Later Never Comes

Later Never Comes

What I saw driving past my childhood home this morning.

They say you can’t go home again. First of all, I don’t know who “they” are. But in this case, “they” weren’t wrong.

My mother-in-law was visiting last week. This morning I drove her to the bus station so she could head on to visit a friend. The bus station happens to be in the town I grew up in. So after I made sure she was safely on the bus and headed where she needed to go, I took the long way back. Past the old neighborhood. Past the house I grew up in.

I haven’t lived there in a little over twenty years. However, it’s where I spent more than half my life. The house itself is well over two hundred years old. It was in my family for about two hundred of those years. When my parents downsized after my brother and I moved out, they sold it. That was well over a decade ago.

What I saw this morning was heartbreaking.

The house looks abandoned. There’s insulation coming out of the windows. You can’t see the driveway anymore. The yard is overgrown to the point that you can’t tell where it ends. There are temporary structures scattered around the property that I don’t have a polite word for. The whole place looks like nobody has lived in it for a long time.

For more than two hundred years, that house was loved and maintained. In the hands of someone who wouldn’t maintain it, it took a fraction of that time to fall apart.

Look, I know turning a story about my childhood home into a business lesson is the kind of move that makes some of you want to roll your eyes. (Cue all the comments.) Roll away. I’m writing it anyway, because the house I drove past today is the same lesson I’ve been trying to teach about AI for years, and I’d never seen it stated quite this clearly until today.

Last week I wrote about why your AI program shouldn’t restart every time a new model ships. That argument is right. It’s also abstract. The house is the same argument made concrete, so I want to come back to it this week from a different angle. The part of building anything that decides whether it’s still standing in ten years is maintenance. And maintenance is the part nobody wants to think about.

What’s happening to the house is happening to your AI program

So here’s the pattern. Companies invest in something exciting. They stand up a new platform, an agent, a fancy workflow. There’s a launch. There’s a deck. There’s a quarter or two of attention. And then the attention moves on to the next exciting thing, because there’s always a next exciting thing, and nobody owns the part where you have to keep the original thing working.

It doesn’t fall apart all at once. It falls apart the way a house does.

This week, you don’t mow the lawn. Next week, you tell yourself you will. Next week comes and goes. The week after, you think, well, it’s almost winter, no point now. Next thing you know, the grass is past your knees and you can’t push a mower through it. You’d have to hire a landscaping company. You don’t have the budget. And even if you did, you’re not sure which landscaping company will treat your lawn the way you would’ve treated it yourself, if you’d had the time.

Inside the house, the same thing is happening. A leaky faucet. Some wiring that needs updating. A deck that needs to be re-stained. None of it is urgent. All of it can wait. You’ll get to it later.

Later never comes.

Here’s the thing. That’s exactly what’s happening with most of the AI tools and platforms companies invested in over the last two years. The setup was the exciting part. The launch announcement was the satisfying part. Then a new model shipped, a new vendor pitched, a new project got funded, and the system that was supposed to be a foundation became one more thing nobody actively owned.

By the time anybody noticed, the workflow was producing slightly worse output than it used to. The team was using it less. Half of the prompts had drifted from the documented version. The vendor pushed an update that changed behavior in a way nobody wrote down. The integration that connected it to the CRM had broken three weeks ago, and nobody had logged in to check.

You can fix all that. Same as the lawn, you’d have to hire somebody. It’d cost more than the maintenance would have cost. It’d take longer than the maintenance would have taken. And you wouldn’t be sure whether the people you hired to fix it would treat the system with the same care you would have if you’d built the time in to do it yourself.

The people need maintenance too

The same thing is true of the people who use these systems. You can’t train somebody once and expect the training to hold for three years. The tools change. The team turns over. New hires come in and inherit a workflow nobody has explained to them. People drift. That’s not a failure of the team. That’s what happens to any skill that isn’t practiced and refreshed.

So if you run this through the 5P Framework by Trust Insights™, the trap is split between two of the Ps: Process and People. Process maintenance is the documentation, the workflow review, the test that the system is still doing what it was supposed to do. People maintenance is the training, the office hours, the resources that are actually used and not buried in a SharePoint folder nobody opens. Both of them have to live on somebody’s calendar. If they don’t, they won’t happen. And the technology starts to look like the lawn.

What a real maintenance plan looks like

It doesn’t have to be elaborate. It has to exist. Here’s what useful looks like.

1. A monthly revisit of each AI workflow. Is it still doing what it is supposed to do? Is the output quality holding? Has anything broken quietly? Thirty minutes a month, per system. That’s it.
2. A quarterly refresh of the training. Not a new training. A refresh. What changed in the tool? What new examples have come in? What questions are people asking? Half an hour to an hour, four times a year.
3. Regular office hours. Somebody available, on a published and consistent schedule, to answer questions in real time. People won’t file a ticket. They’ll drop in.
4. Resources people will actually use. That means a working document in a place people already look. Not a buried SharePoint folder with a name nobody can remember.
5. A named owner. One person responsible for whether the system is still working, who has time on their calendar to maintain it, and who’s allowed to say “not yet” when somebody wants to add the next exciting thing on top of a foundation that hasn’t been kept up.

The moral of the story

As I’m writing this, I’m still emotional about the current state of the house. It was the only home I knew as a child, and there’s 200+ years of family history wrapped up in it. I had no control over what happened to it. Whoever owns it now made their choices, and I have to make peace with that.

What you do have control over is the systems you’re responsible for right now. The AI workflow that’s still working but hasn’t been looked at in a quarter. The platform that was the highlight of last year’s budget and is producing the same quality of output as last year, which is not the same thing as being maintained. The people you trained last spring who haven’t had a refresh since.

When you invest in something, big or small, build the maintenance plan at the same time. Don’t skip it. Don’t defer it. Don’t promise yourself you’ll get to it later, because you won’t.

Build the plan. Name the owner. Put it on the calendar.

Later never comes.

What maintenance are you putting off?

Reply to this email or join the conversation in our Free Slack community, Analytics for Marketers!

– Katie Robbert, CEO

Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.