If you can't say what your AI did last month in business terms, you're not measuring it. You're admiring it.
Actions taken, not proposed. Revenue traceable to work no human started.
Pick three. Baseline them this quarter.
There's a race to put AI agents in chat. I think chat is the wrong home.
An agent that updated a forecast should leave the forecast — not a message about it.
Work belongs where the work is.
techcrunch.com/2026/07/21/ja…
Models are a commodity. Your context isn't.
Everyone has the same frontier intelligence at the same price. That edge lasts a quarter.
Accumulated judgment is the only part of your AI stack a competitor can't buy.
Marketing will be the first department to genuinely run on agents.
High-frequency, mostly reversible, data-dependent — exactly where autonomy pays first.
The scarce skill won't be execution. It'll be judgment written down as a spec.
Four platform shifts since 1989. Every one was going to end developer tools.
Every one moved the craft up a level and demand went up.
What changed: tools now have to be legible to a machine as well as a person.
infragistics.com/
GTM used to be a plan. It's becoming a system.
A plan is a document you defend in April. A system runs while you sleep.
If your GTM only moves when someone opens a dashboard, you have a plan with better graphics.
slingshotapp.io/
Static dashboards are becoming a liability.
Users talk to software now. A grid of charts they have to interpret themselves doesn't read as thorough — it reads as dated.
If a customer asked your product a question today, would it answer?
revealbi.io/
Everyone's arguing about whether to trust AI agents. Wrong question.
The question is how much, at what value, in which direction.
If you haven't written your autonomy thresholds down, your agents aren't governed. They're supervised.
First question I ask at any agent demo: what did it finish?
Not what it found. Not what it recommends. What's done.
Most agent demos are search demos with better manners.
slingshotapp.io/
Most Q4 plans assume the AI you have today is the AI you'll have in December.
That hasn't been true for three years running.
Plan for the curve, not the snapshot.
A dashboard that doesn't end in a decision is an expensive screensaver.
If someone has to leave your app to get an answer, you've added four steps to every decision.
Most people skip them.
slingshotapp.io/
A recommendation is another item on your to-do list. Completed work is a shorter list.
If your AI generates suggestions your team then has to action, you bought an assistant.
You didn't buy leverage.
wordstream.com/blog/marketin…
You can't prevent sprawl. You can decide whether your tools invite it.
Agents are cheaper to create than SaaS apps, and they bill continuously.
If you can't name who owns your agents, you already have a problem.
When I started Infragistics, grit was the whole strategy.
I was wrong about that.
Grit gets you through the hard moments. Intelligence compounds the good ones.
#FounderLessons#Leadership#AI
Every 12 to 18 months I ask: what version of this company needs to be retired?
Not what to fix. What to retire.
The habits that got you here aren't automatically the ones that get you there.
#Leadership#FounderLessons#vNext
The tools that actually get used don't just show you what happened.
They help you do something about it.
Insight without action is just awareness.
#AI#FounderLessons#Leadership#EnterpriseAI
The tools that actually get used don't just show you what happened.
They help you do something about it.
Insight without action is just awareness.
#AI#FounderLessons#Leadership#EnterpriseAI
Your AI isn't underperforming because the model is wrong.
It's underperforming because the data is fragmented.
Centralize first. Everything else follows.
#DataStrategy#AI#Leadership#EnterpriseAI