AI that reads your delivery data, not a chat box bolted on

AI analysis that supports portfolio ranking decisions: proposing scores against defined criteria from the underlying business case and delivery data, explaining why one investment outranks another, and flagging investments whose declared scores are inconsistent with the evidence in their records.

Scoring a portfolio of fifty investments against eight criteria is slow, and the scores are frequently supplied by the people who submitted the investments.

AI that proposes candidate portfolio scenarios against a stated constraint — a budget reduction, a capacity ceiling, a strategic reweighting — and explains the trade-off each scenario makes, feeding the deterministic scenario engine rather than replacing it.

The hard part of scenario planning is not computing the impact; the engine does that.

AI that reads incoming demand, proposes classification and routing, extracts business-case fields from attached documents, and detects that a new request duplicates or overlaps one already in the portfolio.

Duplicate demand is endemic in large organisations — the same need arrives three times from three departments in three vocabularies.

Forecasting of future resource demand from portfolio pipeline, historical delivery patterns and seasonality, producing a probable demand curve by role that extends beyond the explicitly planned horizon.

Explicit capacity planning covers what has been planned.

Detection of dependency risk before it materialises: predecessors trending toward a slip that has not yet been reported, undeclared dependencies inferred from resource and deliverable overlap, and cascade exposure where one task's slip would propagate widely.

Cross-project dependency failure is the commonest cause of programme delay, and the damage is done in the gap between a predecessor beginning to slip and anyone declaring it.

AI that classifies incoming RAID items, proposes probability and impact scores against the tenant's matrix, detects that the same underlying risk appears across multiple projects, and identifies registers going stale.

RAID registers decay.

Automated assessment of schedule health against recognised quality checks — missing logic, negative lag, constraint abuse, excessive float, dangling activities, unrealistic durations, resource-free tasks — with plain-language explanation of why each finding matters and how to fix it.

Most schedules in most organisations are structurally unsound, and the problem is invisible until the schedule is asked to do something.

Duration and effort estimates for new work proposed from actuals on comparable completed work, with the comparable set shown, the variance range stated, and the basis of comparison explained.

Estimates are habitually optimistic and habitually undocumented.

Plain-language explanation of why earned value indices moved — attributing a CPI or SPI change to the specific activities, cost lines and events that caused it, and describing what the trend implies if it continues.

EVM is objective and widely misread.

Extraction of structured work items — actions, decisions, risks, issues, dependencies — from unstructured sources such as meeting notes, minutes, emails and uploaded documents, proposed for confirmation into the appropriate register.

Governance information is created in meetings and lost in documents.