The Puppy Project

Choosing a dog for a household with five children is not a low-stakes decision. Ours range from eleven years old down to two, the family calendar is a contact sport, and everyone had an opinion about the puppy long before the puppy had a name. So I did what nearly three decades in software has trained me to do with any decision that matters. I turned it into a project.

The instinct is simple: take something ambiguous and emotional, and give it enough structure that you can reason about it. Not to remove the emotion, but to make sure the emotion is working with good information instead of against it. A wagging tail is persuasive. It is not, on its own, a decision.

More data than most people ever look at

The litter came with a surprising amount of raw material. Eight weeks of weekly weight logs for each puppy. The measurements of both parents. And a full set of Volhard temperament test scores, the standard aptitude assessment breeders use to read a puppy’s tendencies at around seven weeks.

Most families glance at all of that and pick the cutest face. I wanted to see what the numbers were telling us.

From scattered inputs to a decision

Working alongside AI as a thinking partner, I turned those inputs into a small decision package:

• Adult size projections for each puppy, built from several methods rather than a single guess and anchored to the parents’ size
• A temperament comparison that translated raw test scores into plain-language personalities: the bold one, the gentle one, the big and unflappable one
• A recommendation matched to our real constraints, a busy and active home with two children still in the grab-everything phase
• A side-by-side decision card for the two finalists
• A training protocol with age-based milestones for the pup we chose

None of that was automation. The AI did not make the choice, and it was never asked to. What it did was organize the evidence, hold several methods side by side, surface the tradeoffs, and keep the analysis honest while the judgment stayed with us. That distinction is the whole point, and I will come back to it.

The framework did not make the choice. It made sure that when we met her, we already understood who she was and what she would need.

The decision

The analysis pointed to one puppy in particular. In the spreadsheets she was Miss Mint: the bold, tough, adventurous one. She also happened to have the highest tolerance for handling in the entire litter, which is not a small thing when two of your five children are still learning the difference between petting and grabbing. The data and the family arrived at the same place, which is the most reassuring outcome an analysis can give you.

Then we met her, and that was that. Meet Marley.

Why this is the day job

I run Xyberal Solutions, where a version of this exact pattern is the work. Clients rarely lack data. They lack the structure to turn it into a decision they can stand behind. The most valuable thing modern AI does is not replace the expert in the room. It compresses the distance between a pile of messy inputs and a clear, defensible choice, so the person doing the deciding can spend their judgment where it counts.

Used well, AI is an amplifier for good thinking, not a substitute for it. The questions stay human. The accountability stays human. The work of organizing the evidence does not have to.

That holds whether the decision is a product roadmap, a technology investment, or a Labradoodle.

The spreadsheet has been archived. The dog has not. The training plan starts this week, the chewed shoes are presumably already in motion, and five children are learning that a well-run project and a very good decision can, on occasion, be the same thing.

Brian Humphrey leads Xyberal Solutions, an IT consulting and fractional technology leadership firm that helps teams turn ambiguous problems into structured, AI-enabled decisions.