to train its AI. We never send it the hourly cost you pick — Costiva works out every money figure itself. Costiva does not sell your answers or use them for advertising, and you can delete your AI plan at any time. Please don’t include payment card numbers or health details in your answers.

Yes. A free account includes 1 full AI plan every 30 days, up to 3 section rewrites per plan, and unlimited use of the tracker. Paid plans include more plans and rewrites.

No one pays Costiva to be mentioned. The report describes the kind of tool each idea needs first — often a feature in software you already have or a free assistant. When it names products, they are examples from more than one company, not endorsements, and it never quotes prices.

Ready to find out what AI can do for you?

Create a free account to start

Discover

Plan

Govern

Plan

Build

Test

Govern

For Leaders

Discover

Plan

Govern

For Builders

Plan

Build

Test

Govern

Is this for me?

How it works

Describe your business

Answer a few follow-up questions

Get your report

Track what you try

What’s in your report

Your plan at a glance

Your top 5: what to try first

Everything AI could do for you

How to tell if it’s working

Your first 90 days

Staying safe: guardrails and an AI use policy

Bringing your team along

AI in plain English

Next steps in Costiva

How we worked it out

See a sample report

Your plan at a glance

Do these three first

Your top 5: what to try first

Friendly payment reminders

How to try it

What you could type into an AI assistant

Draft replies to common customer questions

How to try it

What you could type into an AI assistant

Turn job notes into a polished quote

How to try it

What you could type into an AI assistant

Write seasonal posts and reminders

How to try it

What you could type into an AI assistant

Job summaries from technician voice notes

How to try it

What you could type into an AI assistant

Everything AI could do for you

Friendly payment reminders

Draft replies to common customer questions

Turn job notes into a polished quote

Write seasonal posts and reminders

Job summaries from technician voice notes

Write job ads and interview questions

Follow up on quotes that went quiet

Summarize long emails and contracts

Smarter scheduling suggestions

Plain-language maintenance guides for customers

Safety checklists and onboarding guide

Parts list from a job description

Read and file supplier invoices

A website chat helper that books service calls

Predict which furnaces need a fall tune-up

How to tell if it’s working

1. Friendly payment reminders

2. Draft replies to common customer questions

3. Turn job notes into a polished quote

4. Write seasonal posts and reminders

5. Job summaries from technician voice notes

Your first 90 days

Two quick wins that free up your evenings.

Quotes and your busy-season marketing.

Field notes and deciding what to keep.

At day 90, decide

Staying safe: guardrails and an AI use policy