Before an AI front-office audit, bring a list of the tools you use, a few examples of recent customer inquiries, and the rules for booking, payment and follow-up. Include the person who handles that work each day. Leave passwords out of the conversation; agree on the access needed before connecting any accounts.
The useful question is where work gets stuck. A missed inquiry might be a message nobody saw, a quote that never got a reply, or a payment that arrived without the booking being updated. Those are different problems. The audit should show which one you have before recommending a system.
Which records should you bring?
For an initial conversation, choose a few customer journeys you can explain from beginning to end. Include one that went smoothly and one that stalled. Follow the original inquiry through the reply, quote or booking, payment and next contact. If a step happened in a text message or on paper, include that detail.
For the Lights On AI Front Office Audit, the published scope is 90 days of inquiries, bookings or orders, and payments. The first examples help us agree on how to review that period. They are a starting point, not enough evidence to estimate the whole business.
You do not need to clean up every record first. A duplicate customer or an unclear note may be part of the problem we need to see. We agree on which records are needed and how they will be reviewed before sharing them.
What should you write down about your current tools?
List where inquiries arrive and where staff record the next step. That might be email, web forms, a shared calendar, a spreadsheet, an order system or a CRM. Note who owns each account and who uses it. Include the places where someone copies information from one tool to another.
A simple example is enough: “The form sends an email, someone adds the customer to the spreadsheet, and the booking goes into the calendar.” That gives us something concrete to check. A list of software names by itself does not show how work gets done.
Do you need to replace your CRM first?
No. Review the current workflow before choosing replacement software. A new CRM may help, but it will not decide who answers an inquiry, when an unpaid hold expires, or what happens after a customer stops replying. Those rules need to be clear wherever the records live.
If a change is recommended, the plan should explain what the current system cannot do, what would move, and what staff would need to learn. Keeping a tool that works can be the sensible choice.
Who should take part?
Bring the owner or decision-maker and whoever handles inquiries, bookings or orders. They may be the same person. Someone who does the daily work can explain the exceptions that a process diagram misses, such as the customer who replies to an old email or pays under a different name.
Our personal, hands-on approach starts with that conversation. The aim is to understand your business well enough to make a useful recommendation, including what should stay with a person.
What access should you give an AI provider?
Agree on the purpose and permissions before connecting an account. Ask which records the provider needs, whether it can read or change them, who can use the connection, and how you can remove it. Do not send passwords in an inquiry form or ordinary email.
For Google accounts, Google explains how to review and remove third-party connections. Removing access does not necessarily delete copies already held by another service, so ask separately about storage and deletion. The Australian Cyber Security Centre also advises small businesses to check AI outputs and keep human oversight of important business decisions.
How will you know whether the change helped?
Agree on a measure before the build. If the problem is unanswered inquiries, check how many received a reply and how long it took. If the problem is unpaid holds, check which holds were released and whether paid bookings stayed intact. If follow-up is the problem, check the record of what was due and what happened.
A polished dashboard is useful only if its numbers can be traced to the underlying records. A time-saving or revenue claim needs evidence from your business. A few selected examples can explain a problem; they cannot establish the result for every customer.
What should you receive after the audit?
Expect a written account of what was reviewed, where work stalled, and what should be fixed first. Each recommendation should say what it changes, what access it needs, who remains responsible, and how the result will be checked. Unresolved questions should be named.
The Lights On AI audit is listed as ten business days with a written plan and a 45-minute walkthrough. The packages page holds the current scope and price. If you are still deciding where to begin, use the free Front Office Score or schedule a conversation and tell us which part of the process is causing trouble.
Straight answers
What should I bring to an AI front-office audit?
Bring a list of your tools, examples of recent customer journeys, and your rules for booking, payment and follow-up. For the Lights On AI audit, the agreed review covers 90 days of records.
Does my inbox need to be organized before the audit?
No. Duplicate records and unclear next steps can be part of the problem being reviewed. Agree on the records needed and how they will be handled before sharing them.
Do I need to buy a new CRM before using AI?
No. Review the workflow first. A replacement should be recommended only with a clear explanation of what the current tools cannot do and what the change would involve.
Should I email my account passwords?
No. Agree on the access needed and use the account connection or permission process appropriate to the service. Keep passwords out of inquiry forms and ordinary email.
Who should attend the first conversation?
Include the owner or decision-maker and the person who handles inquiries, bookings or orders each day. Their practical knowledge helps explain where work stalls.
What should the written audit plan include?
It should explain what was reviewed, the problems found, the recommended order of fixes, required access, who remains responsible, and how results will be checked.