Sometimes, and more often than a year ago. But the question to start with is which process the software supports, and what that process costs you, rather than what the AI can do. If you can name the process and put a number on it, you can decide whether to keep the software, build AI around it or replace it. If you can’t, no tool is going to make that decision for you.
Most businesses reach this question through a renewal bill. The annual invoice arrives, it’s gone up again, and someone asks whether AI could do the job instead. Here’s how we work through it with clients, in enough detail that you can run it yourself.
We touched on the short version in the two questions business owners keep asking about AI, and one line from that post is worth repeating because it holds up every time: the businesses getting real value started with a problem, not with the technology. The ones who started with the technology mostly ended up with a subscription and a story.
Where should you start?
With the gap between what you pay for and what you use. We call it the subscription-gap test, and it takes about an hour with whoever actually uses the software every day.
- List what you pay for. Every platform, every seat, every add-on, with the annual cost. Include the ones that are charged to a credit card rather than an invoice.
- Mark what you actually use. For each one, which features does your team open in a normal week? Be strict. “We might use the reporting one day” goes in the unused column.
- Count the workarounds. Every spreadsheet that exists because the software can’t do something, every export that gets reformatted by hand, every step where someone copies information from one screen into another.
The workarounds are the important column. They’re where the software stopped fitting your business and your staff filled the gap themselves.
Once you’ve found a gap, pick the one process that sits in the middle of it and run three questions over it.
How often does the process happen?
Frequency is what makes automation pay. A task that happens fifty times a day repays setup costs quickly. A task that happens four times a year almost never does, however tedious it is when it comes round.
Daily beats weekly, and weekly beats monthly. Anything rarer usually belongs on a checklist.
How much of it is judgment, and how much is moving information?
Break the process down step by step and sort each step into one of two piles.
The first pile is information moving: copying a customer’s details from an email into a CRM, pulling numbers from three systems into one report, checking an order against a stock list. This is AI territory now, and it’s usually the bulk of the hours.
The second pile is judgment: deciding whether a quote is worth chasing, reading a tricky customer complaint, weighing up which job gets the crew first. That stays with a person. What changes is the information in front of them when they make the call, because the first pile has already been done.
Work that moves information is where AI earns its keep. Work that weighs things up stays with a person, with better information in front of them.
What is the time actually worth?
Put a number on it. “We’d save heaps of time” can’t be weighed against a quote.
The arithmetic is simple. Hours the process takes each week, times the loaded cost of the person doing it (wage plus super, leave and overheads), times the weeks in a working year. As an illustration: six hours a week at a loaded cost of $55 an hour, over 48 working weeks, is $15,840 a year. That’s the ceiling on what fixing that process is worth, before you count errors, delays or the better work that person could be doing.
If automating it costs a fraction of that, it’s worth doing. If it costs more than two or three years of the time it saves, leave it for now. Either way you’ve made a decision with a number behind it, which a general ambition to “do more with AI” never gives you.
What are the three answers you can land on?
Run the questions and you’ll end up at one of three places.
Keep the software. It fits the process, your team uses most of it, and the price is fair. That’s a good result: you’ve renewed on purpose rather than by default.
Build AI around it. The software is fine at its core job, but the work between it and your other systems is eating hours. This is the most common answer we see. AI and automation handle the moving of information between the platforms you already have, and nothing gets replaced. It’s often the cheapest of the three and the quickest to show a result. If your CRM is one of those systems, what CRM integration actually involves covers the practical side, and our AI and application integration work is built around this middle path.
Replace it with something bespoke. The gap is wide, you’re paying tens of thousands a year, and your process has been bent out of shape to fit the tool. Then a purpose-built application can cost less over a few years than the subscription, and it fits the way you work. It’s a bigger decision with a longer payback, and it starts with a proper scope. We’ve set out what that involves in what scoping a custom web application involves.
What have we learned the hard way?
We’ve been adopting AI inside HyperWeb for about three years, and not all of it has stuck.
Early on we paid for a platform that was meant to take work off us. It didn’t. It still wanted the same repetitive data entry, and every record keyed in by hand was another chance to get something wrong. What it was asking for was exactly the work a database lookup should have been doing, with AI handling the entry and the errors that came with it. We’d bought the tool and left the actual problem sitting where it was.
The lesson has been the same each time. When we started with a process we already knew was costing us, the result lasted. When we started with a tool that looked promising, we usually paid for it for a few months and then cancelled it. The three questions exist because we needed them ourselves. The longer version of that sequence, purpose before the platform, is set out in our 4P framework for AI adoption.
Frequently asked questions
Can AI replace the software my business pays for?
Sometimes. The test is the gap rather than the technology: how much of the product you use, how many workarounds you run, and what the subscription costs against what a purpose-built alternative would. Start with one process, not a platform decision.
Which business processes should be automated with AI first?
High-frequency, low-judgment work: moving information between systems, assembling reports, triaging enquiries. Work that needs someone to weigh things up should stay with a person, with better information in front of them.
How do I work out if automating a process is worth it?
Put a number on it. Hours the process takes per week, times what that time costs, against the cost of automating. A named process with a named cost is a decision you can make. A vague ambition to “use AI” gives you nothing to decide.
Do I need to replace my software to get value from AI?
No. Often the win is AI working between the systems you already have. Replacement is the answer when the gap between what you pay for and what you use is wide and the workarounds are eating staff time.
Want us to run it for you?
This is the process we run in the second half of an AI readiness review. We look at your systems and your software costs with you, run the three questions over the processes that matter, and give you a short list of what’s worth doing first, what’s worth leaving alone, and roughly what each would cost.
Or if you’d rather talk it through first, get in touch with the team.
By Brendan Brooks, Founder and Managing Director of HyperWeb. Brendan has spent 25 years building websites and business applications, and the last three putting AI to work inside his own agency and his clients’ businesses.
