AMBER overall

GO-LIVE

Programme data
Claude · reasoning
Slack
Gmail
Azure DevOps

Claude reads the current programme data, writes a concise status update, and posts it straight to your #programme-status Slack channel.

Programme data · embedded
Anthropic API · Claude writes
Slack · #programme-status

AI Programme Assistant

Ask anything — Claude reasons across the programme data in real time.

Predictive Risk Engine

Risks the AI surfaced that are not yet in the register — found in the white space between workstreams.

AI Inbox Triage

Claude reads your unread email, summarises each one, suggests what to do, and drafts replies — you create the draft in Gmail with one click.

Gmail · read + draft
Anthropic API · Claude triages

Delivery Analytics — Azure DevOps

Starts at programme level. The engine reads your live work items, works out how your board is actually organised — however many levels deep it goes — and computes the delivery view in code: velocity trend, story & build completion, open P1 defects and ADO hygiene. Open any workstream to see the same figures for the work beneath it. Claude narrates the figures it is handed; it is never given the work items.

Azure DevOps · Boards (metrics computed in code)
Anthropic API · Claude narrates

Populates Azure DevOps with ~38 work items — user stories, bugs, tasks and builds — across four markets (APAC · LATAM · NA · EU) as Area Paths, over seven sprints (five completed · current · future) with story points, so the per-market analytics above are real. Re-seeding? Clear first to avoid duplicates.

Area Paths · 4 markets
7 sprints + story points

Go-live readiness

Pick a market — cutover is a per-market event. A weighted go-live confidence is computed from that market's live ADO defects, test progress, risk posture and milestone health, with a decisive go / no-go call. Rollback is shown as readiness states, not a fabricated probability.

AI proposes the shape. Code computes the dates.

Describe a programme in your own words and the AI drafts its shape — workstreams, phases, how long each takes and what waits for what. Every date is then worked back from your go-live by the scheduler, never supplied by the model. When a go-live moves, replanning recomputes every date and shows you each change, and why, before anything is saved.

Shape · proposed by AI, reviewed by you
Dates · computed in code

A paragraph, a goal and a date, or just a list of deliverables. Uses your organisation's own Anthropic key.

Shutdowns and holidays (optional)

One range per line, e.g. Year-end shutdown: 2026-12-21 to 2027-01-01. A single date is a single day. Used only to flag dates that land in a closed period — no date is ever moved for it, and nothing here is saved.

Use the plan template, or your own plan in your own format. You will see exactly what was read — and can correct any column — before anything is saved.

Pick a programme and the date it now has to land on. You will see every date that would move, by how much, and the dependency that fixed it. Nothing changes until you apply.

Shutdowns and holidays (optional)

One range per line, e.g. Year-end shutdown: 2026-12-21 to 2027-01-01. A single date is a single day. Used only to flag dates that land in a closed period — no date is ever moved for it, and nothing here is saved.

Resource Utilisation & Capacity

Two questions in one view: are people correctly loaded — who is booked beyond their hours, whose capacity is sitting idle — and is the work left achievable with the people assigned, at the rate each workstream is actually going. Every figure is computed in code from the resourcing and delivery records; the AI writes the narrative only. Demo runs on embedded fictional data shaped like a real resourcing export.

Resourcing export · demo snapshot
Delivery velocity · demo snapshot
Metrics · computed in code

Each workstream's own velocity against the work remaining and the sprints left. Needed is the rate the remaining sprints demand; where that is above actual velocity, the workstream cannot land on its current dates without more people or less scope.

Hours booked against hours available this sprint. Red is booked beyond capacity; amber is capacity nobody is using.

Where a role is over-subscribed in one workstream and idle in another, that is a reallocation rather than a hire.

Financial Command Centre

One view across both finance sources: period spend (planned vs actual, CapEx/OpEx — the ServiceNow lens) and the market money picture (budget, committed, actuals, balance — the leadership workbook lens). Every figure on this page is computed in code from the underlying records; the AI writes the narrative only. Demo runs on embedded fictional data shaped exactly like both production sources.

ServiceNow financials · demo snapshot
Budget workbook · demo snapshot
Metrics · computed in code

Solid orange = actuals to date · dotted orange = run-rate forecast · navy = phased plan · vertical line = current period.

Filled bars = actual spend split CapEx (dark) / OpEx (light) · outlined bars = remaining plan.

Actuals to date across all markets.

Click a market row to drill in. RAG is computed: red = run-rate forecast exceeds budget · amber = runway or contract-coverage flag.

Closed loop with ServiceNow

Projects live in ServiceNow — the PMO's system of record. The app reads them live over OAuth, and publishes status back: it computes the five RAG categories from real programme data, Claude writes the narrative, it writes that onto the record, and pings the PM on Slack. Authenticated exactly as we'd connect to a production instance — registered OAuth app, no stored password in the calls.

ServiceNow · OAuth 2.0
Slack · PM notify
Anthropic · narrative

AI Reporting Automation

Pick a deck type, select markets, choose extras — generate a branded slide deck per market or as a full programme pack. Download or save to Gmail.

Weekly Status

~10 slides · team & stakeholders

Operational detail: RAG, workstreams, RAID, milestones.

Steering Committee

~5 slides · executives

Decisions, exceptions, risks above threshold.

Leadership Update

~5 slides · leadership

Outcomes, value delivered, roadmap.

Each selected market generates its own set of slides — Project Overview (with its milestones), Weekly Summary and workstream detail. Markets are never combined on a slide. De-select any market not relevant for this audience.

Programme data · embedded
Azure DevOps · live
Download · .pptx
Gmail · draft

Configure your programme

Point at your own plan, map your columns once, and set how reporting works — no code, no developer. The dashboard adapts to your configuration.

Have a project schedule? Import it on Plan & Replan. This section is for a sheet with one row per workstream or RAID item.

Every plan names its columns differently. Instead of rebuilding the tool for each one, you map your columns to the fields needs — once, on this screen. then reads your live plan through that mapping, and everything downstream (RAG, risk sizing, decks) keeps working unchanged.

ⓘ This defines and saves the mapping, and previews it live on real sample rows. Reading a live Smartsheet through it (API token + storage) is the final wiring step — the mapping you build here is exactly what that connection uses.

Data source

1Upload your plan

⬆

Drop your plan here, or browse

Excel (.xlsx) or CSV — from Smartsheet, Monday, MS Project, Google Sheets, anywhere

No file to hand? Load a sample sheet · or paste column headers

How your organisation divides its year into periods. then shows a programme's money in the periods your finance team uses — calendar months, 4-4-5 weeks, 13 four-week periods, or your own period table.

The currency your organisation reports in. Each programme keeps the currency its money arrived in; converts it at rates the European Central Bank publishes or your finance team enters — never at a rate an AI produced.

One of three presets for your whole organisation, chosen by your admin. What a role may not see is removed by the server before it reaches anyone's screen, and a file carrying money its sender may not change is saved without that money.

Whether a closed period locks, the lists your money is broken down by, and whether each programme's cost includes your own staff.

Programme name

Domain

Programme owner

Phase model

Programme structure

Single project

One project, one delivery team

Multi-market rollout

Each market is a separate project

Portfolio of programmes

Multiple independent programmes

Select which tools are active for this programme. The AI reads from and writes to each connected tool automatically.

Plan management

Execution tracking

Status destination

Communication

AI provider

Audiences

Health scale

Publish schedule