Q2 AI Roadmap Alignment
What this meeting produced
"We now know what ships in Q2 and what we're explicitly not doing — and that's the part the team usually leaves the room arguing about."
Meeting highlights
Fastest decision
"Move the on-prem deployment to a Q3 follow-up" — landed in 4 minutes once Anna shared the cost model.
Most impactful contribution
Priya's framing of "two evaluation tracks: regression + safety" reshaped the rest of the agenda.
Most engaged discussion
Pricing model for the pilot — 22 minutes, all six participants spoke at least twice.
Decision under pressure
The "ship behind a flag" outcome was reached with two minutes left in the slot — group chose moderate confidence over delaying.
Personal contributions
Session feedback
Summary
What we set out to do
Align on the Q2 AI roadmap with a shared answer to three questions: which two pilots ship, how we evaluate them, and what we explicitly defer to Q3.
Where we landed
- Two pilots are committed for Q2: the support-triage assistant and the internal knowledge search. Both ship behind a feature flag with a 4-week soak period.
- Evaluation runs on two parallel tracks — regression against last quarter's gold set, and safety via the new red-team rubric Priya introduced.
- On-prem deployment is deferred to Q3. The cost model Anna shared made the trade-off concrete: ~6 weeks of platform work, with no committed pilot customer to anchor it.
Open tensions worth naming
- Pricing for the support-triage pilot is contested. Per-seat vs. per-resolution split the room 4–2 — the group chose per-seat for the pilot, with a 60-day review.
- Two participants flagged that the safety evaluation rubric is still drafty. Priya owns hardening it before the first pilot release.
Energy & engagement
Most engaged stretch was the pricing discussion (22 minutes, full participation). Lowest energy was around the deployment topic — partly because the conclusion ("defer") felt foregone once the cost model landed.
Decisions 5
Action items 9
Parked items 6
Time analysis
Suggested follow-up meetings 3
Based on parked items, contested decisions, and action item dependencies.
The per-seat pricing was the only contested decision — revisit it after 60 days of pilot data instead of letting it drift.
- Pilot revenue + usage data
- Per-seat vs. per-resolution — revisit
- Decision: keep, adjust, or switch model
Two participants flagged the rubric as drafty; the rest of the roadmap depends on it. Tight loop, small group.
- Walk through rubric v1
- Edge cases + scoring guidance
- Sign-off or next iteration
Reopen the deferred on-prem decision once Anna's refreshed cost model and Lukas's customer pipeline are in.
- Updated cost model
- Pilot customer pipeline
- Go / no-go for Q3 scoping
Want this for your workshops?
WorkshopBuddy runs as a private engagement with NeuroForge. We’ll set up a live walkthrough on a real session of yours.