THE METHOD

What 20 Years of Selling Software Taught Me About Reading a Product Roadmap

2026-09-20

Twenty-plus years in enterprise software, split between selling it and later managing the vendor partnerships behind resold solutions, means sitting through more roadmap presentations than anyone should have to. Most of them look the same: a quarter-by-quarter grid, a row of checkmarks, a slide that says coming soon next to something that's been coming soon for three straight decks.

The first real tell for whether a roadmap item is genuine: can the vendor name the specific customer or problem driving it, or do they retreat to customers have asked for this. A specific answer means someone is actually building toward a deadline they're accountable for. A vague one means it's a placeholder to keep a deal warm.

There's a second, quieter version of that same tell worth naming, learned from being on the other side of the table managing partner relationships: watch what happens when you ask the same roadmap question twice, a quarter apart. A genuine item moves, even if it slips. It gets a new date, a narrower scope, or an honest we deprioritized this for a specific reason. A fake item just repeats the same slide verbatim, because nobody actually owns it, and nobody has done any work on it between decks.

The AI tool space runs on the exact same pattern, just faster. Reading through the newsletters we track for this site's Radar page, coming soon: agents shows up constantly with no specifics attached, while the launches that actually ship tend to describe one narrow, concrete use case instead of a sweeping capability. The specificity is the signal, not the ambition.

The compressed timeline changes the stakes, though. In enterprise software, a roadmap slide that never shipped might cost you a renewal conversation eighteen months later. In AI tools, the equivalent slide might be describing a capability you're already deciding whether to build a workflow around this quarter. The tell is the same, but the cost of missing it arrives much faster now, which is exactly why it's worth being more disciplined about checking for it, not less.

The second tell, one that took longer to learn: watch for a feature that quietly disappears from the next update instead of shipping, with no acknowledgment it was ever promised. That's not an oversight. It's usually a sign the scope got cut somewhere internally and nobody wanted to announce it.

There's a related pattern worth flagging separately: pricing tiers that reference a capability as coming soon for six months or more without qualification. In our experience managing vendor relationships, that phrasing almost always means the feature is stuck behind a dependency the vendor doesn't want to name publicly, whether that's an integration partner, a compliance review, or an internal resourcing fight. It rarely means the feature is imminent.

The practical version of this for anyone evaluating an AI tool: judge it on what it does today, and treat every roadmap slide as marketing until it actually ships. It's exactly why this site keeps a hard line between the Tool Directory, which has been tested, and the Radar page, which is just what's been spotted and promised so far. If you take one thing from two decades of watching this pattern play out, it's that the gap between promised and shipped never closes on its own. Someone has to actually build it, and until they do, the slide is a hope, not a feature.

One more version of this worth flagging, specific to how vendors talk in demos versus decks: watch for the phrase in beta with select customers attached to a capability with no named customer and no specific date for general availability. In our experience, that phrase can mean genuinely early and promising, or it can mean built once for one customer and never generalized, and there's usually no way to tell which from the outside except by asking directly and watching how specific the answer is.