Everyone has been talking about AI for a while now, but when I speak with most business owners they’re still trying to figure out if AI is useful and where it could ever make a difference in their business. Our market has built up this huge expectation that AI will transform how we work and we feel this obligation to make it work, but most of us are at a loss for how it could save money or improve our operations.
The bottom line is the Jetsons are still a pipe dream – we’re still washing our own dishes, emptying our own garbage, cooking our own meals, and getting stuck in traffic.
A Mixed Bag
As a web design company, we’re the “tip of the spear” on testing AI within our day to day work. As you can guess, AI has done a great job in speeding up our coding, helped us solve and helped us learn new ways to solve things that previously would have taken hours or even days. On a rough estimate it helps with about 50% of our web development, and shaves off about 50% of our time when/where we use it (so a 25% gain overall).
In our exploration we’ve found AI surprisingly helpful with some things, but shockingly bad at other things. AI has been amazing at web and software development, content writing, and annotating meetings, but we’ve also found AI is really bad at graphic design, coding the visual layout of a site, and doing its own fact checking. It’s been a bit of a mixed bag for us – discovering where it’s useful and also finding out where it fails.
The Big Wins
So if you aren’t in web development, is it worth chasing AI solutions as a business? Absolutely.
Alongside the debatable benefit of AI replacing our entry-level workers, we have seen some real wins that we previously couldn’t have.
For our team the big wins are:
- When you can do something that was previously impossible (even small things)
- When you replace a subscription or software cost
Call us old fashioned, but we’re not really thrilled when we replace people with technology, however we love it when we can replace headaches and business costs with new tech.
Doing the Impossible
For me, these are the hidden gems within the world of AI – when you can utilize AI to build an application that couldn’t have been built 5 years ago – either because the coding languages are too niche (eg. Google Drive automations) or because LLMs are better at processing language. Here are a few we’ve built this year:
Cloning Folders
In our work we use Google Drive. But as a company which loves templates and clear file structures, we were more than frustrated when we found out that Google can’t (won’t) clone a folder structure with a perfectly organized set of files and templates in it. We also don’t employ anyone who specialized in building Chrome extensions… until AI came along.
We now have an in-house (and completely secure) folder cloning software which can spin up a new client folder any time we need.
Tagging 17,000 Posts
We also work with a lot of clients who have us rebuild existing websites. One of our recent builds included a website with over 17,000 articles – and a client asking if we could tag each article with the applicable label for sorting them. In the past this would come down to hoping each article has one of the keywords in the content (and hopefully not all the keywords). Today with AI we’re able to scan all 17,000 posts with an LLM that can assign the best match to each article.
Building a Site Map
Another example is something as simple as building a site map for an existing website. We haven’t found a single online tool we’ve liked, and ChatGPT gave us different results every week – sometimes helpful, but often missing things. But with a combination of code plus AI, we can get the results we need in the format we want every time.
We’ve built a lot of these micro-programs with the help of AI, and also built micro-programs that send information through AI for processing. Between the two, we’re finding great business cases that are legitimately changing how we do things.
Replacing Subscriptions
This is the second, and more ambitious area we’ve been looking at. With so much of our world turning our software and services into little vampires that bleed our businesses, we’ve been looking at ways to replace monthly costs with proprietary software.
Project Management
Over the years we’ve tried 5+ different project management subscriptions, the latest being ClickUp. They’ve been necessary, but also a huge source of frustration for our team:
- Changing the interface and moving things around every year
- New features breaking old features (eg. wiping out the automations we set up)
- And of course the cost – $320/year per team member adds up really quickly when everyone on your team needs it
Considering how simple our needs are, we’re now looking at building our own – not to sell – just to use internally. It will mean our interface isn’t changing all the time, our automations won’t be overwritten, and if we divert our down-time into development time, we could save thousands of dollars a year on this.
Plugins
We’ve also been looking at our website plugins. They don’t cost very much, but for simple things, like writing meta data for SEO, this is another area we can cut costs and increase what we automate.
But not Everything
Unfortunately there are limits to what we can reasonably replace with simple AI-assisted applications – some good examples being the Adobe Suite and our password management software. Programs which are high complexity (eg. Adobe Photoshop) or high risk (all your company passwords) would require far more development hours than cutting your subscription would save (even with Adobe jacking rates up over $1100/year per user).
There are also migration costs to consider. For example switching from Google Drive to your own system could look attractive, but migration would entail months of endless headaches.
Terminating Tedium
As a business we’ve begun booking monthly meetings with our team just to uncover new tedious tasks which we can terminate with these micro-programs. In each meeting we review how well these solutions have been working, get updates on the new solutions we’re creating, and assign new micro-programs to build.
It also brings the full team into the concept of “working on the business” rather than just “working in the business”. It’s been amazing to see the new ideas that come out of our team, and a huge blessing to see everyone brainstorming together on these things.
Our business looks a lot different than it did a year ago, and it’s going to look a lot different a year from now. So even if you haven’t started using Claude or ChatGPT or Google’s Gemini in your own toolkit, it’s well worth booking a meeting (or several) with your team to start asking how AI and micro-programs can be used within your company.
Chances are, you’ll uncover more opportunities than you expected.
Have a Micro-Program You’d Like to Build?
We’d love to help!
