Overview & my contribution
Talking SCADA began in Innovation in Practice (University of Skövde) as an AI-enabled service to help people understand hidden energy problems in buildings. I refined the same concept through product, requirements and business coursework at Blekinge Institute of Technology. It explores multi-agent decision support centred on human judgement; it is not a launched product.
The work builds on my industrial domain grounding and applied AI studies. Across the coursework I produced an innovation report, productification and commercialisation plans, a product protocol, a PRD, an SRD, a business plan, and value and sourcing analyses — connecting user needs, technical choices, delivery and a viable offer to see how they fit together.
Problem & users
Why
Buildings use around 40% of the EU's energy. Heating, ventilation, cooling, sensors and schedules influence each other, so a deviation can come from a changed setpoint, a faulty sensor, equipment running unnecessarily or heating and cooling working against each other.
The challenge is understanding what the data means and where to investigate.
Who it is for — users & customers
Users: building operators, facility and energy managers and service teams with varying HVAC/BMS expertise. Customers: property owners, facility-management companies, multi-site enterprises and municipalities. System integrators and development teams were also considered.
Pain
Investigations mean navigating trends, alarms and settings across several systems and interpreting relationships that are not directly visible.
Proposed job to be done
When my building uses more energy than expected, help me understand what changed, why, and where to investigate — so I can act without manually analysing every system.
Concept & proposed user journey
What the concept is
In everyday terms, a way for the building to explain itself: instead of reading trend curves and alarm lists, the person responsible asks a question and gets a plain-language answer about where energy is wasted, the likely cause and what to check next.
Technically, an explanation layer for SCADA/BMS combining time-series data with a semantic model of building systems. Users can inspect the sensors, alarms and relationships behind each answer before deciding what to do next. It does not control physical systems.
Proposed user journey
- Energy deviation
- What changed?
- Cross-system context
- Possible causes
- Supporting evidence
- Where should I investigate?
Product decisions & MVP scope
NPD & product definition
- Opportunity: market pull, technology push and underused operational data, framed in a Product Innovation Charter (PIC) covering background, focus and direction.
- Concept evaluation: originality, usefulness and producibility, with technical and commercial Go/No-Go criteria to decide before committing further resources.
- Product definition: product protocol (market, whole offer, strategic fit, risks) and a PRD with scenarios, requirements, MVP scope and acceptance criteria, with parallel product, marketing and financial work and iterative feedback.
Requirements & traceability
The SRD separates domain, functional, data, quality and regulatory requirements, defines the system boundary, and links elicitation findings to goals and requirement IDs. The layer stays read-only; decisions stay with people.
Prioritisation
I ranked 22 requirements and used Cost-Value-Risk on the critical ones. Explanation generation and anomaly detection ranked highest — core value and highest uncertainty. Data access and system context are enablers; explainability, usability and security were prioritised from the start.
MVP definition
One SCADA integration, selected HVAC subsystems and a web-based conversational interface explaining energy anomalies.
Success measures
Intended measures: investigation time, energy waste and operational insight, with acceptance criteria and quality targets for explainability, reliability and security. None have been measured yet.
Why AI and multi-agent?
AI makes operational data and specialist context easier to interpret. Validation should show whether multi-agent gives enough benefit over simpler retrieval-based support to justify the complexity.
Value proposition & business assumptions
- Lower energy costs and emissions through earlier detection of hidden waste.
- Specialist knowledge made accessible — less dependence on individual experts.
- Clear, evidence-linked explanations and faster, better-informed decisions.
- Support for energy-performance goals and reporting.
Intended gains, not measured outcomes.
Positioning & go-to-market
PESTEL and VRIO framed external drivers and capabilities. Positioning centres on cross-system diagnostics and understandable explanations, introduced through early adopters and system-integrator partners with onboarding and ongoing service.
Business model & value assessment
A Business Model Canvas connects customer segments and value propositions with channels, partners, costs and revenue. The proposed offer combines decision support with integration, onboarding and ongoing service; the business plan proposes tiered SaaS packaging from building-level explanations to portfolio diagnostics.
Software Value Map weighed usability and explainability against development cost and technical debt — estimates within the academic case.
Assets & Make/Buy/Share
Build in-house where it differentiates — domain graph reasoning, root-cause analysis, traceable explanations — and reuse AI components and external infrastructure elsewhere. Pricing and gains remain subject to market validation.
Commercialisation & life cycle
Parallel technical and marketing launch preparation, staged rollout, performance monitoring and feedback loops, with pricing and features evolving across the product life cycle.
Evidence & validation approach
Documented academic work
- Innovation report — problem framing, PESTEL, VRIO, Design Thinking.
- Productification and commercialisation plans — PIC, concept evaluation, market tests, lifecycle.
- Product protocol and PRD — offer, scenarios, MVP scope, success measures, acceptance criteria.
- SRD — stakeholders, elicitation, traceability, Cost-Value-Risk; context diagram, data model and mockups.
- Business Model Canvas, Software Value Map, business plan and Make/Buy/Share analysis.
Documented elicitation
Interviews, observation and a simple chat-based prototype showed that investigations required combining multiple views and inferring system relationships. This observation shaped requirements for graph context and referenced explanations — so users could see the evidence behind each answer.
Next validation step
Run one narrow workflow with representative users and real building data, compared with existing tools and simpler AI. Measure investigation time, explanation quality, customer value, operating cost and willingness to pay. Energy and cost benefits remain to be measured.
How the case evolved
From Innovation in Practice at Skövde, the concept continued through my Industrial Economics and Management specialisation at Blekinge Institute of Technology (Aug 2025 – May 2026):
- Product Management: NPD, concept evaluation, product protocol, PRD, launch and life cycle.
- Product & Requirements Management: SRD, traceability, prioritisation, Software Value Map and Make/Buy/Share.
- Strategy & Business Models: business plan with positioning, packaging, revenue logic and market entry.
Methods & domains
- PESTEL
- VRIO
- Design Thinking
- NPD
- PIC
- Product Vision
- Product Discovery
- Jobs to Be Done
- PRD & SRD
- Requirements Traceability
- MVP Definition
- Acceptance Criteria
- Cost-Value-Risk
- Software Value Map
- Business Model Canvas
- Make/Buy/Share
- Go-to-Market
- Product Life Cycle
- Multi-agent AI
- SCADA
- BMS
Reflection & next steps
Carrying one concept through innovation, product, requirements and business work showed how early framing shapes later scope, priorities and commercial assumptions — and that useful innovation depends on user understanding and cross-functional collaboration.
Images
Academic product case based on general industry and domain knowledge.