Version 6.3.5 answers a question that gets asked at every sprint planning meeting: what should we commit to next sprint? The headline is Next Sprint Planning — a capacity-aware, Monte-Carlo-backed forecast that's now visible to every customer under Plan → Next Sprint Planning. This release also tightens Cycle Time so the chart admits when the data is unreliable rather than reporting a confident-looking floor, and brings three AI Metrics features that came directly out of customer feedback.
New: Next Sprint Planning
Plan → Next Sprint Planning produces a single screen that reads your team's last six sprints and the upcoming sprint's Azure DevOps Boards capacity, and turns both into a recommendation. No extra configuration — the data is already there, the new view just makes it readable. A typical recommendation card looks like this: "Planned Work: 4 recommended · P50 7 · Safe — · 7.0 day median across 164 closed items." The number on the left is what the model thinks the team can confidently commit to once capacity, history, and reliability are folded in. The line on the right shows where that number came from.
Five things land on the screen at once:
- Capacity-adjusted commitment forecast per bucket — Monte Carlo over the team's own throughput, scaled by the upcoming sprint's effective capacity, with PTO and team days off already included.
- Cycle-time reality check — median and P85 days per bucket, with explicit signalling when the data is too thin or too tight to trust.
- Backlog risk scan — for items already pulled into the sprint, flags missing estimates, blocked work, and items with no parent feature or epic.
- Commitment reliability lens — the historical fraction of committed work the team actually completes, applied to the recommendation so the number is honest rather than aspirational.
- Capacity card — effective percentage and per-member breakdown, read straight from Azure DevOps Boards capacity.
Predictions are advisory — that footer stays in the view because the goal is informed planning, not autopilot. The recommendation anchors the conversation in the team's own data; it doesn't replace the planning meeting.
Cycle Time now admits when the data is unreliable
Next Sprint Planning leans on cycle time, and cycle time has a well-known failure mode: same-second start and close timestamps. They show up everywhere — freshly-imported backlogs, teams whose workflow auto-closes items at creation, projects that bulk-update statuses through the API. Until 6.3.5, the chart would report a confident-looking 0.1-day floor for those buckets, and any view downstream would treat that as real.
6.3.5 changes the contract. When the start and close timestamps collapse to the same day across a bucket, the chart now says "⚠ same-day timestamps — cycle math unreliable." When fewer than three items make up a bucket, "⚠ thin sample" replaces the otherwise-confident "✓ fits in sprint" claim. Customers running Next Sprint Planning on early-stage data see honest signals rather than bogus precision.
The trade-off is intentional. A view that sometimes admits it doesn't know is more useful than one that always sounds confident — it's the same posture as the "Predictions are advisory" line on Next Sprint Planning. Tools that fake precision get trusted once and ignored after.
AI Metrics: three customer-feedback features
AI Metrics picks up three additions in this release. They came out of feedback rounds with customers using AI Metrics in production, and they're now native to Agile Analytics — no separate install.
- Inline vs Chat / Agent breakdown card on the Overview tab — stacked bars across generated outputs, accepted outputs, and lines added, split between IDE inline completions and chat / agent workflows. Tells you where your AI value is actually coming from.
- Side-by-side Team Comparison card on the Teams tab — pick up to five teams via chip selectors and compare them on six metrics at once (members, acceptance rate, suggestions, usage intensity, net LOC, and low-adoption count) as horizontal bars. No more flipping between team pages to compare by hand.
- Link-shareable sub-tabs — the AI Metrics view now uses hash-routed deep links (#/ai-metrics/teams, #/ai-metrics/users, and so on) that survive refresh and back/forward. The "let me send my exec a link to the Actions tab" request is now possible to honour.
Updating
The extension auto-updates from the Marketplace, so existing customers don't need to do anything — the in-app What's New dialog will surface the new view on first load after the upgrade and link straight into Plan → Next Sprint Planning. New installs can grab v6.3.5 from the Marketplace listing at https://marketplace.visualstudio.com/items?itemName=Baytek.agile-analytics.