The Everything SaaS
Announcing a new service from Kelsus
I usually write about whatever is interesting to me and then only marginally connect it to our business in the conclusion. This time, it’s a full-on sales pitch. I’m brewing some coffee because you’re about to be closed (and coffee is for closers).
As a mentor for several incubators, I’ve had a front-row seat to the entrepreneurial reaction to GPT. And I’m so glad I’m not running a startup that just raised VC money on the promise of some solution delivered on the GPT platform. During this week’s keynote, Open AI basically drained all the moats—or at least showed they have no intention to help anyone maintain any kind of technology moat.
Here in Startup Win! recently, I’ve written a lot about how the key thing that AI unlocks is completely customized software that does exactly what your business needs it to do. Meanwhile, incumbent SaaS providers like Salesforce or QuickBooks always prioritize whatever will sell the most seats of their product which may not align with your business’s specific needs.
It used to be that writing custom software for all of your business’s specific needs was grossly too expensive. SaaS companies spread the software development and maintenance cost across a ton of customers, and survive off this economy of scale. AI drops the price of one-off software solutions to a point that it becomes worth it to automate things that used to be more affordable with inexpensive manual labor. And yet there are a bunch of entrepreneurs running around trying to use the same SaaS playbook. It’s the wrong move.
The right move is to start saying yes to software perfectly customized to your business operations and build it for yourselves. We’ve already seen this to be true here at Kelsus. Over the years, I’ve been approached by developers that wanted to automate some aspect of HR (like timesheets) or client management, and I’ve always said no. I myself have wanted to automate certain aspects of invoicing, payroll, or onboarding but I’ve always reminded myself that a few hours a month doing boring manual work is still cheaper than having some custom software that we need to maintain.
Well, now it’s worth it. We’ve been hard at work inside Kelsus automating all the things.
And now the pitch: We’d like to do this for your company.
We believe that it doesn’t come naturally for many developers to become efficient with GPT because it requires changes to how you plan and work that many developers are reluctant to make. Your internal IT team is already doing business as usual, and isn’t thinking in this new way.
We also believe that many managers and line operators just aren’t aware of the vast array of opportunity for automation within the business. Some areas to think about:
Do you still rely on people to extract information from PDFs?
Do you have to nag people to do some manual thing on a regular basis? (We recently wrote a Slackbot called WorstNightmare that nags people that haven’t done their timesheets, but it first checks the vacation calendar.)
Do you rely on people to put information from another system into another?
Do you have financial people that make sure your invoices “look about right”?
Are there any analysts who are responsible for reading a bunch of stuff and keeping an eye out for something unusual or important? (We wrote a system called Kelsus Constellation that scrapes websites and intelligently filters the results looking for promising businesses to acquire.)
We have a million more ideas, but I said I’d keep it short.
We’d like to pair one of Kelsus’s product managers with direct support from me and a team of two Kelsus developers to start automating the living hell out of every operational process at your business. They’ll be responsible for not just building new automation tools but also for maintaining all the ones they’ve already built. And over time, again because AI is so damn good these days, they’ll organize the systems they build into a clean, clear, extensible architecture rather than a nasty mess of stray cats cluttering up your AWS accounts and angering your IT and security teams.
This is a call to anyone with enough authority at a business to see the value this could provide, and allow us to work outside of the cadence of your regular IT and product development team.
One more thing I’ll throw out there. If you want one or two of your own team members to join us, we’ll be happy to have them along so long as we use our experience and knowledge to drive the process. Another consultant friend refers to this as “happy to have them join so long as they understand they're coming to our party.” The benefit of this, would be of course for these developers to pick up some of our techniques for producing highly customized software at alarmingly fast rates.
Thanks for reading, and I know this is too short to have answered all your questions, so please get in touch and I’ll answer the rest (not via a bot).
Replying to this email is the easiest way to get in touch.
—Jon Christensen
P.S. You might wonder about our view on the balance between automation and oversight. Here’s how we think about this. Anywhere something is automated (especially with GPT), it should be treated as though a person is doing that thing. And as with anything that any person (no matter how conscientious) does mistakes are possible. So, if mistakes are irreversible and costly, then work should be double-checked by other people.
So we think when automating things, if mistakes are not that big a deal, let them be fully automated; if mistakes are expensive, then make sure that real humans are in the loop—even if all they have to do is double check GPT’s work or monitor an errors queue.