A CPO at a 320-person B2B SaaS told me last week that she has started joking that her roadmap reviews are an oral tradition.
Every quarter she runs a review with her four Group PMs. Every quarter she finds herself re-telling them the strategy the leadership team picked at the offsite. They find themselves re-telling her what their teams have been hearing from customers. By the end of the meeting, everyone has been brought up to speed for the third time on the same set of things. The roadmaps that ship the following week look almost — but not quite — like the strategy the company is supposed to be executing.
She said: “None of it lives anywhere. I leave each review feeling like I just spent ninety minutes doing what a document should be doing.”
That meeting is happening in a lot of product orgs right now. If you run one, you know.
It’s worth naming the two specific streams of context that keep breaking, because they break for different reasons in opposite directions, and they need different fixes — but the fixes turn out to share the same substrate.
The two streams
The first stream is alignment. It flows top-down. A strategic decision gets made — at an offsite, on a Tuesday call, in a Slack DM between the CPO and the CEO. The decision is supposed to reach every PM’s roadmap conversation, every PRD that gets written next week, every prioritization debate. Sometimes it does. Often it doesn’t. The decision lives in the CPO’s head and in some offsite slides. The PMs work from neither.
The second stream is feedback. It flows bottom-up. Customers say things on sales calls. AEs post in a Slack channel. Support tickets accumulate themes. NPS verbatims pile up in a Notion page nobody opens. The PMs hear about some of it — usually from another PM, or from the loudest AE — and write the PRD from memory. The roadmap that ships reflects last month’s loudest signal, not the actual texture of what customers are telling the company.
If you’re a CPO, here’s the thing that should make you nervous: those two failures aren’t separate problems. They’re the same problem from opposite directions.
Both of them are context that doesn’t travel. Alignment is context made at the top that doesn’t reach the bottom. Feedback is context made at the bottom that doesn’t reach the top. The reason your roadmap review feels like the same meeting every quarter is that the meeting is doing both jobs — re-distributing strategy downward and re-summarizing customer pain upward — by hand, every quarter, with nothing accumulating in between.
What an AI agent can and can’t do here
Before going further, a note for anyone whose product team hasn’t deployed agents in the workflow yet: an AI agent, in this article, is software that uses a language model to do real work for the team — drafting a PRD section, summarizing a sales call into roadmap implications, comparing this quarter’s customer feedback to last quarter’s. The agent is only useful if it can read the team’s actual context. The whole article is about what that context needs to include.
Most product orgs I’ve talked to are using agents the way marketing orgs were using them a year ago: per prompt. The PM opens Claude, pastes in the strategy doc, pastes in the relevant Gong call summaries, pastes in last quarter’s PRD for reference, and asks for a draft. They get something usable, sometimes. They spend twenty minutes pasting context every time. The agent never gets better at being useful for this team. Every PM does it slightly differently. Nothing compounds.
This is the same failure mode marketing went through. The model isn’t the problem. The model gets dramatically more useful the moment it has persistent read-access to the strategy decisions, the call transcripts, the past PRDs, the customer feedback themes — the actual context the team works from.
The two streams give us a useful way to think about what to wire first.
The alignment direction (top-down)
Every strategic decision the CPO makes is one of two things. Either it’s the kind of decision that affects how a PM writes their next PRD — we’re prioritizing the SMB tier over enterprise this quarter; the platform team is the constraint until we’ve shipped the rebuild; this feature isn’t on the table because legal flagged it — or it’s not. The first kind is the kind that needs to travel.
The current way it travels is: a slide at the offsite, a recap in Slack, a mention at the next 1:1, a re-mention at the roadmap review when it becomes clear that not everyone is on the same page. The slide ages out. The Slack thread sinks. The 1:1 mention reaches one PM at a time and doesn’t reach the Group PM filling in for them when they’re on parental leave.
A Director of Product Ops at a Series-D company told me last month that she started keeping a running doc of every strategic decision the CPO made — in 1:1s, at offsites, in the weekly. She did it because PMs kept asking her what they’d decided, and she didn’t have a clean place to point at. Six months in, that doc is the most-referenced document in the product org. The CPO didn’t know it existed for the first four months.
That doc is a substrate. It emerged because a human noticed the gap and filled it. The same substrate, set up deliberately and read by the team’s agents, is what alignment looks like when the context can travel.
The feedback direction (bottom-up)
This is the harder one for most product orgs, because the signals are noisier and the existing tools are partial.
You probably have a Gong account. You probably have a Slack channel where AEs post customer feedback. You probably have a Notion page with NPS verbatims. You probably have a support tool with ticket themes. You have all the inputs. The problem is that none of them reach the PRD when the PM is writing it.
A VP of Product at a dev tools company told me at a Pavilion breakfast in May: “My team has the Gong account, the Slack channel, the Notion page. None of it reaches the PRD. The PM writes the PRD from memory and from the last conversation they had — which is usually with another PM, not a customer.”
The fix isn’t more tools. It’s that the PM’s agent — the one drafting their PRD, the one summarizing the customer-call themes, the one comparing this quarter’s feedback to last quarter’s — can read from those existing tools as persistent context. Not as a one-off paste. As background the agent always has, every time. When that’s wired up, the PRD that comes back doesn’t read like one PM’s memory. It reads like the team’s collective ear for what customers have been saying.
Where I’ve seen this work, the change has been qualitative more than quantitative. PRDs start carrying language customers actually use. Edge cases that came up in three different calls last quarter start showing up in the requirements section. The PM doesn’t have to remember everything. The agent has read everything.
Why these are the same problem
You’ll have noticed that both directions need the same thing. The agent has to have persistent read-access to context that currently lives in places it can’t read. The alignment substrate is “decisions the CPO has made.” The feedback substrate is “things customers and the field have said.” Different content, same plumbing.
This is also why the right place to start is usually the feedback direction. Most product orgs have given up on automating the bottom-up flow — the assumption is that customer signal is too noisy and too contextual to automate, so it stays in Gong and Slack and Notion forever, and the PM does the synthesis manually every PRD. The wins from changing this are immediately visible: the next PRD carries language and texture it didn’t carry before, and the PM can feel the difference.
The alignment direction is also worth wiring, but it requires more curation discipline up front — the CPO or a Director of Product Ops has to actually keep the substrate current as decisions are made. The feedback direction can read from tools that are already capturing the data; it just needs the read-access. Start there. Notice what changes in the PRDs in three weeks.
The Group PM is the curator, not the bottleneck
If you’re a CPO reading this, the temptation is to centralize all of this in your seat. Don’t.
Your Group PMs and Directors of PM are the people closest to both streams. The Group PM running the platform cluster knows which strategic decisions affect platform PRDs and which don’t. The Director of PM running the growth cluster knows which customer-feedback themes are platform-relevant versus growth-specific. They are the curator-owners of their function’s slice of the substrate — the same way the CMO’s function heads curate their slice on the marketing side.
This is a gift, not a delegation. You stop being the alignment-explainer at every roadmap review, because the Group PM’s substrate already carries the strategy decisions the cluster needs. You stop being the feedback-summarizer, because the Director of PM’s substrate already carries the customer themes the cluster is hearing. The roadmap review changes shape: it becomes a conversation about the cases where the streams disagree, where the strategy points one way and customer feedback points another. Those are the conversations a CPO and her function heads should be having. Not the re-distribution of last quarter’s decisions.
What to try this month
One CPO I spoke with last month picked the feedback direction and started with a single input — Gong call transcripts piped into a place the product agents could read from. The first month was bumpy. The data team had questions about which transcripts to include. There was a permissions argument about which PMs could read which calls. There was a week where they thought it was hooked up and it wasn’t. By the end of the second month, the PRDs her team was drafting carried customer language she’d been waiting two years to see in PRDs. She doesn’t have a clean before/after metric yet — she said she’s not even sure how she’d measure it. But she said the next quarterly roadmap review is going to feel different, and she could feel that before she could prove it.
That’s the shape of the start. Not a transformation, not a clean win, not a dashboard. A single input wired into a place the team’s agents can read, and PRDs that start sounding like the team actually knows what its customers are saying.
Pomegranate, the piece I’m building, is one way to make that practical — it’s the substrate that carries your team’s voice and current direction across every agent in product, marketing, sales, and CS. There are other ways to set this up; the point isn’t the tool. The point is that until the alignment substrate and the feedback substrate are both wired, your roadmap review will keep feeling like the same meeting every quarter. The CPO with the oral-tradition joke didn’t need a new meeting template. She needed her team’s context to live somewhere it could be read from.
The ongoing cost is small but real: about five to ten minutes a day from each Group PM, on average, curating their slice of the substrate as new decisions land and new themes emerge. Hide that cost and the substrate decays; name it and it compounds.
You’ll feel the change first in the PRDs. The feedback you’d give, already in the draft. The roadmap review changes after that.