One of your employees shows you something they made with AI. It took five minutes. You can see why they’re excited.
You’re also wondering whether it’s correct, what information they put into it and whether anyone has already sent it to a customer.
You have a business to run. There may be quieter periods when your staff have time to try things, but you haven’t had much opportunity to learn about AI yourself. Perhaps you’re interested. Perhaps you would happily leave the technology to someone else, provided you could trust them to do a good job.
When I speak with business owners, that question of trust comes up often. They recognise that AI could help. They’re less certain about who should take it on and how they would know whether the work was any good.
Some have enthusiastic employees they aren’t quite ready to trust with it. Others expect their staff to figure it out during downtime.
I think there is room to give people a chance. You need a way to assess what they can do, enough support for them to learn and some agreement about where their authority ends.
How much confidence do you have in your staff?
An owner may tell me their employees don’t have the skills to take on AI. Sometimes they mean technical skills. Sometimes they mean the judgement that comes from understanding the business.
Those are different concerns.
An employee might learn a tool quite quickly while still needing help understanding why a particular customer gets special treatment or why a seemingly small change to a quote affects the profit on the job.
Another employee might understand the operation thoroughly but have little confidence with technology.
I also hear concerns about expectations. The owner hired someone for a particular role, at a particular rate of pay. Asking that person to investigate AI and change how the company works may feel like a lot to expect.
That deserves a conversation. If the responsibility grows, discuss the time, support and recognition that go with it.
But I wouldn’t assume the employee is incapable before giving them a reasonable opportunity. Someone who deals with the same awkward process every day may have a very good idea of what needs fixing. They may never have had the tools or permission to try.
A little trust from the owner can give them that opening.
Give them a problem they recognise
“See what you can do with AI” leaves a lot for an employee to work out.
They might spend an afternoon creating social posts when the business is losing enquiries because nobody follows up. They might reorganise a spreadsheet when the underlying problem is that the information going into it is incomplete.
Ask your staff where they lose time. Compare their answers with the problems you see as the owner.
Suppose customers regularly request quotes without providing enough detail. Your team then spends time calling them back before anyone can prepare the estimate.
You could ask an employee to use AI to help draft a better enquiry checklist. Give them your requirements and a fictional customer example. Then sit down together and decide whether the proposed questions would help.
That is an assignment they can understand and you can assess.
Quiet periods can make room for this work, but spare time alone won’t teach someone how to do it. They may need training or a colleague to help. They also need to know that reporting an unsuccessful experiment is an acceptable result.
Otherwise, you risk getting demonstrations of what worked and hearing very little about what didn’t.
What if you would rather hand it over?
Trying AI yourself can help. You get to see how it responds to a vague explanation, what it misses and how much the answer changes when you provide more context.
I would encourage an owner to have that experience. I also know some owners don’t have the time or interest.
At some point, you have to rely on someone else’s expertise. You already do that in other parts of your business.
There is an approach to delegation that takes this quite far: admit you don’t know how to do something and give another person room to work out the method. I have seen it work. The owner doesn’t always need to understand how the work is done before someone else can produce a useful result.
How far you can go depends on the person and the assignment.
You still need to explain what you want to accomplish and make decisions about spending, access and responsibility. You also need someone who can assess the finished work. That could be you, an experienced employee or an outside practitioner.
If nobody involved can judge whether the result is dependable, you need help before putting it into use.
An established business has customer relationships and working processes that took years to build. Give people room to investigate, with clear limits on what they can change while they are learning.
Meet the person who will do the work
You may decide that your employees need outside support, or that you want someone else to lead the work.
Your choice might be an independent consultant or an agency. Either can be suitable. Much depends on the person assigned to you and whether they can give your business the time it needs.
If you speak with an agency, ask to meet the practitioner.
The salesperson may explain the offer well. You also want to speak with the person who will examine your processes, work with your staff and deal with problems after the initial demonstration.
Ask them to walk you through a relevant piece of work. What was the business trying to do? What did they change? What needed correcting?
Then talk about your business. Notice whether they ask questions before prescribing a solution. Can they explain what they would investigate first? Do they recognise where they need more information?
You may not be able to judge every technical detail, but you can assess whether someone listens and gives understandable answers.
Find out how much access you will have to that person. Who helps when your employee gets stuck? Who maintains the work? What happens if the practitioner is unavailable?
An individual may give you close attention but have limited capacity. An agency may have more people available, although the person doing your work may be carrying several other accounts. Establish who will be involved and what time they have committed.
Someone local may be a better fit

At this stage, there is a good reason to consider someone local who can come to your office.
Getting started may involve sitting at your computer, setting up the tools and showing you how to use them. They may need to help configure your phone so you can dictate instructions, review work or use AI between appointments.
You can do much of this remotely. You can also fly a consultant in. But if you expect to need hands-on help, having someone nearby makes it easier to arrange another visit when you get stuck.
I find that meeting face to face matters to some owners, particularly when they are already uncertain about trusting someone with their business systems. They want to sit with the person, ask questions and watch what they are doing.
Working together also gives the practitioner a chance to see where you struggle. Something that seems obvious to them may be the step that stops you using the tool once they leave. Sitting beside you while you try it can help them catch that.
Being local doesn’t establish someone’s competence. You still need to assess their work and agree on what they can access or change. But their willingness to come in, explain what they are doing and help you practise may be an important part of whether the arrangement works.
Ask about that before hiring them. Do they offer office visits? Will they help you and your staff with the actual setup? Are those visits included in the fee?
If you know you will need someone beside you at the computer, make that part of the assignment from the beginning. I describe what that help can involve in Have you found the power up button on your ChatGPT yet?, including setup, permissions and working through a task together.
Ask how they keep learning
Whoever advises you needs to keep up with AI as it changes.
Ask what they have been using recently. What have they tested that proved useful? Have they changed their mind about a tool or stopped recommending something?
Someone doing the work should be able to give you specific examples, including occasions when the technology disappointed them.
Reading announcements can tell them what has been released. Using the tools tells them whether a capability is dependable enough for the work they are proposing. In What I Actually Do With AI All Day, I describe the projects I work on and the time I spend testing and reconsidering the tools I use.
Nobody knows exactly where AI will settle. I would be interested in how a consultant makes decisions under those conditions and how they would adapt something they built for you.
There are enough questions here for a separate article about assessing an AI practitioner. For this first conversation, I would want to hear how they work, see a relevant example and understand what they propose to do for my business.
Agree on what people can do without asking
Whether an employee or consultant leads the work, be clear about what they may access, change and send.
An internal task can still have serious consequences. A mistake in a draft meeting agenda is quite different from a mistake in the spreadsheet used to calculate prices.
For an early experiment, choose work where errors are easy to identify and correct. Keep the original information intact. Agree on what can be entered into the AI tool before anyone uploads customer records, employee information or confidential supplier details. Canada’s privacy regulators provide guidance on using personal information with generative AI, including limiting what is collected and used.
Preparing a customer reply and sending it are separate permissions. So are suggesting a price and approving a quote.
Name the person who checks the work. Staff should know when they can proceed and when they need to ask.
As their ability develops, you can give them more responsibility. Keeping every small decision with the owner can eventually hold up the very work you wanted to improve.
Ask to see the corrections

When someone brings you AI-assisted work, look at it using the standards you already apply.
Would you send that email? Does the checklist include the information you need? Has the quote promised something you cannot deliver?
Then ask them to show you what they had to fix.
That conversation can tell you more than the finished document. You will see whether they noticed missing details, checked questionable claims and understood why a correction mattered.
Important facts should be checked against the original information, even when the answer sounds convincing. Prices need checking against the approved price list. A summary needs comparing with the notes it came from.
Ask about the time involved too, including your own review. A draft produced in five minutes may require forty minutes of repair.
The benefit may also extend beyond saving time. Perhaps your team can finally examine months of enquiry notes and identify questions customers keep asking. That information could help you improve the website or change how you prepare quotes.
Judge whether the completed work was useful and whether you could repeat the process with confidence.
You don’t have to wait until AI settles down
Waiting is a decision too.
While you wait, a competitor may be learning how to answer enquiries faster or reduce the time spent on routine administration. Your own business may have useful information sitting in files that nobody has time to examine.
Depending on the business, there could be thousands of dollars in savings or considerable value in understanding that information better. I can’t tell you what that is worth in your company without knowing more about it. Neither should someone trying to sell you a solution.
You can begin by talking to someone and choosing one problem worth investigating. If you would like help with that first assessment, Salient AI’s consulting work starts with the process, the people involved and what you want to improve.
If an employee wants to try, give them a defined assignment, suitable support and time to show you what they can do. If you need outside expertise, meet the practitioner and agree on a limited first piece of work with a clear cost.
Review the result together. What helped? What needed fixing? Did the person tell you openly about the difficulties? Would you trust them with a little more responsibility?
You won’t answer every question about AI through that first assignment. You will know more about the person doing the work, the support they need and whether their approach is useful to your business. That is enough to make a better decision about what to do next.
