Markides Advisory

Use CaseSurrogate AI Lead Sprint

CustomerMid-Size Structural
Engineering Firm (Anonymized)

Selected examples of how we partner across growth, operations, and AI.

About these examples. These examples are illustrative. Some are drawn from prior work, some combine more than one engagement, and some were created to demonstrate how the work runs. Identifying details have been removed or changed, and some of this work predates Markides Advisory. What is real is the structure, the reasoning, and the format of the deliverables. Any figures shown are directional rather than audited.

02Current situation

A well-run firm where the constraint is senior engineering time.

Growth has tracked the regional construction market rather than outpacing it, and leadership is explicit that no aggressive expansion is planned.

The business

A structural engineering firm serving commercial and multifamily developers, roughly 26–30 full-time staff. Service lines: structural design, permit drawings, and construction administration support, delivered across a steady pipeline of concurrent projects rather than a few marquee jobs.

Systems and data

Projects run through a combined project-management and time-tracking system; engineering work happens in a standard CAD environment; KPI reporting is finished in Excel from a monthly export. The project system is treated internally as a trusted single source of truth for both project and financial data.

Ownership already formalized

The founder is the principal engineer and delegates most day-to-day operations to an operations director, a role created roughly three years earlier specifically to formalize ownership of process, reporting, and resourcing. Changes are reviewed and approved in a quarterly leadership meeting, with a monthly project financial review and a weekly resourcing meeting.

The binding constraint

Engineering talent, not cash, not tool budget, limits how much work the firm can take on. Cash is healthy and tool tolerance is moderate, but leadership has been explicit that it does not want the firm to become a testing ground for tools without a clear business case.

Markides Advisory - Use CaseCurrent situation

03The trigger

A permit resubmission that made the pattern impossible to file away.

The incident

A scheduling scare on a multifamily project for a repeat developer client

A permit resubmission, caused by inconsistent use of the firm's own QA checklist, pushed the project close enough to a hard deadline that the operations director had to personally intervene to keep the client relationship undamaged. It was not the first resubmission of its kind, it was the one that made the underlying pattern impossible to treat as a one-off.

The resource behind the pattern

Senior engineers spending a disproportionate share of their time on drawing QA

The firm's scarcest and least replaceable resource, given how difficult engineering talent is to hire in this market, was absorbed by review work rather than the technical judgement only they could do. Every project the firm added made that bottleneck slightly worse.

The founder's non-negotiable

Open to AI and automation to relieve pressure on senior staff, on one clear condition: nothing would be tested without a named owner and a defined success measure in place first. The firm's reputation rests on technical accuracy, and the founder had seen enough industry chatter about scattered, ungoverned AI pilots at peer firms to want no part of that pattern.

Why an outside-led sprint

The operations director had already been quietly researching options relevant to permit tracking and QA support, without running a formal pilot. Rather than picking a tool and testing it informally, they proposed a short sprint, a structured, evidence-based view of where to start before spending any of the firm's moderate tool budget or asking senior engineers to change how they worked.

Markides Advisory - Use CaseCurrent situation

04Mandate and guardrails

The task was precision, not repair.

Surrogate AI Lead Sprint, run as a founder-plus-operator intake: a founder interview plus operations-director input. The operations director sponsored; the founder retained sign-off through the existing quarterly leadership review.

Objective

The fastest, cheapest, directionally reliable view of what to do first in AI, automation, and analytics, targeted specifically at relieving pressure on senior engineering time and closing well-evidenced operational gaps, not at fixing governance that was already sound.

The firm had nearly every strong-fit signal for this sprint type at once: clear existing ownership, mature and trusted data, low tool overlap, and a founder genuinely open to AI so long as it was governed. What it lacked was a structured way to decide which of several plausible ideas, permit tracking, QA review support, budget-risk reporting, deserved its limited appetite for structured testing first.

Explicit boundaries

Boundary 01

No changes to the core CAD or project management systems.

Boundary 02

Every test must have a named owner and a defined success measure before it starts.

Boundary 03

No more structured tests running at once than the firm's existing reporting cadence and ownership model can absorb, roughly two at a time.

Markides Advisory - Use CaseEvaluation

05Evidence base

Rich evidence, and one gap that mattered less than usual.

Collected

Founder interview with the principal engineer, growth intent, resourcing constraints, and guardrails for any new testing.
Operations-director interview, ownership, reporting mechanics, workflow pain points, and prior research into automation options.
Project financial export, active projects, margins, and utilization by role.
Existing KPI reporting templates, including the monthly Excel roll-up the operations director already produces.

Intentionally not collected

No junior-engineer-level workflow interviews were conducted.

The gap was logged explicitly rather than assumed away, though it mattered less here than in other engagements, since the operations director's existing process ownership already gave strong visibility into how drafting and QA handoffs actually work in practice.

Evidence thickness

Rich

Strong, internally consistent visibility into operations, resourcing, and financial reporting.

Overall confidence band

The highest band the sprint uses

Reflecting both the quality of the data reviewed and the firm's already-mature ownership structure.

Where visibility is thinner

Sales and marketing behaviour

The firm's business is mostly repeat and referral work, with little active selling to observe.

Markides Advisory - Use CaseEvaluation

06Evaluation

Seven move families, and most of them were already fine.

Each family was evaluated independently against severity, leverage, effort, information readiness, and confidence, then compared across families for overlaps, magnifiers, conflicts, and sequencing dependencies.

Move family	What the evidence showed	Read
Leadership and ownership	In strong shape before the sprint began. The operations director explicitly owns AI, analytics, and automation decisions with founder sign-off; activity is coordinated through an approved tool list and an established quarterly leadership review. The one open gap: no formal AI use policy written down yet.	Severity low · confidence high
Workflow redesign	12–15 active projects at a time, 4–6 new proposals per month. Drawing QA review stood out sharply: heavily dependent on senior engineer availability, estimated at 15–20% of senior engineer time. Permit submission tracking a second clear pain, tied directly to resubmissions from inconsistent checklist use. Fee proposals judged likely to stay manual.	Severity high · leverage high · effort low-medium
Reporting and KPIs	Meaningful KPIs already tracked, project margin, utilization by role, proposal win rate, on-time permit submission rate, through a reasonably mature monthly Excel dashboard taking about half a day to assemble. Definitions standardized, so cross-team inconsistency is minimal. The gap: no early warning for projects trending over budget before month-end close.	Severity medium · leverage medium-high · effort low
AI-assisted sales and marketing	Little weight in this sprint. New business is repeat- and referral-driven, proposals built from an already-efficient standard template, and throughput is explicitly not a current constraint.	No move considered; none needed
Data plumbing	The project management system is a genuinely trusted single source of truth for both project and financial data, and most duplication between systems has already been removed. One mechanical break: the manual monthly export into the Excel dashboard.	Severity low-medium · confidence high
Tools and experiments	No formal AI or automation experiments yet, though the operations director has been actively researching options. Tool overlap minimal by design; appetite for structured testing strong given mature ownership and reporting discipline, roughly two structured tests absorbable at once.	Capacity: ~two concurrent tests
Risk and guardrails	One guardrail governed everything: any test must have a named owner and a defined success measure before it starts. No changes to core CAD or project systems. The founder's broader concern was scattered experimentation and vendor noise, not AI itself.	Binding on every move

Markides Advisory - Use CaseEvaluation

07The choice

Why three moves, not the default two.

What the synthesis ruled in

Workflow redesign and reporting both pointed at the same underlying strength: data and ownership already mature enough that new moves could be additive and low-risk rather than corrective. Nothing here required fixing governance first. So the highest-confidence leverage sat in relieving the firm's real constraint, senior engineering time, and closing two well-evidenced, low-risk gaps the firm's own reporting had already surfaced.

What was explicitly set aside

Fee-proposal automation. Considered and ruled out: the evidence suggests proposals should stay manual, since one senior person's judgement is central to producing a credible one. This is a process the firm does not actually have a gap in.

The co-equal test, applied

The operating system defaults to two primary moves, reserving three only where the evidence shows genuinely co-equal, tier-one options rather than a primary-plus-secondary pattern. This engagement met that higher bar.

Permit tracking and the QA bottleneck were clearly tier-one on leverage-over-effort grounds from the start.
The budget-risk early-warning view showed comparably high leverage against a low effort cost, with evidence as strong as the other two.
No clean way to rank it third without understating a real, already-quantified gap, and no clear fourth candidate behind it.

Forcing a false primary-versus-secondary ranking would have distorted the firm's real decision picture, not clarified it.

Markides Advisory - Use CaseChoice

08Recommended moves

Three co-equal moves, each with an owner and a success measure.

Move	What it is	Why now	Profile	Owner
01 - Automate permit submission tracking	Replace manual permit-status tracking with a structured, semi-automated tracker tied to the existing project management system, with automated status flags.	The strongest automation candidate the sprint identified, tied directly to the resubmission errors already occurring, the same pattern behind the trigger incident.	Leverage high · effort low-medium · value medium Scope 2–3 weeks · traction 1–2 project cycles · confidence high	Operations director, consistent with existing ownership of process and reporting. Highly startable internally.
02 - Reduce the senior-engineer QA bottleneck	A standardized QA checklist plus light AI-assisted first-pass markup review, so senior engineers review flagged items rather than starting from scratch.	QA review consumes 15–20% of senior engineer time, the firm's single scarcest resource, given its stated engineering-talent constraint.	Leverage high · effort medium · value medium-high Scope 3–4 weeks · traction 2–3 project cycles · confidence high	Operations director sponsors; a senior engineer champion co-owns technical design. Mixed internal and external support.
03 - Build a mid-project budget-risk early-warning view	A simple mid-cycle threshold flag pulled from existing project data, surfacing projects trending over budget before month-end close, reviewed at the existing weekly resourcing meeting.	Leadership explicitly lacks an early-warning indicator between reporting cycles, despite already having high-confidence project and financial data.	Leverage medium-high · effort low · value medium Scope 2–3 weeks · traction 1–2 reporting cycles · confidence high	Operations director, extending existing reporting ownership. Highly startable internally.

Markides Advisory - Use CaseAction sequence

09Tests and cautions

Each move carries a test, a caution, and a known gap.

Deliverables at close: a slide deck built for a single leadership meeting, a one-to-two-page written summary including the rationale for three co-equal moves, and a roadmap showing all three as parallel, independently startable tracks.

Move 01, permit tracking

Test: run the tracker on the next two to three projects and compare resubmission rate to the historical pattern, with a named owner and defined success measure in place before starting.

Caution: the founder's guardrail requires a named owner and success measure before this or any test starts, it should not begin informally.

Missing data: no breakdown yet of how much of the resubmission pattern is checklist-driven versus jurisdiction-specific variability.

Move 02 - QA bottleneck

Test: pilot on a subset of active projects and track senior-engineer QA hours against the 15–20% baseline, with a defined stop-go rule.

Caution: QA review is safety- and accuracy-critical. Any AI-assisted step must stay a first-pass aid, not a replacement for senior engineer judgement.

Missing data: no detail yet on which specific QA failure modes are most common, which would sharpen checklist design.

Move 03, budget-risk view

Test: run the flag alongside the existing monthly dashboard for one quarter and check whether it would have caught any project that later ran over budget.

Caution: included as a legitimate third move on the evidence, but sequenced slightly behind moves 1 and 2, which more directly address the binding talent constraint.

Missing data: no historical record of past budget overruns to calibrate the warning threshold against.

Sequencing logic

All three can start immediately given existing ownership and reporting cadence; none depends on another completing first. Permit tracking and the QA bottleneck sit slightly ahead in practical terms, with the budget-risk view running as the firm's second concurrent test.

What is deliberately not claimed

Directional impact expected: a reduction in senior-engineer QA time and permit resubmissions, plus earlier visibility into budget risk. No specific dollar or hour savings are claimed beyond the ranges the firm itself already reported.

Markides Advisory - Use CaseAction sequence

10Outcomes

Three tracks run in parallel, and the last governance gap closed.

Move 01, permit tracking

Piloted on the next three projects

Resubmissions tied to formatting or checklist inconsistency dropped to zero across the pilot, and the operations director extended the tracker to all active projects rather than running a further pilot cycle.

Move 02 - QA bottleneck

Ran alongside normal QA for one full project cycle

Senior engineers reported reviewing flagged items rather than starting from a blank drawing, and tracked QA hours came in modestly below the 15–20% baseline, enough signal to keep the tooling in place and extend it to the full active project list.

Move 03, budget-risk flag

Caught one project trending over budget three weeks early

Roughly three weeks before the prior monthly cadence would have surfaced it, giving the operations director time to address it with the client before it affected margin.

What happened next

With all three moves showing traction, the founder authorized drafting the firm's first formal AI use policy, closing the one governance gap the sprint had flagged but deliberately not elevated to a standalone move. The operations director began evaluating whether the QA tooling could extend into fee-proposal support, without disturbing the guardrail that proposals stay judgement-led.

What it demonstrates

Judgement in a different register than a turnaround case: knowing when not to recommend governance fixes, when three moves genuinely earn co-equal status rather than forcing a false ranking, and when to set aside a plausible-sounding move because the evidence did not support it.

Markides Advisory - Use CaseOutcomes

Markides Advisory

Book a discovery call to learn more.

markidesadvisory.com•michael@markidesadvisory.com

Book a Call