AI-written roadmaps can look more rigorous than they are
Pushing back carries political risk, especially when the author is your VP
Ask about customer evidence, editorial judgment and assumptions
Keep the discussion about the strategy, not the AI tool
Use the answers to build a learning loop around the roadmap’s biggest risks
Hey folks,
I imagine this is a scene that’s starting to play out in more and more team meetings. Let me know if this sounds familiar.
Your VP, the one who is always asking you for the latest product roadmap, shows up at your weekly check-in, opens a laptop and shares a roadmap “they” created.
It’s unusually clean. The rationale is organized, the sequencing makes sense, the market framing is competent in a way that quarterly planning documents usually are not and somewhere in the first two minutes it becomes obvious to everyone in the room that Claude, or another AI model of the same kind, wrote most of it.
Nobody says anything, of course. That would be a career-limiting move in most cases.
The VP asks for reactions and the awkward silence that follows is indicative of a group of people who all have the same objection and none of whom want to be the first to speak out.
So, what do you do in this situation?
AI-written roadmaps are harder to push back on than human-written ones. Here are three reasons why.
The first is that polish reads as rigor.
We are trained to treat a well-argued document as evidence that someone did the hard thinking and distilled their core ideas into this polished deliverable.
That’s not necessarily true anymore.
Of course, your VP could have done that, but you don’t know and that’s risky.
The second is the political risk that comes with critiquing the boss.
This is the thing that keeps most people quiet.
Most of you work in organizations that spent the last eighteen months telling everyone to use AI. Questioning the output of the AI tool feels, in the moment, like questioning the new mandate.
The person who does it more than once can easily start getting mentioned as the one who “struggles with the new way of working.”
That’s a real risk and shouldn’t be discounted when figuring out why people won’t give feedback to a superior.
The third is that you have no idea which parts of that document your VP disagreed with and kept anyway.
They may have fought Claude on the sequencing and lost interest halfway through, or read every line, or skimmed it once on the way to the meeting.
All three behaviors produce an identical-looking document.
Of the three, this one is probably the riskiest to navigate.
To get around this new dilemma, your goal should be to ask questions that would be reasonable no matter who wrote the document.
And don’t mention the AI tool in your questioning and critique.
The moment you say, “Well, because AI wrote this,” you’ve converted a roadmap strategy conversation into an argument about AI, and you will lose that argument in front of your VP even when you’re right.
The following three questions start you down the path of giving feedback without limiting your career. The questions are worded a specific way. Your mileage may vary as to how to finesse them into your particular conversation.
“What customer evidence is behind this?”
This is the question you would ask a human author.
Since an LLM can produce confident market rationale out of pattern recognition rather than real customer data, this is a strong place to start.
Your VP could come back with clear customer feedback that drove their thinking. They could also come back with a research round from eighteen months ago or a competitor’s positioning page.
In this latter case, at least you know where the foundation lies for this document.
“Which parts did you struggle with and decide to keep anyway?”
This one, worded well, reveals how much human effort was put into the document.
It assumes your VP did the editing rather than accusing them of not doing it. It’s a compliment, really. And then it tags on a clarifying question at the end.
Anyone who truly worked through the editing process of the document will have an answer ready and will usually enjoy giving it, because it is likely the part they’re proudest of.
“What would have to be true for this to work?”
I’ve borrowed this one from Roger Martin‘s strategy work for years and it’s super useful in this scenario.
It moves the conversation from whether the document is “good” or not to what the document is betting on.
It also surfaces the assumptions the LLM inherited from the broader internet’s consensus strategy approaches rather than your organization’s specific conditions and market realities.
None of these questions require you to have or voice an opinion about AI.
Ask the three questions, note what you get and pay attention to the second one in particular.
If your VP cannot name a single thing they overrode, they have handed you a document that, while it has their name on it, doesn’t have a clear owner yet.
That’s good information to have at the outset rather than at your post-delivery retrospective.
It’s also your opportunity to jump in and “own” the future iterations of the roadmap.
The first and third questions speak to the assumptions, or risks, that the proposed work is based on.
This is your opportunity to then share with your VP what you believe to be the biggest of those risks and how, using AI tools (well done there), you can build a learning loop to help de-risk them and make the new roadmap even more solid.
This way, not only are you volunteering to help your boss move their ideas forward, you’re doing it in line with the corporate mandate to adopt more AI into your ways of working.
At the same time, you’re ensuring that, as the roadmap evolves, it’s taking an increasingly evidence-based, customer-centric approach.
The issue isn’t whether AI wrote the roadmap.
The issue is whether someone can explain the customer evidence, the judgment calls and the assumptions the roadmap is betting on.
If this comes up for you, give these questions a shot. As I said earlier, word them in a way that makes the most sense for your context and relationship with the VP.
I’d love to hear how this lands and if this helps you bring your strategy and roadmap deliverables into better alignment with your customers’ needs and overall corporate strategy.
Jeff
jeff@senseandrespond.co – Got any questions? Email me and ask me anything, or just hit reply to this email.
P.S. Join us on September 10, 2026, for our first free webinar of the fall. Josh and I are taking on a big question: “What does product management look like in 2027?”
Free, live, and we’ll take your questions.
Can’t make it? Register anyway and we’ll send the recording.
I am on vacation in August. Every year both Josh and I take time to unwind, think about the business, NOT think about the business and just blow off some steam.
This year I am headed to Asia, China and Korea to be exact and I can’t wait to share with you what our adventures look like.
I’ve never been to mainland China. I did go to Hong Kong a few years ago and that was magical. I have been to Korea once, for work and don’t remember much, so I am excited to be heading over there in a non-work capacity.
Pics and updates when I get back.
So, for now, to hold you over, here’s a recent photo of our dog, Loki.
By the way, two of my favorite ways to interact with folks in our industry are through public speaking and podcasts. If your organization is looking for a compelling, high-energy, and fun keynote for an upcoming event, keep me in mind. You can see me in action here.
And, if you run a podcast or know of one that would benefit from having me on as a guest, please send me an email or a referral. I’d love to do more of these as well.
I’ve picked up The Odyssey. You know why. The movie is awesome. Going back to the original text is fun too.
The Strokes have a new record out called Reality Awaits. Love this band, but I’m not loving this record so much. I’m working on increasing my love for it, though.
The Bear, Season 5. The characters on this show are so good. I just want to hang out with all of them. I’m sad this show is ending. It will be hard to see these actors in other roles. They’ll always be Syd, Carmy, Richie, Nat, Fak, T, Marcus, Ebra, Garry, and Luca to me.
Interested in working together? Please reach out.
In case you need it, here's a description of what I do.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.