Automation often happens around rather than through a problem.
For a long time there was an argument that cashiers would be the last to be automated: greeting the customer, handling the cash, passing on the order. These tasks seemed hard for a robot to do (and they are): therefore the cashier was safe!
But of course there are no cashiers now. Your order you burger through the app, the order is passed to a screen in the kitchen, and payment happens in the cloud.
The system automated ‘around’ the cashier:
- abstracted the job
- cut it into pieces
- threw some away (Good bye human interaction!)
- divided the rest between some systems (banking, mobile apps, APIs)
- transferred the data entry task to the customer
Those companies spending millions on humanoid cashier robots missed the point.
Be the water: automate the easy, flow around the hard.
Giraffe has the same thing. Take analytics for example. Analytics is basically a pivot it table. It filters, aggregates and groups the properties in the Giraffe model. We can polish this thing until it shines. Jobs that use to take 10 clicks now take 4 clicks. Defaults that used to be annoying, are no longer annoying.
But, there is a high risk analytics will be automated ‘around’. That the whole sequence of defining the same calculations again and again is redundant given widespread models and agents.
This example is illustrative: first, I drew this building in Giraffe
Second, I copied the JSON that defines this building into chatGPT, and asked it for an investor ready proforma.
.png)
Now, I don’t know if this is right. The LLMs are prone to hallucination. But good ‘prompt engineering’, or even a different finance application could easily automate ‘around’ analytics.
Our team has built so many automations for so many organizations - if you're keen to talk reach out!
