Proof-of-concept solutions
Real education challenges translated into clear, consultancy-style solution designs.
Stephanie Clifton · The AI Elective
Designing practical AI adoption for education.
I explore how Microsoft AI technologies could help schools and trusts reduce workload, improve decision-making and build sustainable AI capability. The AI Elective documents my consultancy-style proof-of-concept solutions, adoption frameworks and learning journey.
Proof-of-concept solution designResponsible adoption thinkingHonest learning in public
What you will find here
A portfolio of thinking, design and practical exploration
The work is deliberately transparent about its stage: these are independently developed prototypes and implementation concepts, not claims of live organisational deployment.
Microsoft architecture
Thoughtful use of Copilot, Copilot Studio, Power Platform and connected services.
Adoption frameworks
Discovery, prioritisation, governance, piloting, measurement and scaling.
Learning in public
Honest reflections on decisions, mistakes, certifications and evolving practice.
Powered by Microsoft
An integrated AI and automation architecture
Familiar Microsoft services working together as one secure, governed ecosystem rather than a disconnected collection of tools.
Productivity foundation
Microsoft 365
DataverseAI and agent layer
Microsoft Copilot
Copilot Studio
Azure AI FoundryAction, apps and insight
Power Automate
Power Apps
Power BI
Microsoft FabricSecurity & compliance
Enterprise controls, privacy and governance.
Human oversight
People remain accountable for important decisions.
Responsible AI
Transparent, inclusive and proportionate by design.
Measurable impact
Every concept is linked to clearly defined expected outcomes.
Designed around outcomes
AI that earns its place
Time Saved
Creating organisational capacity through intelligent automation of repetitive, low-value work.
Measurable Value
Every concept begins with a business problem and defines the expected outcomes, implementation approach and measures that would be used to evaluate impact.
Responsible AI
Built with governance, transparency, data protection and meaningful human oversight from the outset.
From the Staffroom
A learning journal in progress
The latest reflections from the journey into Microsoft's AI ecosystem. This section updates automatically whenever a new Staffroom entry is added.
Enter the Staffroom
Musings from the learning journey.
Education AI Solution Library
Practical concepts for every part of school and trust life
A collection of independently developed Microsoft Power Platform concepts organised by operational need. Every solution is presented at the same level, with transparent status, governance, architecture and projected impact.
New to AI adoption?
Start here.
Explore the library by the scale and complexity of the change.
Proof of ConceptEarly-stage solution concept↗
Central Services
Marketing & Brand Management Agent
Create consistent school communications while protecting brand quality and governance.
1Brief2Create3Review4Publish
Why I built this
Trust marketing teams need to balance local voice, speed and consistency. I built this to explore how schools could create content more confidently while central teams retain clear brand and publication oversight.
Microsoft CopilotPower AutomateCommon questions
Does AI publish content automatically?
No. Drafts are routed to named school and central reviewers, and publication remains a human decision.
Is a human still involved?
Yes. Local context, accessibility, brand approval and final publication all remain under human oversight.
What Microsoft tools does this use?
Microsoft Copilot, Power Automate, Microsoft 365 and an approved content or asset store such as SharePoint.
How long could a prototype take?
An initial controlled workflow could typically be explored in 4–8 weeks, depending on existing brand assets and approval routes.
What licences might be required?
This depends on the organisation’s Microsoft estate and chosen connectors. Licensing would be confirmed during discovery.
Working PrototypeFunctioning prototype available↗
School Admin
AI Enquiries Email Agent
Reduce repetitive enquiries while keeping every response under human oversight.
1Receive2Classify3Draft4Approve
Why I built this
School teams lose valuable time answering recurring questions. I built this to test whether AI could reduce repetitive drafting while ensuring a person remains responsible for every response.
Copilot StudioPower AutomateCommon questions
Does AI make decisions automatically?
No. It classifies and drafts, but it does not make consequential decisions or send responses autonomously.
Is a human still involved?
Yes. Every draft is held for review and approval by a named colleague before sending.
What Microsoft tools does this use?
Copilot Studio, Power Automate, Microsoft 365 and, where appropriate, Dataverse or SharePoint for approved knowledge.
How long could a prototype take?
A focused prototype using a limited set of approved enquiries could be explored in 2–4 weeks.
What licences might be required?
Licensing depends on message volumes, connectors and the existing Microsoft environment and would be confirmed during discovery.
Working PrototypeFunctioning prototype available↗
School Admin
Admissions & Open Evening Assistant
Provide instant answers to prospective families while reducing administrative workload.
1Ask2Retrieve3Answer4Escalate
Why I built this
Prospective families often need the same information outside normal office hours. I built this to make approved admissions guidance easier to access without removing personalised support for complex questions.
Copilot StudioMicrosoft 365Common questions
Does AI make admissions decisions?
No. It answers routine information questions only and escalates individual, sensitive or judgement-based enquiries.
Is a human still involved?
Yes. Staff own the source content, escalation routes and responses to individual cases.
What Microsoft tools does this use?
Copilot Studio, Microsoft 365 and Power Automate, with approved website and document content as knowledge sources.
How long could a prototype take?
A bounded assistant for open-evening and routine admissions questions could be explored in 2–4 weeks.
What licences might be required?
Requirements vary by channel, user access and knowledge-source design and would be assessed during discovery.
Proof of ConceptEarly-stage solution concept↗
School Admin
School Trip Agent
Simplify trip administration by bringing information, documents and communications together.
1Propose2Prepare3Track4Approve
Why I built this
Trip administration is spread across forms, emails, payments and checks. I built this to explore how one coordinated workflow could reduce chasing and make readiness easier to see.
Power AutomateDataverseCommon questions
Does AI approve school trips?
No. The educational visits lead and other authorised staff retain all approval and risk decisions.
Is a human still involved?
Yes. Staff review medical, safeguarding, payment and consent information and make the final readiness judgement.
What Microsoft tools does this use?
Power Automate, Dataverse and Microsoft 365, with Power Apps potentially used for a structured trip interface.
How long could a prototype take?
A prototype for one trip type or process stage could typically be explored in 4–8 weeks.
What licences might be required?
Licensing depends on Dataverse, Power Apps and connector choices and would be confirmed against the organisation’s estate.
Proof of ConceptEarly-stage solution concept↗
Leadership
Executive Briefing Agent
Turn meetings, reports and key updates into concise leadership briefings in minutes.
1Collect2Summarise3Review4Decide
Why I built this
Leaders receive more information than they can reasonably absorb. I built this to explore how approved sources could be turned into concise, traceable briefings without replacing leadership judgement.
Microsoft CopilotPower BICommon questions
Does AI make leadership decisions?
No. It organises and summarises approved information; leaders remain accountable for interpretation and decisions.
Is a human still involved?
Yes. A leader reviews sources, checks emphasis and approves the final briefing.
What Microsoft tools does this use?
Microsoft Copilot, Microsoft 365, Power BI and Power Automate, depending on the source systems.
How long could a prototype take?
A briefing based on a small number of consistent sources could be explored in 4–8 weeks.
What licences might be required?
This depends on Copilot access, Power BI capacity and data-source connectors and would be confirmed in discovery.
Proof of ConceptEarly-stage solution concept↗
Leadership
Meeting Action Tracker
Capture decisions automatically and keep actions visible, owned and on track.
1Record2Extract3Assign4Track
Why I built this
Important actions can disappear into minutes and inboxes. I built this to make ownership, deadlines and progress visible after the meeting ends.
Power AutomateMicrosoft 365Common questions
Does AI assign actions automatically?
No. It proposes actions from approved records, but the meeting owner confirms wording, ownership and deadlines.
Is a human still involved?
Yes. The chair or meeting owner approves actions before they are created or assigned.
What Microsoft tools does this use?
Power Automate, Microsoft 365, Dataverse and optionally Power BI for progress reporting.
How long could a prototype take?
A prototype for one recurring meeting cycle could be explored in 2–4 weeks.
What licences might be required?
Requirements depend on where minutes, tasks and reporting are stored and would be confirmed during discovery.
Proof of ConceptEarly-stage solution concept↗
Teaching & Learning
Teaching & Learning Assistant
Give teachers faster access to trusted guidance, resources and design principles, rooted in your organisation’s resources.
1Brief2Retrieve3Refine4Approve
Why I built this
Teachers need quick access to trusted organisational guidance without losing professional autonomy. I built this to create useful starting points grounded in an organisation’s own resources and design principles.
Microsoft CopilotCopilot StudioCommon questions
Does AI replace teacher planning?
No. It provides editable starting points grounded in approved resources; teachers retain professional control.
Is a human still involved?
Yes. Teachers review, adapt and approve every resource before classroom use.
What Microsoft tools does this use?
Microsoft Copilot, Copilot Studio and Microsoft 365, connected to approved curriculum and teaching resources.
How long could a prototype take?
A prototype focused on one subject, phase or resource set could be explored in 4–8 weeks.
What licences might be required?
Licensing depends on Copilot access, knowledge sources and user numbers and would be confirmed during discovery.
Proof of ConceptEarly-stage solution concept↗
Pastoral & Safeguarding
Attendance Analysis Assistant
Identify attendance patterns earlier so staff can intervene with greater confidence and impact.
1Import2Analyse3Prioritise4Review
Why I built this
Attendance information can show risk before it becomes obvious, but patterns are easy to miss. I built this to surface meaningful signals earlier while keeping pastoral interpretation with staff.
Power BIPower AutomateCommon questions
Does AI decide which pupils need intervention?
No. It surfaces patterns and priorities; attendance and pastoral professionals interpret context and decide what action is appropriate.
Is a human still involved?
Yes. Staff validate data, consider individual circumstances and own every intervention decision.
What Microsoft tools does this use?
Power BI, Power Automate, Dataverse and Microsoft 365, connected only to appropriately governed attendance data.
How long could a prototype take?
Because data quality, thresholds and governance must be validated, a controlled prototype and pilot may take 8–16+ weeks.
What licences might be required?
Requirements depend on data architecture, Power BI capacity and secure connector access and would be confirmed during discovery.
Proof of ConceptEarly-stage solution concept↗
SEND & Student Support
EHCP Assistant
Update staff and help them navigate EHCP documentation more efficiently while supporting consistent practice.
1Secure2Extract3Review4Support
Why I built this
EHCPs contain vital information but can be difficult for busy staff to navigate consistently. I built this to help authorised colleagues find relevant provision more quickly while preserving source context and professional judgement.
Copilot StudioDataverseCommon questions
Does AI interpret or change an EHCP?
No. It helps authorised staff navigate information and links outputs back to source wording; statutory interpretation remains professional.
Is a human still involved?
Yes. SENCOs and authorised colleagues review context, correct extraction and decide how information should inform practice.
What Microsoft tools does this use?
Copilot Studio, Dataverse, Power Automate and Microsoft 365 within a tightly controlled access model.
How long could a prototype take?
Secure document handling, permissions and quality assurance make this a strategic 8–16+ week exploration.
What licences might be required?
Licensing depends on secure storage, Dataverse capacity, connectors and Copilot Studio use and would be confirmed during discovery.
No matching solutions found.
My AI adoption approach
A repeatable path from problem to responsible scale
This is the process I would use to turn an opportunity into a governed pilot and, where the evidence supports it, a sustainable implementation.
Discover
Understand users, workflows and friction.
Prioritise
Balance value, readiness and risk.
Design
Shape the human-led solution and architecture.
Govern
Set controls, ownership and escalation.
Pilot
Test safely with a defined user group.
Measure
Evaluate quality, value and experience.
Scale
Expand only where evidence supports it.
Why I built The AI Elective
I chose to design the work before I had permission to deploy it.
I created The AI Elective to give myself a dedicated space to explore what effective AI adoption could look like across education.
Using realistic educational scenarios and consultancy-style methodologies, I've developed a portfolio of proof-of-concept solutions that demonstrate how Microsoft AI technologies could support schools and trusts. Each concept explores discovery, process redesign, solution architecture, governance and implementation planning, while remaining transparent about its current stage of development.
This portfolio reflects how I approach AI adoption: understanding the problem first, designing with people and process in mind, considering responsible governance from the outset, and focusing on measurable outcomes. While these concepts have not yet been deployed in a live organisation, they represent the way I think about solving real-world challenges and the type of work I hope to deliver at scale.
Start with the workflow
Understand the people, friction, risks and desired outcome before selecting a tool.
Keep humans accountable
AI supports professional judgement; it does not replace safeguarding, governance or decision ownership.
Measure the difference
Define pilot hypotheses around time, quality, capacity, consistency, experience and organisational value.
The AI Elective