Open recording · Passcode: 2R$x@Etv
Recorded 23 April 2026 · ~47 minutes · Luc van de Boom & Willem Spruijt
What this webinar covers
Why Scaler built Lumi, how it's hosted (AWS Bedrock, EU data residency, no training on client data), and how it's evolved from a support chatbot into an AI layer integrated across the platform. The second half is a live demo: home-screen chatbot, asset summaries, decarbonization roadmap with AI measure suggestions, portfolio-level data coverage analysis, and page-aware Lumi prompts. Closes with a Q&A on adoption, forecasting, and where to find Lumi.
Articles to read alongside the recording
- AI in Scaler: overview — what AI in Scaler does, where to find each feature, and how it keeps your data secure
- Using Lumi: writing effective prompts — write prompts that get accurate, useful answers
- AI for data quality and analysis — outlier detection, meter analysis, forecasts, asset list queries, and bill scraping
- AI for action and reporting — CapEx measure suggestions, AI report reviewer, and Investor DDQ answers
Key takeaways
- Lumi runs on AWS Bedrock with EU data residency for EU clients. Client data stays inside Scaler's infrastructure, isn't used for model training, and is scoped to the user's authorisation just like the rest of the platform.
- Lumi is more than a chatbot. It's an AI layer integrated across pages, with page-aware suggestions on collection, analytics, performance, roadmap, and reports.
- Every Scaler user gets 15 free Lumi prompts a month. Higher plans get unlimited. Use the 15 to practise, build a feel for prompting, and give feedback — even before licensing the full package.
- Lumi can forecast at month, year, and 10-year horizons. Forecast accuracy depends on how much historical data is in Scaler — clients with multiple years of consumption data see meaningfully better results.
- If Lumi isn't visible in your account, activation is usually a setting your Account Operations Manager can flip in minutes. In rare cases, your organisation's agreement with Scaler restricts AI use — check with your legal or admin team.
Chapter timestamps
Open chapter list — scrub to the timestamp in the recording
Zoom recording links don't deep-link to a specific time, so use these as a guide to scrub forward in the recording.
- 00:57 — Why we built Lumi
- 03:25 — Security: AWS Bedrock, scoped authorisation, EU data residency
- 05:52 — Accuracy: low temperature, observability, no training on client data
- 07:33 — How Lumi evolved: support chatbot → agent → proactive UI
- 14:25 — Demo: home screen chatbot and portfolio analysis
- 17:08 — Asset summary panel
- 19:30 — Decarbonization roadmap with AI measure suggestions
- 21:11 — Portfolio-level data coverage analysis
- 23:04 — AI data analysis: outliers and like-for-like
- 25:54 — Lumi as a data analyst on the collection portal
- 27:02 — Performance ranking with Lumi insights
- 32:23 — What's coming: PDF and audit document parsing
- 34:51 — Q&A: adoption advice for non-tech-first organisations
- 41:29 — Q&A: can Lumi forecast?
- 44:35 — Where to find Lumi (home screen and the bottom-right widget)
💡 This is a snapshot of what was covered on 23 April 2026. AI features in Scaler continue to evolve quickly — for current functionality always defer to the linked articles above.