Engineering Diagnostic

Rebuilding an unmanaged delivery model, and recovering ~43% of annual delivery cost

~43%: of annual delivery cost recoverable, drawn from the client's own data
~60% higher: true all-in delivery cost versus the visible spend
~9%: of monthly cost banked from the first fix

A noisy field of scattered points on the left passes through a violet diagnostic scan band; one pink-ringed signal node is isolated on the right, labeled the constraint.

Case file

Industry: Early-stage AI SaaS
Region: Switzerland
Service: Engineering Diagnostic
Team: Wayne ArendseDelivery & Transformation LeadYurii KotulaCEO & Engineering Leader
Systems: Product platform
a Teams bot and AI modules
CI/CD
single-tenant architecture
enterprise frame of SOC 2 Type II / ISO 27001 and Swiss data residency

Client context

The client is an early-stage Swiss software company. It is pivoting from a consulting-services business into an enterprise AI product, and has committed to enterprise-grade security: Swiss-default data residency, encryption, least-privilege access, and no training on customer data.

The client is moving decisively into that market which screens vendors on multi-tenancy, security certification, and delivery predictability. With its enterprise commitments set, the question worth answering was how ready delivery already was to back them at scale. The engagement put that in numbers, drawn from the client's own delivery data rather than assumption.

Business problem

Delivery ran on individual contributors with no delivery owner. One role was expected to cover leadership, architecture, code review, meetings, and hands-on development at the same time. Coordination, prioritisation, and technical trade-offs were handled informally rather than through a delivery and technical governance structure.

The texture showed in the day-to-day. Around half of all merge requests shipped with no peer review, and releases went out ungated, at will. Kanban ran without a formalised cadence and there was no committed release plan. In one session, 27,000 lines of AI-generated code entered the codebase, unreviewable at that scale. Three client-side stakeholders coordinated sign-off by hand, and client executives spent roughly 240 hours a month keeping delivery moving.

The stakes

Total modelled exposure across six categories approached a full year of delivery cost.
Roughly 39% of all-in delivery cost was high-confidence avoidable waste, drawn from the client's own delivery data.
Client-side executives spent ~240 hours/month keeping delivery moving instead of running the business.
32.5% of delivered work was rework, re-done at full delivery cost.

What was broken

One role covered leadership, architecture, code review, meetings, and hands-on development at once.
~50% of merge requests shipped with no peer review, and releases shipped ungated.
27,000 lines of AI-generated code entered the codebase in a single session, unreviewed.
The product ran on a single hardcoded tenant.
The Teams bot and the AI modules had never been reviewed, on either side.
No committed release plan existed; post-release defect rate and sprint velocity were not tracked.

Diagnostic approach

The gap sat across the business at once, a strategic pivot, an operating model, and a codebase, so the diagnostic ran on three levels together.

Strategy
What is the delivery gap worth?
The pivot to an enterprise product depends on delivery credibility the current model cannot yet show, so the diagnostic put the gap in money before anything else. It mapped the consulting-to-product pivot to the delivery capabilities enterprise buyers screen for, established the true all-in cost of delivery rather than the visible team spend alone, and separated confirmed waste from modelled exposure.
Operations
Where does the time actually go?
Delivery ran on individuals with no cadence and no owner, so the coordination cost landed on client executives. It traced ~240 client-executive hours/month spent keeping delivery moving, measured rework at 32.5% of delivered work, and found Kanban run without a formalised cadence and no committed release plan.
Technology
Can the system ship safely?
Code shipped without review or gates, and the architecture could not onboard a second enterprise customer. It found ~50% of merge requests shipping with no peer review, a single hardcoded tenant blocking multi-tenant onboarding, and the Teams bot and AI modules unreviewed, with no stable DEV/Demo environment.

What changed

The true cost of delivery is now a figure, not an impression.
The engagement replaced a visible monthly spend with an all-in cost roughly 60% higher, built from the client's own data and a Swiss IT/PM benchmark. Of that, around 39% is confirmed avoidable waste. The decision now rests on numbers.
A three-role operating model, proposed with ownership drawn.
The proposed model splits today's single overloaded role into a Client Executive Sponsor for sign-off, a Technical PM for operations, and a Tech Lead for architecture and quality. Each carries a defined access level and decision boundary. It is designed to remove the single point of failure the diagnostic found.
A measurement baseline that already banked one saving.
Six delivery metrics now carry baselines and targets, tracked sprint over sprint. The first fix the measurement surfaced, a redundant pre-daily removed and the team daily trimmed to 20 minutes, saves around 9% of monthly delivery cost, at zero incremental cost.

Decisions

Confirmed waste leads, not the biggest number.
The case could have opened on the full modelled exposure. It opens instead on the ~39% of cost that comes from the client's own data, so the figure carrying the decision is the one that cannot be argued down.
Split the overloaded role rather than hire a CTO.
A full-time CTO was costed at a six-figure annual salary and rejected. It over-corrects on cost and stage, takes 60–90+ days to hire, and still does not run daily delivery. Three defined roles fix the actual failure, which is ownership, not seniority.
Fix the operating model before multi-tenancy or certification.
SOC 2, ISO 27001, and a multi-tenant rebuild are real blockers, but scoping them first spends money before delivery can be trusted to execute. The path gates them behind a working cadence and full review coverage.
Prove the measurement before asking for the full engagement.
The meeting saving was banked from a retrospective, not projected. It shows the tracking works before any larger commitment is made.

Results

Measure	Before	After
Delivery cost visibility	Delivery seen only as the visible monthly spend	True all-in cost established ~60% higher, ~39% of it avoidable waste
Meeting overhead	Redundant pre-daily plus an unbounded team daily	~9% of monthly cost recovered, realized
Code-review coverage	~50% of merge requests reviewed	Target 100%, tracked sprint over sprint
Release planning	No committed release plan	Published release plan, met per release, under the proposed model

Strike Team

Senior engineering leaders find the real constraint in a fixed-fee engineering diagnostic, then stay to fix it. Project rescue, fractional CTO, architecture and business cases.

About the Strike Team