Notes From the Work
Why AI rollouts underdeliver, how to tell whether one is working,
and when a model is the wrong tool for the job.
Why Nobody Trusts Your Dashboards
When two reports disagree, people stop trusting both. Why dashboards lose credibility, the four usual causes, and how to win the trust back for good.
When a Startup Should Bring In Outside Engineers
Hire, contract, or bring in a small firm? How startups can add engineering capacity without losing control of their own codebase.
What to Automate First (and What to Leave Alone)
Not every manual task is worth automating. How to pick the ones that pay back, which to leave alone, and how to automate without creating new problems.
Signs Your Team Needs Its Own Internal Tool
When an inbox, three spreadsheets and a group chat are your workflow, a small internal tool can pay for itself. The signs, and what to build first.
What Hiring a Small Software Firm Looks Like
From first call to hand-off: what a project with a small, engineer-run software firm should look like, what to get in writing, and the red flags.
Custom Software or Off-the-Shelf? How to Decide
Most small businesses should buy software, not build it. How to spot the few cases where custom wins, and the cheaper middle ground most people miss.
How to Connect Your CRM, ERP and Accounting Tools
Re-typing orders between your CRM, ERP and accounting system is a tax you pay every day. How integration works, what it costs, and where it breaks.
Signs Your Spreadsheet Has Quietly Become a Database
Spreadsheets are the most successful business software ever made. How to tell when one has turned into a database, and what to move it to.
What a Two-Week AI Audit Actually Looks At
A plain description of what gets examined during an AI Second Opinion, in what order, and what the verdict at the end can and cannot tell you.
Nobody Agreed What "Working" Meant Before It Shipped
The most common reason an AI rollout cannot be evaluated is that no baseline was ever recorded. Here is how to reconstruct one after the fact, and what to do differently next time.
Your AI Pilot Isn't Underperforming. Your Data Is.
When an AI rollout disappoints, the model is almost never the culprit. It's usually one of three things underneath it — and you can check for all three yourself before paying anyone to look.
When You Don't Need Machine Learning
Machine learning gets applied to problems a query would answer, and skipped on problems it would genuinely solve. A practical way to tell which one you have.
Recognize your own situation in any of these?
Tell us what's going on. You'll hear back from one of the engineers who'd do the work.
Start a project