Midshift AI Process Platform
Building a productised, AI-powered process improvement platform from scratch — taking 20 years of BA and Lean Six Sigma methodology and making it accessible, automated, and affordable for organisations that have never been able to justify the consulting invoice.
A consulting methodology. Turned into a product.
Midshift is a Changeable-built SaaS platform that runs process improvement engagements end-to-end — from stakeholder interviews through to formatted deliverable documents — using Claude as the analytical engine and a structured nine-phase methodology as the governing framework.
The platform was designed to solve a specific market failure: SMEs, councils, and public sector organisations that need serious process improvement work but cannot justify a traditional consulting engagement that starts at NZ$30,000 and takes months to deliver.
Midshift makes the same frameworks Big 4 consultants use — Lean DOWNTIME, BABOK, BPM CBOK, and ADKAR — accessible through a guided, AI-powered platform that a non-technical process owner can use without training or facilitation.
Structured phases
Every engagement runs the same methodology. Intake, interviews, current state analysis, opportunity identification, future state design, automation assessment, and professional delivery.
Professional deliverables
Process Improvement Report, AI and Automation Assessment, Implementation Guide, and SOP — each a standalone document formatted for board or leadership presentation.
Human gates at critical points
AI does the heavy analytical work. Humans confirm before the platform proceeds. The methodology is designed so AI and human judgement operate in the right sequence.
Built as a real SaaS product
Full authentication, MFA, Stripe payments, tiered access, and an admin layer — not a prototype. A live platform at app.midshift.co.nz from day one.
Process improvement has a cost and access problem.
Every SME has broken processes. Most know it. Almost none have done anything structured about it — not because they don’t care, but because the options available to them are either too expensive, too slow, or too generic to be useful.
Traditional consulting engagements involve weeks of stakeholder interviews, workshop facilitation, analysis, and report writing. The methodology is sound but the delivery model is inaccessible for organisations without a large budget and a patient timeline.
Internal improvement attempts stall without a governing framework. Without a structured methodology, organisations loop back to symptoms without reaching root causes. The same problems resurface six months later.
The opportunity was to productise the methodology — not simplify it — and remove the access barrier without removing the rigour.
The design constraints that shaped the build
- The platform had to be usable by a non-technical process owner with no BA background
- The methodology had to remain intact — not watered down for ease of delivery
- AI-facilitated interviews had to feel like talking to a knowledgeable colleague, not filling in a form
- Human judgement had to remain in the loop at the moments where it matters most
- Deliverables had to be client-ready and board-presentable without post-processing
- The platform had to be buildable and maintainable by a solo non-developer founder
- Pricing had to be accessible to the SME and public sector market — not consulting-tier
- The build had to be live, functional, and sellable from the first engagement
How the platform was built
The build combined AI-assisted development, a disciplined no-code-first approach, and a methodology-first design philosophy — in that order.
Methodology productisation
Twenty years of BA and process consulting practice was translated into a structured nine-phase platform methodology. Each phase has defined inputs, outputs, and approval gates. The AI does not improvise the methodology — it executes it. Lean DOWNTIME, BABOK, BPM CBOK, and ADKAR are embedded in the analytical engine, not bolted on as labels.
AI-driven adaptive interviews
Stakeholder interviews are the core data-gathering mechanism of any process engagement. Rather than a scripted questionnaire, Midshift uses Claude to drive adaptive conversations — probing based on what stakeholders actually say, following unexpected threads, and ensuring all analytical objectives are met before the interview closes. No two interviews follow the same path.
Human-in-the-loop architecture
The platform is explicitly designed so that AI analysis and human approval operate in sequence, not in parallel. Critical transitions between phases require explicit human sign-off before the platform proceeds. This is not a usability feature — it is a methodological requirement. The human gate is where client context, professional judgement, and organisational knowledge enter the analysis.
Four-document deliverable architecture
The output layer was redesigned from a single Word report into four structurally distinct documents, each serving a different audience and purpose. The Process Improvement Report is diagnostic and board-ready. The AI and Automation Assessment is a Layer 2 analysis built on validated future state. The Implementation Guide bridges to action. The SOP documents the confirmed future state procedure.
AI-assisted solo build
The entire platform was built by a single non-developer founder using Lovable for app generation, Claude Code for targeted code fixes, and Claude.ai for architecture decisions and session planning. Every build session was logged to Notion. The build process itself became a live demonstration of how AI enables a subject-matter expert to build a professional SaaS product without a development team.
Production-grade from day one
Midshift launched as a fully authenticated platform — MFA, Stripe payments, tiered subscription access, email notifications, and an admin layer — not a prototype or an MVP with placeholder features. The platform was built to be sold from the first engagement, with the full methodology running end-to-end on launch day.
What Midshift delivers
A rigorous, AI-powered process improvement engagement — accessible to any organisation, not just those with a consulting budget.
Not months
A traditional process improvement engagement takes weeks. Midshift runs the same methodology — AI-facilitated interviews, structured analysis, professional deliverables — in hours of active work, with a 24–48 hour turnaround.
Not $30,000
The Pro tier runs five process engagements per month for NZ$150. The same methodology applied by a senior BA in a consulting context would cost tens of thousands per engagement. The platform changes who can access rigorous process improvement work.
Client-ready documents
Every engagement produces four distinct professional deliverables — each formatted, structured, and ready to present to leadership or a board. No manual assembly required after the platform has run.
Analytical dimensions
Each improvement opportunity is rated across eight dimensions — impact, time saving, error reduction, implementation effort, organisational readiness, dependency risk, strategic alignment, and quick win potential. Every score is accompanied by visible reasoning, not a black-box number.
Automation opportunity built in
Every future state step is assessed against an automation taxonomy — RPA, AI, workflow orchestration, system integration. Time reduction estimates, FTE saving projections, and tool recommendations are generated for every viable candidate, with explicit confidence levels.
A build others can learn from
The Midshift build is documented as live content for the Zero to AI podcast and education platform. Every architectural decision, every technical challenge, and every lesson learned is captured — making the build itself a practical resource for professionals learning AI-assisted development.
Productising expertise is harder than building features.
The technical build of Midshift was complex — authentication, payments, AI integration, document generation, email notifications. But none of that was the hard part. The hard part was deciding what the methodology should do, in what order, with what human involvement, and what constitutes a good output.
Most AI-powered tools make the same mistake: they use AI to do something faster without asking whether the thing being done is the right thing. Midshift was designed methodology-first. The AI executes the framework. The framework does not bend to what the AI finds easy.
The result is a platform where the quality ceiling is set by 20 years of consulting experience — not by what a language model will produce if left to its own devices.
What made this work
- Starting with the methodology, not the technology — the platform runs a consulting framework, it does not invent one
- Treating human-in-the-loop gates as a design requirement, not a limitation to work around
- Building for a real market gap — not a generic “AI for business” proposition
- Using AI-assisted development to enable a solo non-developer to ship a production-grade SaaS product
- Designing the deliverable architecture before writing a single line of application code
- Keeping the free tier genuinely useful — one full engagement, all nine phases, at no cost
- Logging every build session and decision so the build itself became reusable knowledge
- Testing and iterating on the AI interview engine with real conversations before committing to the architecture
Have a question about AI product design?
Common questions about building AI-powered products, productising consulting methodology, and what Changeable can help with.
What kind of processes can Midshift analyse?
Any repeatable business process where you have stakeholders who can describe how the work actually gets done. Invoice approval, staff onboarding, client intake, procurement, compliance workflows, service delivery processes — the methodology applies consistently across process types, industries, and organisation sizes. The AI interview engine adapts its questioning to the specific context provided at intake.
How is this different from just using Claude directly?
Using Claude directly gives you a capable AI. Midshift gives you a governed methodology executed by a capable AI. The platform structures the engagement, enforces the analytical sequence, manages stakeholder interviews, applies the Lean/BABOK/BPM CBOK/ADKAR frameworks consistently, and produces formatted client-ready documents. The methodology is the product. Claude is the engine.
Who is Midshift for?
SME owners and directors who know a process is broken but haven’t had the time or budget to address it. Operations and business managers who need a structured improvement plan to present to leadership. Councils and public sector organisations with efficiency mandates and limited internal capacity. Independent consultants who want to deliver higher-quality process work faster. The free tier is designed to be a genuine first engagement — not a teaser.
Can Changeable build something like this for another organisation?
Yes. The Midshift build demonstrates what is possible when domain expertise is combined with AI-assisted development and a disciplined product design process. Changeable works with organisations that want to productise their own methodologies, automate their consulting or service delivery processes, or build AI-powered tools around their proprietary knowledge. The starting point is always a Decision Clarity Session.
How long did the platform take to build?
The full platform — from first commit to a live, authenticated, payment-enabled SaaS with the complete nine-phase methodology running end-to-end — was built over several weeks of focused evening and weekend sessions by a single non-developer founder. The AI-assisted development approach compressed a build that would traditionally require a team of developers into something achievable solo, at a fraction of the time and cost.
Is the build documented anywhere?
The Midshift build is captured as live content on the Zero to AI podcast and education platform. Architecture decisions, technical challenges, lessons learned, and the practical experience of a non-developer building a production SaaS product with AI assistance — all documented in real time. The goal is to make this build a practical learning resource for mid-career professionals exploring AI-enabled product development.
What frameworks does the platform use?
The methodology is built on four established frameworks: Lean Thinking (DOWNTIME waste categories), BABOK (Business Analysis Body of Knowledge), BPM CBOK (Business Process Management Common Body of Knowledge), and ADKAR (change management and implementation readiness). These are embedded in the analytical engine — not referenced in passing — which means every opportunity assessment, every automation recommendation, and every implementation roadmap reflects the same rigorous analytical standard.
Start with a free Decision Clarity Session.
A Decision Clarity Session is a no-obligation conversation where we listen to where you are, what you are trying to achieve, and what is getting in the way. If you are thinking about productising a methodology, building an AI-powered service platform, or using AI-assisted development to bring a product idea to life — you will leave with a clearer view of what that actually takes.