Free tool

The AI integration checklist

Thirty things to get right before you connect your business systems to AI. Check them off, see your readiness grade, and find out what to fix first.

Part 1 of 6

Data access and connections

AI can only work with data it can reach. This is the foundation everything else stands on.

You can get to your customer listIf your customer data lives in a system nobody can export from, nothing can be automated on top of it.How we connect your systems →Your systems can talk to each otherMost small businesses run five to ten tools that never share data. Confirm which ones have a way to connect.How we connect your systems →You know which records are the real onesWhen the CRM says one thing and the spreadsheet says another, automation just spreads the confusion faster. Pick one source of truth first.Your data is backed up somewhere you controlBefore you connect anything new, make sure one bad day cannot wipe out the records your business runs on.Keeping data in your building →Old, duplicate, and dead records are cleaned upAI trained on messy data gives messy answers. A cleanup pass before you build saves months of strange results later.

Part 2 of 6

Logins and permissions

Most small business breaches come through old accounts and shared passwords, not hackers in hoodies.

Every system has a named ownerWhen something breaks at 9 PM, 'I think Sarah set that up' is not a plan. Every login needs a human responsible for it.How we plan the rollout →Former employees can no longer get inOld accounts are the most common way small businesses get breached. Remove access the day someone leaves.Admin logins are separate from daily loginsThe account that can delete everything should not be the one used to check email. Separation keeps one bad click from becoming a disaster.Two-factor is on for the important accountsPasswords get stolen. A code on your phone stops most break-ins cold, and it takes five minutes to set up.You know who can see customer dataWrite down who has access to what. If you cannot list it, you cannot protect it.Keeping data in your building →

Part 3 of 6

Human approval gates

The difference between AI that helps and AI that embarrasses you is deciding up front where a person steps in.

A person approves anything involving moneyQuotes, refunds, and payments should never go out on AI autopilot alone. A human glance takes seconds and prevents the expensive mistakes.How we automate carefully →A person approves anything a customer seesThe first AI-written message that goes to the wrong customer undoes a year of trust. Keep a human in the loop on outbound communication.How we automate carefully →You wrote down what the AI is never allowed to doMake the no-go list before you build: no deleting records, no changing prices, no promising timelines. Boundaries first, automation second.Someone reviews the AI's work weekly at firstNew systems need supervision like new employees do. A 15-minute weekly review catches bad habits before they become expensive ones.Your team knows when it is the AI versus a personCustomers deserve to know. Staff need to know when to step in. Nobody should have to guess.

Part 4 of 6

Cost controls

AI bills grow quietly. The businesses that get burned are the ones with no limits and no one watching.

You know what each AI tool costs per monthSubscriptions multiply quietly. List every AI tool, what it costs, and who uses it, or the bill will surprise you.How we control AI spend →There is a monthly spending limitAI usage bills can spike tenfold in a bad month with no warning. A hard cap means the worst case is a number you chose.How we control AI spend →Cheap tasks do not run on the expensive AISorting emails does not need the top-tier model. Routing simple work to cheap options cuts most AI bills dramatically.How we control AI spend →One person owns the AI budgetIf everyone can sign up for tools and nobody watches the total, costs grow every month. One person, one number, reviewed monthly.You can see what each dollar boughtA bill with no detail is a bill you cannot question. If you cannot trace spend to results, you cannot manage it.How we control AI spend →

Part 5 of 6

Monitoring and alerts

Automations fail. The question is whether you find out in minutes or in weeks.

You get told when something breaksSilent failures are the expensive kind. If an automation stops Tuesday and nobody notices until Friday, that is three days of lost work.How we build for reliability →You can see what the AI did todayA simple daily log, who it contacted and what it changed, keeps everyone honest and makes problems easy to trace.Customer complaints reach a human fastWhen the AI gets something wrong with a customer, the fix needs a person immediately, not a ticket queue.You test the important workflows monthlySystems drift. A monthly ten-minute test of your key automations catches rot before your customers do.You know exactly who to call when it breaksWhether it is your vendor or your IT person, know the name and number now, and make sure they pick up.

Part 6 of 6

Rollback and ownership

The endgame question: if everything went wrong tomorrow, would you still own your business?

You can turn the AI off without breaking the businessIf flipping the switch off means chaos, you do not have a tool, you have a dependency. Always keep a manual fallback.How we plan the rollout →You own your data and your workflowsIf your vendor disappeared tomorrow, could you keep running? Your logic should live where you control it.How we plan the rollout →You can switch AI providers without rebuildingProviders raise prices and change terms. If switching means starting over, they own you. Build so swapping is a settings change.How we avoid lock-in →How it all works is written downHow it is built, how it is connected, who to call. If the knowledge lives in one person's head, you do not own the system.A 90-day review is on the calendarWhat works today needs tuning in three months. Put the review on the calendar now or it will not happen.Start with a diagnostic →

Your readiness grade

Not yet, and that is fine

You have 30 items left. Start at the top with data access and connections. Most owners finish this list in a week of spare moments.

Show my 30 remaining items

○ You can get to your customer list (Data access and connections)
○ Your systems can talk to each other (Data access and connections)
○ You know which records are the real ones (Data access and connections)
○ Your data is backed up somewhere you control (Data access and connections)
○ Old, duplicate, and dead records are cleaned up (Data access and connections)
○ Every system has a named owner (Logins and permissions)
○ Former employees can no longer get in (Logins and permissions)
○ Admin logins are separate from daily logins (Logins and permissions)
○ Two-factor is on for the important accounts (Logins and permissions)
○ You know who can see customer data (Logins and permissions)
○ A person approves anything involving money (Human approval gates)
○ A person approves anything a customer sees (Human approval gates)
○ You wrote down what the AI is never allowed to do (Human approval gates)
○ Someone reviews the AI's work weekly at first (Human approval gates)
○ Your team knows when it is the AI versus a person (Human approval gates)
○ You know what each AI tool costs per month (Cost controls)
○ There is a monthly spending limit (Cost controls)
○ Cheap tasks do not run on the expensive AI (Cost controls)
○ One person owns the AI budget (Cost controls)
○ You can see what each dollar bought (Cost controls)
○ You get told when something breaks (Monitoring and alerts)
○ You can see what the AI did today (Monitoring and alerts)
○ Customer complaints reach a human fast (Monitoring and alerts)
○ You test the important workflows monthly (Monitoring and alerts)
○ You know exactly who to call when it breaks (Monitoring and alerts)
○ You can turn the AI off without breaking the business (Rollback and ownership)
○ You own your data and your workflows (Rollback and ownership)
○ You can switch AI providers without rebuilding (Rollback and ownership)
○ How it all works is written down (Rollback and ownership)
○ A 90-day review is on the calendar (Rollback and ownership)

Book a Free Diagnostic Book Directly

Keep going

Two tools, one starting point

The checklist tells you whether you are ready. The calculator tells you what the busywork is costing you and what to fix first.

AI Opportunity Calculator

Put in your hours and your hourly cost. See what busywork costs you per year, ranked by what to automate first.

Try the calculator

Free AI Diagnostic

Thirty minutes with us. We map your operation, name what is costing you the most, and give you a written assessment either way.

Book the diagnostic

Questions we hear

Checklist questions, answered

How long does the checklist take?

About 15 minutes if you know your systems, longer if you have to hunt down logins. That is useful information too: if you cannot answer a question, that box is doing its job.

Do I need every box checked before using AI?

No. But the unchecked boxes are your risk list. Most businesses start building with around 20 checked and a plan for the rest. The grade at the bottom tells you where you stand.

Who should fill this out?

The person who would get the call if something broke. Usually the owner or the office manager, not the vendor selling you the system.

Is my progress saved anywhere?

Only in your own browser. Your answers never leave your computer and we never see them.

Engagement starts here

Know your score. Now map the work.

Bring your checklist results to a free 30-minute diagnostic. We will look at what is unchecked, tell you what actually matters for your business, and give you a written plan either way.

Book a Free Diagnostic Book Directly

Not limited to what's listed. Every engagement starts by assessing what your business actually needs, and we build whatever it requires.