AI Transformation
Director

A case pack from a real, anonymised 300-person company: one P&L, one org chart, three pages of interview notes. The participant finds where money leaks, estimates the weekly cost (frequency × time × hourly cost), and names the three processes they would examine first.

Then a 30-item applied check on AI and systems: which tasks need a model and which need a rule, what a system of record is, what breaks when an agent writes to a CRM. Scored against the reference, not against other participants.

The five dimensions of the entry assessment. Each produces a score from 0 to 100; the composite places the participant on one of four stages.

Twenty scenarios taken from real engagements. A founder demands lead generation when fulfilment is broken. A board member calls AI “a fad” in front of the chair. A pilot misses target at the halfway mark. A department head “has no issues”.

The participant ranks four responses and justifies the top choice in two sentences; scoring is against a panel of practising Directors. Then a calibration test: ten estimates with 90% confidence ranges (how long, how much, how many). Over-confidence here predicts missed deadlines and broken promises to clients.

The Director Trait Inventory, a 360 from colleagues, and a live 30-minute simulator session with the founder of Case 1, who opens with “everything is running smoothly, your software cannot help us”. The session is recorded and scored on composure, listening, candour and authority. The participant watches it back and writes what they would change.

The Director Trait Model. Eight traits, each with a failure mode on both sides. Directors fail far more often from a trait than from a missing technique.

Placement. The composite score places the participant on one of four stages and a matching track.

The Director is the company's external CEO for transformation, a co-founder of the change rather than a subcontractor. Four contrasts set the identity:

• IT builds to a specification; the Director writes the specification of the business.
• A consultancy hands over a diagnosis; the Director stays for execution.
• A vendor sells one template; the Director designs the operating model every tool must fit.
• An FDE builds the system right; the Director decides which system should exist.

The five kinds of companies that sell AI, and why only one changes how a business works.

Organisations were designed when information was expensive to move: small steps, specialists, managers who coordinate. That layer now creates overhead. The core skill of the role: separating information-moving work (AI does it) from judgement work (people do more of it).

The five waves of AI, from automation to the AI-native organisation. Why transformation is a dependency chain and not a portfolio, and what breaks when a step is skipped — usually months after go-live, and usually disguised as a technology problem.

Who the Director talks to: board, CEO, CFO, COO, department heads, frontline, IT, legal, vendors. Each needs its own language. The six accountabilities, and why one person cannot hold them all in a large company. Where the Delivery Lead and the Architects begin.

Five sentences that separate a Director from an order-taker:

• “AI does not solve this.”
• “This is not the right sequence.”
• “A second person is needed here.”
• “That deadline is not achievable.”
• “We stop this project.”

Operations before automation. Organisation before technology. Capability before dependency. Audit before advice. Build for evolution, not for today's AI.

The participant's own unlearning comes first: the tool takes an afternoon, the unlearning takes a year — for the client and for the Director. The expert's blind spot: the more transformations you have run, the faster you pattern-match, and the more likely you are to miss the fact that makes this company different.

Framing: the stated request versus the real problem (“we need more leads” usually means “we cannot see our own pipeline”). One-way and two-way doors: decide reversible things fast, irreversible things slowly. Second-order effects across departments. The pre-mortem.

Knowing when a number looks right but is not — the invoice re-pricing that found 70% unapproved discounts with no AI involved. Traps: sunk cost, the shiny tool, local optimisation, the highest-paid person's opinion.

How to decide with 70% of the information and say what the other 30% is. The Decision Journal: decision, rejected options, what would prove it wrong, confidence. Who is accountable when the AI is wrong.

Ethical judgement and walking away: the client red flags that disqualify an engagement, and the cost of taking the wrong client. The kill decision: stopping a project you started.

A 12-month engagement is a stamina test. Emotional regulation and personal triggers: the founder who interrupts, the board member who mocks, the department head who smiles and stalls. Boundaries: scope, hours, hand-holding.

Ego and attachment to one's own solution. Impostor feelings in front of technical staff, and the opposite risk, over-explaining technology to compensate. Recovery routines. Review of the Week 1 trait plan.

What each audience fears and wants. The CEO: legacy, control, being outpaced. Department heads: tracking burden, loss of authority, burnout. Frontline staff: replacement. IT: territory. Legal: exposure. The board: fiduciary risk.

Protocols: the disarming opener, complementary validation, the employment-safeguard script (“this expands your capacity, it does not replace you”). What never to say: never rubber-stamp a broken routine, never amplify a grievance without a probe, never explain back-end mechanics — speak only in visible outputs (“a message arrives in Telegram”, “the invoice fills itself and sends”).

Authority without a title. Strategic re-anchoring: guiding a CEO from the solution they asked for to the root cause they have. Handling “we have no issues” by probing the person's own priorities. Two departments fighting over one resource. Giving hard feedback to a founder.

A coaching conversation in four moves: goal, reality, options, commitment. The signatory rule: compliance flows down from the decision-maker; subordinate resistance is converted, then escalated, never ignored.

Qualification criteria: annual revenue of at least $5M, so workflows and teams are complex enough to need redesign. Disqualifiers: a founder closed to change, a leader without decision authority, a mega-conglomerate or rigid government body for end-to-end work (training-only engagements excepted). The target: a founder-led or tight co-founder company with centralised flexibility, even across hundreds of sites.

The 48-hour contact protocol: direct channel to the decision-maker, systems overview, independent market research, logged pain points, the second meeting booked, credentials sent before it.

Option 1, Done-For-You Systems: the MagnaQore team designs, builds and deploys selected systems. Option 2, Internal AI Department Builder: a 6–12 month transformation that audits, co-builds, hires the permanent AI team and trains the company.

The Option 1→2 pipeline: a Done-For-You engagement is scoped so that its proved value becomes the business case for Option 2.

Dual-layer pricing: business outcome plus physical deliverables, set with the technical lead. 3–4 instalments. Retainers sized by number and complexity of systems. The fixed-fee rule: never lower the rate, remove deliverables instead.

Mandatory clauses: technical feasibility verification before any proposal; the third-party technology disclaimer; data governance and residency; phased IP transfer on milestone payment, with the methodology retained by the firm; force majeure; the signatory rule. Refund boundaries: full before the audit, partial after it, none on client-side failure.

The 7-day pre-engagement window: shared resource folder, full org chart with duties, system access, calendar visibility, mutual NDA, and the leadership self-audit questionnaire. Early warning signs that justify exiting before Day 1.

An organisation cannot govern what its leaders do not understand. The seven questions most companies cannot answer:

• Who can change a process?
• Who can remove an approval?
• Who can change a KPI?
• Who can give AI access to data?
• Who can change a role?
• Who can stop a project?
• Who is accountable when the AI is wrong?

Adoption shows people are using AI; readiness shows the organisation knows what to do with it. Designing the leadership programme: 8 sessions at 2 per week — AI fundamentals and sector cases, governance and responsible AI, risk and decision frameworks, leadership simulations, the board AI strategy blueprint. Every director assessed before and after.

The “zoom out” method: lift the board from features to the ecosystem. Framing for owners: market position, proprietary platforms, operating models that can be packaged and sold. Handling the hostile director and the over-enthusiastic one in the same room.

Leaving with a mandate, not applause: a named sponsor, decision rights, a budget envelope for the assessment, and a reporting cadence.

The working contract with the CEO: what they decide, what they delegate, what they hear first.

The first 90 days: mobilise and map (days 1–30), diagnose (31–60), quantify and decide (61–90). The weekly executive review starts in week one, not after the audit. The sponsor map: who carries the programme when the CEO is absent.

The enterprise assessment runs on eleven tracks: operational friction, process architecture, AI readiness, transformation readiness, opportunity per company, data readiness, technology landscape, people capability, governance and legal exposure, business value, risk. Most have nothing to do with technology.

The ingest: org schematics, job descriptions, knowledge bases, software inventory verified by IT. The standard project folder. The Delivery Lead as corporate memory. All transcripts fed into a shared, access-controlled knowledge base.

The sequence: CEO (vision, friction, expectations), COO (handoffs, weak teams, reschedule interviews by real priority), IT director (systems, security, access), general counsel (data, IP, storage law), then frontline (the manual reality and the workarounds). Open questions; why “tell me about your processes” fails.

The decoder: “we need more leads” means no visibility; “maximise efficiency” means redundant reporting; “we are overwhelmed” means poor role distribution. The 5-Why, the soft closing probe, and the stop-digging standard. Contradictions between accounts are data, not noise.

Physical signals: walking between desks, huddles to check paperwork, repeated follow-up emails — information living in heads, not systems. Shadow processes.

Department interdependence: Sales, Finance, HR, Support, Legal, Operations. External touchpoints: carriers, brokers, suppliers. An anonymous employee instrument. Mapping the seven decision-rights questions against what actually happens.

Human-centric, step by step: where the person pulls data, follows up, stores the result. Real templates and screenshots linked into each node. The blueprint target standard: an outsider could do the employee's job tomorrow from the map alone.

Validated live with the department director and the people who do the work. Where the process tears: manual transfers, double entry, approval waits, work in private messages.

The three readiness tiers:

• Process-mature (disciplined but manual): automate directly.
• Operational chaos (memory leaves with the employee): build the process foundation first.
• Hyper-controlled (layers of approval): condense the policies first.

Data readiness: duplicates, gaps, records nobody owns. Technology landscape and legacy blockers. Governance and legal exposure: contracts, data residency, outbound-communication law.

A heatmap of function × friction type. Value streams, systems of record, handovers and decision rights drawn on one page: business architecture v1. This is the picture the operating-model redesign in Block 4 will change.

Value × feasibility scoring across all opportunities found (37 in the source case, 19 needing no new data).

The quick-win filter: active financial bleed, friction consuming 70%+ of a role's day, no dependency on unbuilt architecture. Model or rule for each. The baseline: four numbers per initiative, measured before anything is built.

The client portal: findings, maps and loss calculations as a polished interactive page, never raw backends.

The board readout: the number, the cause, the sequence, the decision required. What to do when the CEO rejects the finding. The decision on what proceeds.

The sequence: mindset → process → unlearning → redesign → only then the technology. Three tiers, plus a track for internal IT:

• Executives and directors: 8 sessions, 2 per week, on process design, data structure and automation logic.
• Champions: workflow mapping, tool selection, practical systems.
• Frontline: 6 sessions on daily use and data hygiene.
• Internal IT: a 1–2 month co-development track.

Every session ends with something working, not a certificate.

One senior champion per department, chosen for curiosity, peer influence and process history — not by the department head's preference.

The activation routine: apply the lesson to live work the same evening, report back next session.

Project-based evaluation: every manager builds and deploys one working asset by session 8, judged pass/fail on whether it moves their real KPI. The employee capability development plan.

Most companies are built like a machine: remove a part and it stops. AI needs something closer to Lego. The core question is which processes should exist at all.

Three moves: automate the information-moving, redesign the judgement, then change what the person is measured on. Every handover was once a saving; now it may be a cost. Removing approvals and coordination meetings before adding tools.

One system of record per entity: client, order, shipment, invoice. Named process owners, including across company boundaries in a group. Where the human stays in the loop and why.

The rule the Director enforces later in build: the development team never reorders a business process to simplify the code.

The seven questions answered for the to-be model. A responsibility matrix for AI-related decisions. Approval layers removed or kept, with reasons. Escalation paths that work when the Director is not in the room.

Role by role: what is automated, what judgement remains, what the role becomes. In the source case, eight years of data moved into one operating system with eleven defined roles.

Capacity scenarios and the ethics of redeployment. How to announce a role change so people hear growth, not threat.

A KPI tree from board to company to department to role. Outcome metrics over activity metrics. Adoption measured as changed behaviour, not log-ins. Quality and risk tracked beside cost and time.

Incentives aligned with the new roles — otherwise people keep doing what they are paid for.

The unlearning plan. The resistance map. The communication cascade through executives, champions and frontline. The first-week anxiety curve of every launch. Reinforcement through KPIs, rituals and visible wins.

One plan per company or unit, translating the target model into that business and sequenced against what the assessment actually found. This is the step that stops every technology question landing on the owner's desk.

The roadmap follows the dependency chain, with a gate between stages and criteria that must be met to pass it. Named owner, budget and baseline on every initiative. What is deliberately left off, and why.

Three months of evidence come before any build decision. Six stages, twelve steps, five gates: the order can flex; the dependencies cannot.

Presenting to the board: the sequence, the cost, the return, the risks, the decision required. Trading scope without breaking the chain.

Then the mid-point reassessment: the Week 1 instruments re-run with new scenarios, and a one-to-one with an assessor.

One intake and approval gate for every AI request in the company. A portfolio register: what exists, who owns it, what it costs, what it returns.

Standards that stop three teams building three versions of the same thing. One named owner per initiative.

The policy set: acceptable use, data, vendors, ethics and responsible AI. Board oversight structure. The risk register and the one-page risk report the owner actually reads.

The Director is not the lawyer, but must spot the problem before the code is written.

Outbound-communication law: in the US and UAE, calling restricted lists can carry criminal penalties, so filters are designed in. Consent disclosures at the start of recorded or automated calls. Data protection and residency: the split architecture — protected records stored in-country, model calls through an encrypted layer. Emerging AI regulation at board level.

Verifying a tool through the vendor's enterprise terms, data processing agreement and security reports. Negotiating a custom data-isolation clause. Preferring vendors already approved inside the client.

The third-party disclaimer in practice, and the pivot plan when an API or a regulation changes overnight.

The warning-signs matrix by phase:

• Leadership missing calls and delaying payment before Day 1: exit and return the deposit.
• Department heads refusing to release staff for mapping: pause and escalate to the CEO.
• Stakeholders dismissing the system over first-week errors: compare against the human error rate.
• Scope disputes as invoices approach: suspend access until settled.

Protecting the methodology as the firm's IP.

Operating knowledge that is remembered rather than recorded is a succession risk and the biggest limit on what AI can safely do.

The knowledge audit: what lives in heads, chats and private drives, and what leaves when a person resigns.

The company's procedures, terminology, constraints and standard cases turned into a governed, permissioned source with an owner and an update cadence.

The rule “no source, no answer” for anything an AI tells a client or an employee.

The management rhythm that keeps the transformation alive: intake gate, weekly executive review, monthly department reviews, quarterly board review, knowledge updates, and the four-week optimisation cycle. Each forum with an owner, inputs, outputs and a duration.

The spot-check test: any employee can pull an accurate history of any client from the system on request.

Build where the way of working is the advantage. Buy where mature technology already solves it. Rent commodity capability, and stay portable.

Total cost over three years, lock-in, exit cost. The principle: the technology should be replaceable, the capability should remain.

What a Director must understand without coding: systems of record, integrations, APIs, data flows; model versus rule; agent versus fixed workflow; document search with citation versus a file in context; evals, monitoring and rollback. Reading an architecture diagram.

Three questions that test any technical judgement, from a person or an agent: where will it break, what happens at ten times the volume, which option was rejected and why. The reference stack used in the labs (Google Cloud, Gemini and Workspace), taught alongside the portability rules.

From opportunity to solution brief: business outcome, process map, acceptance criteria, baseline, constraints, red lines. Technical feasibility checked before the client hears a promise. Estimating and pricing with the team.

When the Director steps in — undocumented process variations, or a team changing the business order to simplify code — and when the Director stays out of the sprint.

The four-role foundation, reporting lines and the nine standing responsibilities. Growth phases by trigger.

The budget: built per client from local salary levels and the hiring plan, not from a fixed figure, and tied to a target return on the ROI control panel. Fractional versus internal leadership.

Scorecards per role, built from the certification standards. An interview loop with a practical test for each role: the Delivery Lead onboards a simulated department, the Systems Architect designs a CRM lead-scoring integration, the Revenue Architect designs a prospecting system.

The Director's hiring veto. Growing internal talent: champions who become Delivery Leads.

The weekly rhythm: Monday priorities, protected build time Tuesday to Thursday, Friday demos and retrospective. Capacity rules: up to 4 custom rollouts per person, up to 10 template rollouts per team.

Settling disagreements between Delivery Lead and Architect through a senior technical review or structured user testing. Handing department leadership to an internal head.

Build only where value was proved. Workflow logic is locked before the build; no new features during testing.

Rollout in three moves: a 15–30 minute video guide, an all-hands live walkthrough, weekly optimisation for 1–3 months. The anxiety curve: days 1–7 resistance, day 7 one-to-one manager check-ins, day 14 cross-department review.

Long-term financials cannot be read in two weeks, so the Director tracks proxies: errors removed in document-heavy work, pipeline velocity in client-facing work, admin hours reclaimed.

Usage dashboards by employee name. Comparing system accuracy with the human error rate, not with perfection.

The halfway-mark decision tree:

• External cause (API change, outage, client delay): document it, issue a formal delay notice, pivot the architecture if needed.
• Low usage: an enforced training track — two-hour live screen-share clinics, tracked by name.
• High usage but poor results: the audit was wrong; re-diagnose with the Architect and redesign.

The kill-switch protocol: never terminate a live system unilaterally unless payment defaults; if the client refuses to use a working system, hand over everything and close the account.

Baselines captured before deployment, or the return can never be proved. A named owner per initiative. Adoption measured as changed behaviour. Quality and risk tracked beside cost and time.

The ROI measurement framework, agreed with finance before the first result arrives.

One page for the CEO, one for the board. Leading versus lagging indicators.

A dashboard built to trigger decisions, not admiration: every metric has an owner, a threshold and the action taken when it is crossed.

Attribution honesty: what AI caused versus what the redesign or the market caused. Counterfactuals. What never to claim. The department scorecard and the monthly ROI review.

Sixty minutes with the CEO, fixed agenda:

• Scorecard — 10 min
• Roadmap and gates — 15 min
• Decisions required — 15 min
• Risks and escalations — 10 min
• People and adoption — 5 min
• Commitments — 5 min

Pre-read sent 24 hours before. A decision log. The rule: the CEO is never surprised by something the Director knew.

Coaching the CEO: giving up information-moving, owning decision rights, modelling the new behaviour in public. Coaching department heads: from coordinators to judges — which decisions they keep, which tasks they delegate to people or systems.

A written coaching plan per leader. The hardest conversation: telling a founder that they are the bottleneck.

The four-week optimisation cycle from month 4 onward:

• Week 1: managers evaluate team output; the Director runs one-to-one reviews.
• Week 2: the Delivery Lead plans developer time and collects access and data.
• Week 3: research and safety testing; the plan goes to the executive team.
• Week 4: the Architects ship.

Monthly department audits against the KPI checklist. The quarterly board review.

The handback package: credentials, source code, system maps and data flows, training videos and guidebooks. Training the team, not one individual.

The support agreement: what is included and what is billed. The client owns the build; the firm keeps the methodology, prompts and training materials.

The spot-check test. The unified governance standard: financial reviews automatically trigger updates to AI systems. The autonomous innovation cycle: employees propose improvements without being asked.

The final test: if the organisation still needs an external provider to decide what to build next, the programme did not work. The exit interview with the CEO: the starting state against today, the change in leadership mindset, and referrals.

Opportunity hunting after go-live through monthly audits. Expanding solutions that proved value. The Option 1→2 pipeline in practice. Fractional Director retainers.

Helping the client commercialise its own systems: a SaaS product, a franchise model, a packaged operating model.

Case 5 is released on Monday: a new industry — manufacturing, a healthcare group or a multi-entity holding. Interviews in hierarchy order, maps, loss calculation, opportunity matrix, readiness tier. A test of method, not of speed.

To-be operating model, decision rights, governance essentials, AI department design, 12-month gated roadmap.

Option 1 or Option 2 with reasons, scope, timeline, price, instalments, retainer and the contract clauses that matter for this client.

The eight Director deliverables from Case 1, the Case 5 pack, and the Decision Journal reviewed against what actually happened in the cases.

The Case 5 pack and the full portfolio, defended in 90 minutes: a 20-minute presentation to a simulated board, then 40 minutes of questions from the board and two human assessors — one from MagnaQore, one nominated by the certification partner.

Then a 30-minute pressure round chosen by the assessors on the day — for example, the founder demands to skip governance to go faster, or a director attacks the cost of the AI department. It is scored mainly on judgement and composure: was the position right, and was it held for the right reasons?

The Week 1 instruments re-run. The growth report across five dimensions and eight traits. A personal development plan for the first year in the role.