How to Use It
Run it as one 90-minute session with your engineering, connected-product, and service leads — or as three short ones. Use the tools in this order: questions first, then the customer outcome, then the plan.
- Brainstorm for questions (20 min). Generate questions only. No answers allowed until the list is done. Pick the three that would change the plan most.
- Write the press release and FAQ (40 min). Date it December 2027. Describe what a customer's operations lead can do that they can't today. The FAQ is where the hard tradeoffs surface.
- Check the roadmap (30 min). Hold the draft plan against the checklist. Anything without an owner, a date, and a measure isn't in the plan yet.
Why written narratives. At Amazon, annual planning ran on written narratives in sentences, with no slides. A narrative makes it hard to skip what didn't work last year, and it makes the plan reviewable by people who weren't in the room. The same discipline works for a 200-person equipment maker.
What a Good 2027 Plan Has That Most Don't
- The tech debt behind last year's worst field-service months — named, costed, and owned.
- An explicit boundary for what software and agents may do to fielded units without a person.
- A Q1 that stabilizes before it automates, and automates before it modernizes.
- An AI line with a value test: Gartner advises that every AI investment show business value within 90 days or be cut.
// tool 1
1
Brainstorm for Questions (BFQ)
Better questions make better plans. Spend the first 20 minutes on questions only.
Rules
- Questions only. Anyone who starts answering gets redirected.
- Aim for 30+ questions in 15 minutes. Quantity first.
- Sort into three piles: answered already, answerable with data we have, nobody knows.
- Vote on the three questions that would change the plan most. Those become the plan's opening section.
Starter questions for a connected-equipment OEM
Fleet and field service
- Which three failure modes drove the most truck rolls last year?
- What share of dispatches could have been resolved remotely?
- Which hardware revision costs us the most to support?
- How long does it take to answer "which units, which firmware, which sites" after an advisory?
AI and automation
- Which actions would we let software take on a customer's unit without a person? Which never?
- What would an AI pilot have to show in 90 days to keep its funding?
- Whose data would we be training on, and who owns it?
Platform and tech debt
- Which part of the connected-product stack does only one person understand?
- What would break first if device count doubled?
- Can we roll back a firmware release on one cohort today?
People and money
- Which hire are we making because a process is broken?
- What are we paying for that nobody uses?
The moonshot question
If the installed base doubled by 2029 and service headcount stayed flat, what would have to be true of our platform?
// tool 2
2
Work Backwards: the PR-FAQ
Write the announcement before the roadmap. If the press release isn't exciting to a customer, the roadmap won't be either. (Format from Working Backwards, Bryar and Carr, 2021.)
The press release (one page, dated December 2027)
Headline
What a customer can now do, in their words.
Subheading
Who it's for and the main benefit, in one sentence.
The problem
What the customer's operations lead dealt with before. Use their numbers: downtime, truck rolls, recovery time.
The solution
What changed, described as the customer experiences it. No architecture.
Quote from your leader
Why you did it.
Quote from a customer
What it changed for them. If you can't imagine a believable one, stop here.
The FAQ
External (customers ask)
- What does it cost me, and what do I stop paying for?
- What happens to units already in the field?
- What can it do to my equipment without asking?
Internal (your leadership asks)
- What has to be stabilized first?
- What are we not doing so we can do this?
- What would make us stop?
- Who owns it, and how will we know by end of Q1?
// tool 3
3
2027 Roadmap Checklist
Every item needs an owner, a date, and a measure. Order matters: stabilize, then automate, then modernize.
Stabilize (Q1)
- The tech debt behind the worst field-service months is listed, with a cost and an owner for each item
- You can answer "which units, which firmware, which customer sites" within an hour
- Firmware releases go out by cohort, with a tested rollback
- Knowledge held by one person (provisioning, a legacy service) is written down
Automate (Q2–Q3)
- The first workflow is chosen for low consequence and high volume, such as alarm triage before dispatch
- An autonomy boundary is written and signed: what software may do alone, what waits for approval, what stays with a person
- Success is measured as truck rolls avoided without a rollback, not "% autonomous"
- Each AI investment has a 90-day value test and a stop rule
Modernize (Q3–Q4)
- Platform changes are tied to a customer outcome from the PR-FAQ
- Vendor contracts include an exit plan for data, logs, and policies
- The plan names what you'll revisit in January once Q4 actuals are in
// Want a second set of eyes on your 2027 plan?
Run it with us instead of just a worksheet.
The Roadmap Planning Workshop runs this exact kit live with your leadership team — plus the facilitation to get through the hard tradeoffs in one sitting.