Website Design Company Kelowna BC - Atomic 55

ai advice and professional services

The Time Spent Untangling AI Advice Is Professional Time

There is a growing amount of work arriving in professional inboxes that begins with somebody else’s AI conversation. A customer asks ChatGPT or Claude a question, receives several paragraphs of seemingly plausible, downright agreeable recommendations. What comes next? Hint: it's the opposite of a discount.

Data reflects the past, while human creativity shapes what comes next.
how can i tell if ai advice is bad

There is a growing amount of work arriving in professional inboxes that begins with somebody else’s AI conversation. A customer asks ChatGPT or Claude a question, receives several paragraphs of recommendations and forwards them. Or, and not better/not worse but they might try and assertively explain their position on how you ought to implement what are somehow now “their ideas” to the company they hired.

The answer took seconds to produce, which seems to have created the impression that dealing with it should require about the same amount of time and also, zero effort, really, if we’re getting down to it.

What they neglect to understand is that for the person receiving that email, the work has just begun. Before they can respond responsibly, they need to understand what was asked, what information was supplied, which assumptions shaped the answer and whether any, some or none of the generated recommendations apply.

They may also need to revisit an agreement, retrieve previous correspondence, or explain work that was already discussed and approved. All of that takes professional time, even when the eventual answer is that the original recommendation still stands.

At Atomic 55, we are seeing this become an increasingly frustrating part of providing technical and marketing services. Time that should be going towards customers’ businesses is being consumed by the need to untangle advice generated without an adequate understanding of either.

The promised efficiency of AI becomes difficult to appreciate when the resulting work lands on somebody else’s desk.

Somebody has to establish whether the advice applies.

An AI answer depends on the information available to it.

Customers understandably describe a situation through the parts they know. They see the website, and they know the requested change they want to make. They usually have very little reason to know about the software dependencies, previous customizations, testing requirements or responsibilities that influence how the website is where it is today. They don’t see the backend, server infrastructure, security and custom code that’s implemented.

They don’t account for how all of that functionality dramaticaly changes an outcome. 

And it’s those missing details that change the recommendation considerably. Asking what it should cost to update a website tells an AI very little about the website itself, what needs to change or what the provider has agreed to include. It can also generate wild requests like requesting access to data or systems that you should not be granting any AI access to. 

Yet the answer can arrive with enough detail and confidence to make the customer feel like they have potentially uncovered a problem. Their provider now has to reconstruct the circumstances around a question they were never asked, then assess a recommendation made without those circumstances. That investigation is work, regardless of whether it produces a change.

If the response explains why the advice doesn’t suit the situation, the conversation can start another round. The explanation goes back into the AI, another answer comes out, and another email arrives for review. Everyone involved can spend a considerable amount of time going through this process before returning to a conclusion the provider had already reached with the relevant information. I call it the AI loop. It takes FORever to get things done. 

We need to be more deliberate about the work we create with AI.

Years of working together should count for something; that’s why these exchanges are particularly difficult.  They quickly introduce suspicion into a relationship that has worked well for years. A company has supported a customer, answered their requests and taken responsibility for keeping services running, yet a conflicting AI answer suddenly raises the possibility that the customer has been overcharged or denied something they should have received.

Customers should be able to question their providers, and a longstanding relationship deserves honest explanations. That same relationship also deserves some consideration before an answer generated from a partial description is treated as the more trustworthy account of what is happening.

Your provider knows the history of your website, the work already completed and the arrangements supporting it. They have also been accountable to you through previous decisions and problems. Those facts should carry weight when a chatbot suggests that something could be cheaper, easier or handled differently.

A difference of opinion gives you something to discuss. Establishing that someone has been dishonest requires considerably more than an AI offering an alternative explanation, particularly when it has never examined the work or received the information needed to assess it.

I don’t think we’re saying don’t use AI. Because we aren’t, but there should be some ground rules. 

1) Ask us first. Most likely, it will save you a chat, and you can go spend time doing other things. Don’t you have other things to do? Like, I hate to be blunt, but maybe your own job?

2) Send us the full transcript from the AI. We might be able to send you back the correct questions/prompts to help you get a better explanation of what you’re searching AI for. We might also learn more about what you’re really looking for. We’ll definitely see your complaints 🙂

We got the list. Now. Who’s doing the work?

Generating instructions does not account for carrying them out.

Technology is especially susceptible to this because so much of the visible work happens through a screen. A customer sees someone change a setting, publish an article or upgrade a plugin, and the action itself can look straightforward. An AI can describe the steps just as easily, which makes it tempting to assume that following those steps covers the job. How does that old adage go? You get out what you put in?

Because if it were just that easy, then everyone would be doing it. And there are definitely a lot of wannabe web devs out there, but they’re missing experience, education and experience. Yes, so nice, I stated it twice. A website involves content, images, databases, software, hosting and connected services that all need to function together. Pull a lever over there, and you don’t know what might happen over here. Small changes can drastically affect other parts. That’s why even with AI you need someone with enough understanding to recognize what should be checked, what could go wrong and most importantly, how to respond if it does. The responsibility continues long after the update appears to have completed.

Even routine blog publishing involves judgment about formatting, image size, links (huge!), mobile presentation and how new material fits the existing site. Not to mention optimization and readability. And then our friend maintenance brings us some compatibility, backups, testing and security to the party. How much attention each task requires depends on the website and the circumstances, which is precisely the information a general set of instructions may be missing. It’s also critical that we take into account and understand your business. Like where you were and where you want to be. Because in so many cases we’ve kind of been ride or die in helping you build your company’s sales and marketing direction. We know a few things. 

When a customer forwards along instructions as a directive, somebody somewhere has to evaluate the work before agreeing to do it.  Being able to generate a checklist doesn’t remove that responsibility. It also doesn’t reduce the time needed to exercise the work properly.

There’s a lot of work involved in keeping things running today.

Have you noticed the recent vulnerabilities and outages across technology systems?  The news has plenty of reminders about what happens when the systems behind everyday business require urgent attention. The PHP2Shell vulnerabilities were a significant concern for us, while the separate Microsoft and FAA disruptions this November demonstrated the reach of technical failures. Microsoft and the FAA reported no indication of cyberattacks in those incidents, although the interruptions still affected the millions of people relying on their services.

We also know of a larger local business whose website was reverted to an older backup during its hosting company’s server upgrade in response to the PHP issue. Two months of blog posts and website updates disappeared from the live site, creating another problem for a business that had already paid to produce that work. Then they went down along with Microsoft and had no email or Teams for two full days.

Experiences like these explain why professionals like us spend time checking things that customers rarely see. Recovery, backups, testing and maintenance have never been more important because the consequences of lost time in a world that moves this fast are unfathomable. We’ve moved beyond instant gratification in our entitlement to time not spent on work. So, alongside the steady stream of attempted fraud, suspicious messages, and account concerns, that work already places substantial demands on the people responsible for it.

Adding pages of loosely informed directives to the same workload has a cost. Just remember, before you hit send, the person on the other end may have a finite number of hours, even when the supply of generated recommendations appears endless.

Congrats. Reviewing AI advice has officially made it into the scope of work.

There is a reasonable way to bring AI advice into a professional relationship. Explain what you want to accomplish, identify the concern and ask whether the suggestion would help. Give your provider the opportunity to assess it against the actual circumstances, including anything already established in your agreement or previous conversations.

That assessment may be brief, or it may require investigation. When it involves reviewing a lengthy transcript, evaluating alternative tools, preparing comparisons or repeatedly responding to revised recommendations, it has become an additional professional service. Businesses need to be able to identify and charge for that work instead of continually absorbing it into time allocated for something else.

Customers also deserve to understand that consequence before spending hours gathering recommendations. If your provider is already responsible for advising you on a subject, beginning with a direct question may resolve the issue with far less effort. If you want an independent review, it should be based on enough information to examine the service fairly.

We need to become more deliberate about the work we create with AI. Before forwarding another set of instructions, consider what you are asking the recipient to do and how much professional judgment will be needed to answer responsibly. The seconds you spent generating the advice bear very little relationship to the time someone else will spend untangling it, and that time, whether you like it or not, has now made it into the cost of the conversation.

is ai advice good

Ask yourself this: When technology goes sideways (PHP, Microsoft). Who's your support?

If you’re turning red in the face, don’t worry – it’s not just you!

It’s easy to get caught up in the lure of all these highly agreeable answers at your fingertips. You aren’t alone. Answers arrive and they are designed to make you feel smart, ingenious, and competent. Passing them along seems like the logical thing to do, right? But the thing most people are missing when they’re playing around in their new role of accountant, web dev, doctor, lawyer is the experience and knowledge of what came before. And folks, we have AI too. And we use it. But the difference between my prompt to make things more efficient on a website and your prompt lies in the 30 years of prior knowledge that forms my questions. 

Let us know if you want to talk more about AI and any questions you might have about your marketing and technology. 

TOP