Executive Presentation  ·  2026
AI Initiative
Wirk
Measured results from our own builds. A path to scale them across 160+ developers. The decisions that make it happen.
400h Project
400h → 50h 8x faster
Effort reduction on our own pilot
Dev Workflow
4h → 15min 16x faster
Maintenance screen build time
At Scale
160+ devs → multiplied
One standard, applied across the whole team
02  ·  Where We Are Today
The Problem We're Solving
Three friction points that slow delivery, fragment knowledge, and leave capability unevenly distributed across a 160+ developer organisation.
02 / 16
Uneven AI Adoption
160+ developers. Dozens of AI approaches.
Some developers use AI extensively. Others don't. Most are somewhere in between and experimenting without a standard. The productivity gap between optimised and unoptimised developers widens with every sprint. The multiplier exists. It's just not applied consistently.
8x faster delivery available now  ·  distributed unevenly across the team
Legacy Modernisation Bottleneck
Legacy analysis and migration is slow by default.
Understanding legacy codebases, mapping dependencies, and generating migration plans takes weeks manually. AI can compress this dramatically, but only with a standardised approach that produces consistent, verified output clients can trust.
Legacy modernisation  ·  weeks of analysis  ·  AI compresses to days
Support at Scale
180+ applications. 1,000+ devices. 24/7.
Support operations at this scale are complex. AI-driven triage, documentation generation, and pattern detection across incidents reduces resolution time and surfaces systemic issues before they escalate. Without it, engineers spend time on work that shouldn't require engineering judgment.
ITIL-certified support  ·  AI triage reduces resolution time
03  ·  What Changes
Why AI? And Why Now?
Three results from our own development work. Each measured. Each reproducible with a standardised AI workflow in place.
03 / 16
8x
Faster Delivery
400-hour project reduced to 50 hours. Same developer, same quality standard, standardised AI workflow applied end-to-end.
400h Project
16x
Build Speed
Maintenance and customisation screen builds from 4 hours to 15 minutes. Repeatable by any developer following the workflow. At Wirk scale, this is thousands of hours recovered per quarter.
Dev Workflow
1
Prompt
PM mockup turnaround: 3 days to one prompt. Produces a working, on-brand screen plus the use case and technical spec. ClickUp task created automatically.
PM + ClickUp
04  ·  How It Works
The Scaffold: What Makes It Work
The results on the previous slide are produced by a structured, tested, verified generation system. Not generic AI.
04 / 16
Architecture
Design Token Layer
tokens.css  ·  109 CSS vars (truth source)
8,294-line compiled CSS  ·  multi-tenant brand
Component Layer
7 Razor shell fragments  ·  18 atomic widget primitives
25 menu icons  ·  brand assets
Composition Layer
14 YAML recipes (A-M)  ·  design-system.yaml (Claude reads first)
prompt.md hard rules  ·  14 slot-substituted init templates
Quality Layer
verify.js live-vs-replica probe  ·  38 verified test screens
Returns tokens + shellRects + primitiveStyles
Claude reads the design system first, follows the composition rules, generates on-brand output, then verifies against a live replica. Generic AI gives you a rough idea. The scaffold gives you a production-ready screen.
The Pipeline: 1 Prompt to ClickUp Task
01 ///
PM writes one prompt
Describes the screen or feature needed
↓
02 ///
Scaffold generates on-brand screen
Reads recipes, applies tokens, verifies against live replica
↓
03 ///
Use Case Builder produces spec
Structured use case + technical spec for developers
↓
04 ///
ClickUp task created automatically
Work item tracked from brief to delivery. No manual handoff.
05  ·  The Developer Scaffold
How It Works for Developers
One /scaffold init deploys a project-aware AI infrastructure. Code-specific enforcement. Knowledge that persists.
05 / 16
Scaffold Architecture
scaffold/
├─ CLAUDE.md.tmpl • project identity + key commands
├─ Conventions • Things That Will Bite You
├─ hooks/ • 14 Python scripts, one concern each
guard_style.py • doc_check.py • memory_extract.py
└─ discovery/stack_markers.json • 8 stacks auto-detected
7 Commands  •  Any Project
/scaffold init · Bootstrap CLAUDE.md, docs/, hooks/, settings.json
/scaffold doctor · 26-validator health check. Flags drift + broken files
/scaffold upgrade · Pull latest templates without touching project files
/scaffold repair · Auto-fix broken scaffold state (doctor + fix failures)
/scaffold discover · Re-scan project: update overview + code fingerprint
/scaffold modules · List all modules with their status for this project
/scaffold tour · Walk through what each enabled module does here
Hooks are copied per-project. Self-contained and portable.
The Pipeline: 1 Session to Persistent Knowledge
01 ///
Developer opens any project
One-time /scaffold init deploys CLAUDE.md, hooks, and settings. Never repeated.
↓
02 ///
Session starts. CLAUDE.md auto-loads
Stack, conventions, gotchas, and key commands injected before the first prompt. No re-explaining the codebase.
↓
03 ///
Claude generates convention-aware code
kebab-case files, PascalCase classes, correct API shape, right branch naming. All from context, not instruction.
↓
04 ///
Session ends. Knowledge persists
Retrospective written. Corrections extracted as memory. Gotchas captured. Next session starts with everything the last session learned.
06  ·  Internal Pilots
Proven Results From Our Own Team
06 / 16
Development Results  ·  Scaffold Applied
87.5%
Effort Reduction
400-hour project — 50 hours delivered. Same developer, same quality standard. Applied at Wirk scale across 160+ developers, this multiplier is transformative.
94%
Build Time Saved
Maintenance screens: 4 hours to 15 minutes. Repeatable by any developer following the standardised workflow.
Our own team. Our own work. Not vendor projections.
PM Case Study  ·  Full Pipeline
From brief to tracked work item in one prompt.
A PM wrote one prompt. The scaffold generated a working, on-brand mockup. The use case builder produced the spec. ClickUp created the task automatically. 3 days of work compressed to minutes. The developer starts from a verified, tracked work item with no manual handoff.
BEFORE / AFTER
3 days → 1 prompt → ClickUp task
Mockup + use case + spec + work item. No manual step.
Proof 01
400h Project Efficiency
07 / 16
The Scenario
A development team quotes a large project: 400 hours estimated. With the AI scaffold applied end-to-end, initial UAT delivery came in at 50 hours.
Same developer, same quality standard. The scaffold handled spec generation, code scaffolding, test generation, and documentation at each stage. At Wirk's scale, this multiplier across 160+ developers translates to thousands of additional delivery hours per quarter without additional headcount.
What this means for Wirk
Every large project in the pipeline carries the same opportunity. More capacity per developer means faster client delivery, more projects in parallel, and margin improvement without scaling headcount.
Before
400h
MANUAL WORKFLOW
After (scaffold applied)
50h
AI SCAFFOLD APPLIED
The Result
87.5%
EFFORT REDUCED
Same developer. Same quality. 8x faster. UAT-ready output, not a prototype.
Proof 02
Legacy Modernisation Acceleration
08 / 16
The Scenario
A client has a legacy codebase: undocumented, tightly coupled, built over years. Understanding it, mapping dependencies, and producing a migration plan takes weeks manually.
AI-assisted legacy analysis reads the codebase, maps dependencies, identifies high-risk areas, and generates a structured migration plan with effort estimates. The analysis that took two weeks takes two days. The plan is documented, reviewable, and client-shareable from the first session.
Wirk's service differentiator
Legacy modernisation is already a Wirk service. AI-accelerated analysis makes this faster, more accurate, and more defensible to the client. It also produces the documentation that protects Wirk's team during the migration itself.
01 ///
AI codebase analysis
Reads legacy codebase, maps dependencies, identifies coupling and risk areas. Two weeks of manual analysis in two days.
02 ///
Migration plan generation
Structured migration plan with effort estimates, risk ratings, and recommended sequence. Client-shareable from day one.
03 ///
Documentation as byproduct
Every migration step documented automatically. Protects the team and the client throughout execution. No knowledge locked in one person.
04 ///
Faster scoping, faster delivery
More accurate scoping means more accurate quotes. Faster analysis means faster project starts. Both directly improve margin and client confidence.
Proof 03
PM Workflow & ClickUp Pipeline
09 / 16
The Scenario
A PM writes one prompt. Three minutes later: a working, on-brand, interactive mockup exists. The use case is built. The technical spec is written. A ClickUp task is created and tracked.
Without the scaffold, this PM was spending 3+ days per deliverable with inconsistent results. With it, the entire brief-to-tracked-work-item pipeline runs in one session. The developer starts from a verified, documented task. No manual handoff. No lost context.
The scaffold is the difference
The mockup generator reads the design system, follows composition recipes, applies tokens and components, then verifies against a live replica. The output isn't "an idea." It's a production-ready screen that passes verification. That's what makes the ClickUp task trustworthy enough to send to development immediately.
01 ///
PM writes one prompt
Describes the screen or feature. No wireframe, no spec document, no design brief needed upfront.
02 ///
Scaffold generates verified mockup
Reads design-system.yaml, follows recipe rules, applies correct tokens and components, verifies against live replica. Output is production-ready.
03 ///
Use Case Builder produces spec
Structured use case + full technical spec for developers. Same format every time. No variation, no gaps.
04 ///
ClickUp task created automatically
Work item created with spec attached and context preserved. Development starts from a tracked, verified work item. No manual handoff required.
Proof 04
Support Operations at Scale
10 / 16
The Scenario
Wirk manages 180+ applications and 1,000+ devices across manufacturing and dealer networks, 24/7. Incident triage, root cause analysis, and documentation are constant workloads.
AI-assisted support operations triage incoming incidents, suggest probable root causes based on historical patterns, generate structured incident reports automatically, and flag recurring issues that indicate systemic problems. Engineers spend time resolving issues, not administering them.
The ITIL advantage
Wirk's ITIL-certified three-tier support model is already structured for AI augmentation. AI triage sits at the first tier, routing incidents with context before a human engineer touches them. Resolution time drops. Pattern detection improves. The same team covers more ground.
01 ///
AI incident triage
Incoming incidents classified, prioritised, and routed with suggested probable causes. Engineers start with context, not a blank ticket.
02 ///
Pattern detection
AI surfaces recurring incidents that indicate systemic issues. Problems flagged before they escalate to major incidents or client-visible failures.
03 ///
Automated documentation
Incident reports generated automatically from resolution notes. Compliance documentation maintained without manual effort. Audit trail always current.
04 ///
Same team, more coverage
AI handles first-pass triage and documentation. Engineers apply judgment where it's needed. No headcount increase required to expand coverage.
11  ·  The Direction
Cloud-First Now. Local AI Next.
11 / 16
Phase 1  ·  Active Now
Cloud AI: fastest path to results
Cloud AI delivers capability right now, without infrastructure investment before governance and skills are in place. For Wirk, the immediate opportunity is standardising cloud AI workflows across 160+ developers to capture the 8x and 16x multipliers at team scale.
Why the timing matters
The AI space is maturing, not just growing. Tooling, orchestration patterns, and collective understanding are converging. Each maturity cycle builds on the last. Teams with structured workflows capture each improvement as it lands. The gap widens with every wave.
Phase 2  ·  Industrial AI
Local AI for manufacturing clients
Manufacturing and industrial clients (Volkswagen, Aspen, Tenneco) operate in environments where data sovereignty and latency matter. Local AI deployment on client infrastructure becomes a differentiated service offering once foundations are in place.
Data stays on client infrastructure. No external exposure.
Shopfloor AI without cloud dependency or latency
Natural progression from cloud foundations being built now
In Pipeline  ·  Proof of Direction
Gemma 4 E4b
1
Active Client
First Local AI Deployment
Document classifier POC, approved and in active development. Local inference, no external data exposure.
12  ·  The Initiative
Our AI Initiative: What We're Building
12 / 16
01 ///
Education
Structured AI literacy for every developer: tools, prompting, and responsible use. Built on the failure modes already documented so teams skip the learning curve.
02 ///
Standardisation
Defined AI workflows and proven methods. Codified processes that replicate the 15-minute builds and 50-hour projects across every team, every client, every engagement.
03 ///
Policy & Governance
Clear policies on what tools to use, how data is handled, and what outputs require validation. Especially important for manufacturing and industrial clients with strict data requirements.
04 ///
Measurement
Tracking time saved, quality improvements, and ROI across teams and client engagements. Makes the business case visible and reportable at every leadership review.
05 ///
Team & Resourcing
One person owns standardisation, iteration, and feedback. That ownership model, applied across Wirk's teams and divisions, is how this scales beyond individual effort.
06 ///
Executive Alignment
Three asks: public leadership commitment, approval for the short-term productivity buffer during upskilling, and policy sign-off. These are what move this from individual effort to company capability.
13  ·  Honest Disclosure
Understanding the Adoption Curve
13 / 16
There is a learning dip. We are planning for it.
When developers begin adopting AI, initial productivity drops 20-30% as they map the learning curve. This is expected and temporary. The real budget cost is billability, not cash. The scaffold shortens that dip significantly, because the failure modes are already documented. Teams that follow it recover faster and stabilise higher.
+5× +3× +2× +1× 0 -1× -3× -5× +3x early win (2–5× range) −3x dip (matches early gain) +2x stable (2–3× range) Weeks 1–4 Month 2–3 Month 4–5 Month 6+ BASELINE
Why the field is accelerating
LLM inference costs fell 40x per year (Epoch AI). The same AI capability costs 280x less than two years ago (Stanford AI Index 2025). Every month of delay means competitors access better tools at lower cost.
01 ///
The Dip Is Temporary
Productivity drops 20-30% as developers adjust. Those who persist (especially with scaffolded guidance) recover and exceed their original output.
02 ///
The Scaffold Shortens It
The failure modes are already mapped. The scaffold gives developers a structured path through the learning curve, not trial and error.
03 ///
Stagger the Rollout
One team at a time. At Wirk's scale, this keeps client delivery stable while the standard is refined and proven before broader rollout.
14  ·  The Ask
What We're Asking For
14 / 16
01
Budget Approval
The real budget is billable time, not direct cash. People's billability dips temporarily as they upskill. Approval means accepting a short-term billability buffer while teams clear the adoption curve. The payback begins immediately and compounds with every project.
02
Executive Sponsorship
Visible leadership endorsement to signal this is a company priority, not a team experiment. At 160+ developers, a top-down signal is what creates the conditions for consistent adoption.
03
Dedicated Headcount
One person with clear ownership to drive standards, training, and continuous improvement. This moves the initiative from person-dependent to institution-dependent across all of Wirk's teams and client engagements.
04
Policy & Governance Sign-off
Executive approval of an AI usage policy, formalising what is encouraged, permitted, and required. Essential for Wirk's manufacturing and industrial clients who operate under strict data and compliance requirements.
15  ·  Q&A Preparation
Addressing Your Concerns
15 / 16
"Our clients need consistency and precision. AI introduces risk."
The scaffold is precisely about removing that risk. It doesn't replace discipline. It encodes it. Every output passes a live-replica verification before it's considered done. Manufacturing clients get the same quality standard they always have, delivered faster and with documented context.
"If this is working, why isn't it spreading already?"
Results don't spread on their own. They spread when someone structures the evidence clearly enough that a decision becomes easy. You're experiencing that right now. The initiative formalises that process so it isn't dependent on one person doing it. That's how individual results become company results.
The Data Is Clear  ·  The Results Are Real  ·  The Risk of Delay Is Measurable
We move from pilots
to practice together.
Four decisions. One direction. Secure our competitive position.
→Approve the budget to scale our AI initiative
→Provide executive sponsorship to signal company-wide commitment
→Authorise dedicated headcount to own and drive AI standards
→Sign off on the AI governance policy to protect the business and our clients
WIRK  ·  AI INITIATIVE  ·  2026