Meeting assistance
Meeting assistance supports the entire meeting workflow by automatically transcribing conversations, generating summaries, and identifying action items.
It ensures that decisions and to-dos are clearly captured and easily shareable. Additionally, it provides analytics on participation, speaking time, and meeting quality.
As we operate within the Belgian public sector, meetings often involve a dynamic mix of Dutch/French (and English) spoken interchangeably. It is therefore essential that AI transcription tools fully support multilingual conversations. Not all tools on the market currently handle this complexity adequately, which can lead to inaccurate transcriptions and lost context. Multilingual support is not a nice-to-have — it’s a critical requirement for us.
Maturity levels
|
Level |
Name |
Description |
Technology |
Example tools |
| 0 | Non-existent | Meetings are fully manual: no recording/transcription, no structured summaries, no automated follow-up. Notes and actions are captured ad hoc outside project tools. | none | none |
| 1 | Ad-hoc assist | AI is used sporadically after the meeting to transcribe or summarize recordings. Output is inconsistent, not linked to task systems; multilingual/code-switching largely unsupported. |
|
|
| 2 | Embedded assist | AI is built into the meeting platform: reliable live transcription, first-pass summary, and extracted action items. Multilingual is basic (tolerates occasional NL/FR switching); artifacts are consistently saved and shareable. |
|
|
| 3 | AI-Human collaboration | Assistant co-creates in real time: proposes decisions, confirms owners & due dates, tracks parking-lot items, and adapts from feedback. Multilingual is robust with per-utterance detection; actions/decisions are pushed to project tools. |
|
|
| 4 | Full autonomy | Assistant orchestrates end-to-end: prepares agenda, facilitates/time-boxes, records decisions/risks, assigns actions, monitors closure, and escalates. Multilingual is native-grade (code-switch aware) with policy-driven outputs. |
|
|
| AI Maturity Level: Indicates the level the technology vendors claim to have reached in deploying AI solutions that actually work in real-world applications |
Effort/cost estimation
Effort and cost estimation begins with sizing the project – translating requirements into quantifiable units like function points or use cases. AI enhances this process by interpreting unstructured inputs, applying historical patterns, and suggesting effort and cost ranges with confidence levels.
Maturity levels
|
Level |
Name |
Description |
Technology |
Example tools |
| 0 | Non-existent | Estimates are manual (spreadsheets, intuition) with no standardized sizing or baselines. | none | none |
| 1 | Ad-hoc assist | LLMs are used sporadically to guess size or effort from prompts; outputs are unstructured and not tied to a sizing method or history. |
|
|
| 2 | Embedded assist |
AI is built into the estimation workflow: it interprets stories/use cases into a chosen sizing method and proposes ranges with explicit assumptions. |
|
|
| 3 | AI-Human collaboration | Estimator and AI co-simulate scenarios: apply cost drivers (reuse, team skill, focus/continuity, NFRs (Non-Functional Requirements)), generate uncertainty ranges (P50/P80), and derive schedule/staffing with time–cost trade-offs. |
|
|
| 4 | Full autonomy | Agentic system ingests RFPs/specs/repos, produces end-to-end estimates with self-tuning drivers, updates as data evolves, and outputs team plans/risks/financials. |
|
|
| AI Maturity Level: Indicates the level the technology vendors claim to have reached in deploying AI solutions that actually work in real-world applications |
Management product generation
Management product generation involves creating structured project documents (called Management products in PRINCE2).
AI supports this by drafting, summarizing, and refining content based on project context, previous documentation, and live inputs.
It accelerates document creation, ensures consistency, and reduces manual effort while maintaining traceability and alignment with project standards.
P.S.: Knowledge reuse from meeting assets: meeting recordings/transcripts (cf supra: Meeting assistance) can be reused as organizational knowledge assets to auto-draft PM documents (management products)
Maturity levels
|
Level |
Name |
Description |
Technology |
Example tools |
| 0 | Non-existent |
Project management documents are produced entirely by hand from past examples; no AI support. Knowledge reuse from meeting assets: no capture; nothing indexed or reusable |
none | none |
| 1 | Ad-hoc assist |
LLMs are used sporadically to draft paragraphs or ideas via manual prompts; outputs are unstructured and not tied to project context or templates. Knowledge reuse from meeting assets: manual copy/paste from single recordings or rough transcripts into documents; no search, no citations. |
|
|
| 2 | Embedded assist |
AI is built into the authoring workflow: fills template sections from structured inputs (project metadata, objectives), keeps style consistent, and proposes first drafts for standard sections. Knowledge reuse from meeting assets: transcripts/recordings are indexed (per project) with basic metadata (meeting type, date, participants). PM document templates get snippet suggestions with timestamp links back to the recording/recap; light de-duplication. |
|
|
| 3 | AI-Human collaboration |
The assistant co-creates with the PM: proposes entire documents/sections with rationale, tracks open inputs, reconciles versions, and learns from feedback. Multilingual output and house-style enforcement are consistent. Knowledge reuse from meeting assets: context-aware generation across multiple meetings. The assistant extracts structured fields and auto-fills sections of the PM documents with evidence citations (timestamped quotes). Reviewer accepts/edits; changes improve future drafts. Multilingual normalization applied to official outputs. |
|
|
| 4 | Full autonomy |
Agentic system assembles and maintains PM documents end-to-end: pulls facts from project tools, generates/updates products on events, tracks assumptions, flags risks/changes, and routes for approval. Knowledge reuse from meeting assets: agentic assembly of full PM products using organization-wide recordings and prior project assets. Cross-project pattern mining (e.g., recurring risks and mitigations by domain). Change-aware updates: when new meetings land, the agent proposes redlines to the PM documents and triggers approvals. Policy-aware governance (PII redaction, retention, residency) enforced end-to-end. |
|
|
| AI Maturity Level: Indicates the level the technology vendors claim to have reached in deploying AI solutions that actually work in real-world applications |