Mohab Fathalla | Commercial, technology and quality engineering advisory

01 Beyond the plan

What makes things work is what no plan can specify. Nobody owns it, and that is where value is won or lost.

From my practice

For seven years I have been the person a partnership calls when the contract is silent. A market needs something no agreement priced, and needs it soon. My work is to find the right teams on both sides, settle what sits inside the agreement and what sits on top of it, and carry the request through to delivery and invoice.

Almost none of that work is written in the contract. Most of the value is.

The plan is the contract: what both sides could foresee on the day they signed. What it cannot specify is everything that arrives afterwards. The need nobody priced. The promise made in a meeting that never reached the paper. The day something goes wrong and the contract is silent.

Because it is impossible for a contract to specify every eventuality, the 2016 Nobel committee wrote, the question that matters is which party is entitled to decide when the contract is silent. Most partnerships never answer it.

The Nobel Prize in Economic Sciences, 2016: Hart and Holmström

For its first decades, artificial intelligence tried to make computers intelligent by writing the world down in mathematical models. Every fact, every rule, every state of affairs, described in advance. It never scaled: the world has more states than anyone can write. The breakthrough came when the field stopped describing the world to the machine and let the machine learn from exposure to it.

Partnerships are still run the old way. The contract tries to describe every future state. The partnerships that work make the same shift: stop listing every engagement point, and build the ability to deal with contention and the unforeseen when they arrive.

Each side governs its own side well, and no reporting line crosses the company edge. So what the plan cannot specify fills with private information. The supplier knows the programme is slipping. The client knows the strategy is about to change. Across fourteen studies of complex programmes, status reports were systematically biased, audits made the bias worse, and executives discounted the bad news. Service levels stay green while the relationship turns red. The industry calls it a watermelon.

Not a tighter contract. Four things, each with evidence behind it:

A named owner on each side with the mandate to decide when the contract is silent. In 119 buyer-supplier pairs, the more discretion the person at the boundary had, the more the other side trusted them.
An agreed way to deal with contention and the unforeseen, written into the agreement as guiding principles.
A shared record of what has been asked, promised and delivered beyond the contract. Today that memory lives in inboxes and in 24 different systems.
Governance that decides the next thing instead of accepting the last report.

Perrone, Zaheer and McEvily, 2003. Frydlinger, Hart and Vitasek, HBR 2019. WorldCC with Deloitte, 2023.

A cable-stayed bridge under construction, seen from above, a crane on the unfinished deck — The plan reaches as far as both sides could see. The rest is built from both banks.

A vast flock of birds in flight over open land — The plan reaches as far as both sides could see. The rest is built from both banks.

1 / 5

8.6%

average erosion of contracted value; the best performers lose a little over 3%, the worst more than 20%. WorldCC with Deloitte, 2023, 1,236 organisations.

Almost half

of strategic alliances fail. Dyer, Kale and Singh, MIT Sloan Management Review, 2001.

63% v 50%

long-term alliance success with and without an organisational home for the relationship. Kale, Dyer and Singh, 2002.

WorldCC has tracked the most negotiated contract terms for well over a decade. The same clauses sit at the top year after year, and WorldCC keeps finding a gap between them and the terms that matter most to outcomes. Companies put their most senior attention on the plan and far less on what it cannot specify. WorldCC's own conclusion says it in fewer words: the contract alone is not enough.

In practice

More than €10M a year of client and partner needs turned into commercial value beyond contracted revenue. Advisory on a partnership in pharmaceutical distribution.
Executive-level governance that turned round a failing strategic partner relationship.

The partnership auditSeven questions

Seven questions about what the plan cannot specify in any partnership that matters. A weak answer to any one of them tells you where the value is going.

Who owns what the plan cannot specify? Name one person on each side who may decide when the contract is silent, and say what they may decide without escalating. If the answer is a committee, nobody owns it.
What has been asked in the last quarter that the contract does not cover, and where is it written down? If the list lives in inboxes, the relationship has no memory.
What is the agreed way to deal with contention and the unforeseen? Who calls whom, within what time, and by which principles is the first decision taken?
When did the dashboard last say green while someone senior on either side said otherwise, and what happened next?
What does each side know that the other needs, and what does keeping it private cost? The supplier's real state of delivery and margin; the client's budget, strategy and intent to renew.
What did the last three governance meetings decide about the future, as opposed to accept about the past?
What would the partner say the relationship is worth beyond the contract, and would your number match theirs?

02 Practice

Four disciplines for what the plan cannot specify.

One notion each, what the evidence says, what I do, and an instrument you can use without hiring me.

The risk is private information.

From my practice

For eleven years I sat on the supplier's side. I know what a delivery team writes in the status report and what it says in the corridor. For seven years I have sat on the client's side. I know what a market keeps from its supplier until the renewal. Both sides were acting rationally. Both silences cost money.

What the research says

Bad faith is rare. Private information is universal. The supplier knows the real state of the programme and its margin. The client knows the budget, the strategy and whether it intends to renew. Economists call this asymmetric information, and in outsourcing it runs both ways: clients are unsure of a supplier's capability, suppliers unsure of a client's requirements.

Closing the gap pays both sides. In 753 large contracts, where an independent party sat between client and supplier, suppliers earned more and contract outcomes improved; fewer than 5% of contracts had one. Across 23 years of outsourcing research, communication and knowledge sharing were the most consistently positive predictors of outcomes, ahead of contract detail.

What I do: make the truth cheap to tell early. One record both sides trust. An early-warning convention with no penalty for the first signal. A person on each side who can carry a hard message across the edge before it becomes an escalation. Having held the supplier's private information for eleven years and the client's for seven, I know what each side holds back, and what the silence costs.

One caution from the research: trust does not grow with time. Across 168 tests it follows an inverted U over the life of a relationship. A long partnership needs the record more, not less.

Bapna, Gupta, Ray and Singh, Information Systems Research, 2016. Lacity, Yan and Khan, 2017. Zhong, Su, Peng and Yang, Journal of Management, 2017.

A brass balance on a wooden shelf — Two sides, one record. The balance holds when both can see the same weight.

In practice

Executive-level governance that turned round a failing strategic partner relationship.

The watermelon testThree pairs

Three pairs of measures. A gap inside any pair means green outside, red inside.

Pair	Outside: what the dashboard shows	Inside: what the people say
Service	Service levels met, as a percentage	Experience score from the people who depend on the service, asked quarterly
Demand	Requests formally logged and priced	Requests actually raised, including the ones done as favours or refused quietly
Warning	Escalations recorded	Early warnings given before the escalation; none means the channel is closed, not that nothing is wrong

The rule: if the green needs a footnote, it is a watermelon.

Decide the next thing.

From my practice

When a strategic partner relationship was failing, more steering meetings were not the answer. It turned round when senior people on both sides agreed who decides what, and met to decide the next step instead of reviewing the last month.

When a billing-rate gap surfaced, it was closed within 24 hours of being raised, because the people who could decide already knew and trusted each other.

What the research says

Most governance looks backwards: status, service levels, penalties. Status reports are biased by design, and complex programmes fail one day at a time, after warning signs that were reported and discounted. Governance earns its cost when it decides the next thing.

Size it to the stakes: senior time where the money and the risk sit, clear rules for the rest. Organisations with well-developed minimum viable policy cut the time to make complex or high-stakes decisions in half and found opportunities and problems at three times the rate of their peers. In software delivery, formal approval boards made teams 2.6 times more likely to be low performers, while a clear change process made them 1.8 times more likely to be elite.

Earned autonomy: oversight should fall as evidence accumulates. A partner that shows its work earns the right to run with less of it, and loses that right at once when quality drops. The ladder is written down, and both sides can see where they stand.

Keil, Smith, Iacovou and Thompson, MIT Sloan Management Review, 2014. MIT Center for Information Systems Research, 2023. DORA, Accelerate State of DevOps, 2019.

A hand on a ship's wooden wheel — Governance is the hand on the wheel, not the log of where the ship has been.

In practice

Executive governance behind a nine-figure, five-year renewal.
Commercial continuity through a corporate separation.
About €5M recovered through rate assurance.

The earned-autonomy ladderPartners and agents

The same ladder for a partner and for an AI agent. Evidence buys the next rung; a quality failure costs one at once; the ladder is written down and both sides can see where they stand.

Observe. Acts only when instructed; every action checked.
Recommend. Proposes; a named person decides.
Act with approval. Acts after a yes; a sample of actions is reviewed.
Act with rollback. Acts within set limits; every action can be reversed; exceptions are reviewed.
Autonomous within audited bounds. Acts; the evidence is reviewed on a cadence; one quality failure drops a rung.

What stands behind it: the EU AI Act requires human oversight commensurate with the risks, level of autonomy and context of use, with the ability to override or stop. The human-factors research on automation says people fail to rely on it appropriately unless trust is calibrated to demonstrated capability (Lee and See, 2004). And 5% of companies have the full set of controls for agent autonomy today (BCG, 2026).

Fit for purpose is a business judgement.

From my practice

In eleven years of quality engineering, from avionics to banking and insurance, the programmes that slipped rarely failed on the code. They failed where business, development and testing handed work to each other.

When a telecoms operator's programme was in crisis, quality engineering was how we took back control: one view of what worked, what did not, and what to fix first.

What the research says

Juran defined quality as fitness for use. The current international standard defines it as the degree to which a system meets the stated and implied needs of the people who depend on it. The implied needs are the point. Only the business in the room can state them, which is why business, technology and testing belong together from the first requirement, never in sequence.

The evidence agrees: teams that focus on the user show 40% higher organisational performance, and elite teams fail 5% of changes against 64% for the lowest performers. The most striking finding is older. On one system of fifty million lines of code, how the organisation that built it was structured predicted where it would fail with 86% precision, better than any measure of the code itself. Who builds it, and how they are organised, is a quality decision.

Quality as foresight: on one insurance programme I built a predictive model from development, testing and defect data that showed where effort would protect the launch date and where it would be wasted. It was artificial intelligence in notion rather than in technology, built before the current tools existed.

ISO/IEC 25010:2023. DORA, 2023. Nagappan, Murphy and Basili, ICSE 2008. CISQ, 2022.

Every part fit for its purpose, or the whole stops.

In practice

A 75% reduction in regression on a core-platform migration in insurance, through one model for business, development and quality assurance.
Programme director of group-wide ERP harmonisation for an engineering group; delivery governance for the roll-out of an ERP harmonisation across a global flavours and fragrances group.
For scale: CISQ estimates the cost of poor software quality in the United States at $2.41 trillion for 2022.

Fix the business first, then design for what AI is.

From my practice

My first piece of AI was a spreadsheet. On an insurance programme, long before today's tools, I built a model that read development, testing and defect data and showed which team to focus on to hold the launch date. It worked because it started from a business problem, not from a technology.

My current AI work began the same way: from the gaps between what was estimated, quoted, delivered and billed across a partner portfolio, and from who owned each gap. The models came second.

How I apply it, and what the research says

Fix how the business runs before any technology talk.

Every AI conversation I join starts with the same question: where does the daily work lose money or time? The handoff that takes three weeks. The request re-keyed four times on its way to an invoice. The escalation that should have been an early warning. Sometimes the fix is a process. Sometimes it is a cultural change. Only sometimes is it a model, and you cannot know which until you have looked at the work.

Invest where money is lost. Give it one owner.

We do not digitise everything and follow the trend. The investment goes to the place where money leaks, and one person owns the result with a number to answer for. The record of doing it the other way round is poor: 37% of organisations attribute any profit impact to AI, the same as a year earlier, while 80% of respondents say it improved their own productivity; high performers are stuck at 6%. 56% of 4,454 chief executives report no significant financial benefit. The companies that get returns put 10% of their resources into algorithms, 20% into technology and data and 70% into people and processes, and redesign the workflow rather than insert AI into it.

Design for what AI is.

A language model predicts the most likely next word. It does not think. It forgets everything between conversations. It is confidently wrong. None of this is an argument against using it; it is the design brief. Memory comes from the organisation, in the record of asks, promises and decisions. Truth is verified: if it did not emit evidence, it did not happen. Judgement stays with a named person. Autonomy is earned one rung at a time.

Strategic value is what the business can newly see, remember and promise.

See: every new demand read against the agreement, in scope or incremental, allowed or not; the gap between estimate, quote, actual and invoice surfaced as it opens. Remember: one memory of asks, promises, decisions and outcomes that survives renewals, separations, staff changes and M&A; this is the organisation supplying the memory the model lacks. Promise: verified outcomes with a cost per outcome, and autonomy earned through evidence, by a partner or by an AI agent; this is the answer to a model that is confidently wrong. The bottom line follows from all three.

The constraint is governance, and judgement is the bottleneck.

42% of companies expect to give AI agents real decision-making authority by 2030; 5% have the full set of controls in place. The EU AI Act requires human oversight commensurate with the risks, level of autonomy and context of use. And as analysis gets cheap, judgement becomes the bottleneck: AI widens what the owner of the work beyond the plan can see, and the owner's judgement is the scarce input. The ladder I apply to partners is the ladder I apply to agents.

McKinsey, The state of AI in 2026. PwC, 29th Global CEO Survey, 2026. Brynjolfsson, Rock and Syverson, 2021. BCG, 2024 and Applied AI Index 2026. McKinsey Quarterly, October 2026. EU AI Act, Article 14.

A seal pressed onto a sheet of paper, the stamp beside it — If it did not emit evidence, it did not happen.

Fix the business firstThree questions

Three questions to answer before any technology talk. If the third has no name in it, stop.

Where does the daily work lose money or time? Name the handoff, the re-keying, the wait, the escalation that should have been a warning.
What has to change first: a process, a habit, or a tool? Say which, and why a tool would not fix the first two.
Who owns the result? One person, with the number they will be measured on.

The design rules that follow, for whatever is built: memory comes from the organisation, not the model; truth is verified, or it did not happen; judgement stays with a named person; autonomy is earned one rung at a time.

The AI strategic value mapSee, remember, promise

Three capabilities, each tied to a question a board already asks and to the limit of the technology it answers. The bottom-line column is deliberately last.

Capability	The board's question	The limit it answers	What AI changes	Strategic value	Where the bottom line shows up later
See	What is happening in this partnership that nobody is reporting?	People protect private information	Every demand read against the agreement; estimate-to-invoice gaps surfaced as they open; unpriced needs made visible	Decisions taken on the present, not on last quarter's report	Leakage recovered, incremental work billed, disputes avoided
Remember	What did we promise each other, and what did they promise us?	A model has no memory of its own	One memory of asks, promises, decisions and outcomes across renewals, separations, staff changes and M&A	The relationship survives the people who built it; renewals start from the record	Renewal terms grounded in evidence; fewer concessions bought with forgotten history
Promise	What can we commit to, and how do we know it will hold?	A model states falsehoods with confidence	Verified outcomes, cost per outcome, evidence trails; earned autonomy for partners and agents	Commitments made faster with less oversight; trust calibrated to evidence	Lower cost of control; shorter cycle times

Start in the row where the board's question is most uncomfortable. That is where the first system goes.

04 About

Engineer by training, commercial leader by practice.

I studied Computer Science and Engineering at the German University in Cairo. My thesis built an AI interpreter in modal logic that tells necessary truth from possible truth. It belongs to the era when the field still tried to write the world down, and the question it asked still shapes my work: what must be true for a deal or a programme to succeed, and what only might be.

The first eleven years were on the supplier's side, leading transformation programmes and the teams behind them for Fortune Global 500 clients across Europe and the Middle East, with quality engineering as the discipline that kept each one fit for what the business needed. The last seven have been on the client's side, holding commercial mandates for partner portfolios across five European markets: the master services, transitional services and partnership agreements, and every business need that arose beyond them. Along the way I have worked alongside many chief technology, chief information and chief financial officers, often on programmes they commissioned directly.

I founded MF Business Consulting in 2022 to bring that range to leadership teams. Through it, I advise boards and executive teams across Europe, the Middle East and Africa on cross-border partnerships, M&A and commercially grounded AI adoption.

Education: B.Sc. Computer Science and Engineering, German University in Cairo, 2008
Executive education: Harvard Business School Online: CORe (Credential of Readiness) and Leading with Finance. MIT Sloan: Digital Transformation and Platform Strategies.
Certification: Certified Professional for Requirements Engineering, International Requirements Engineering Board (IREB)
Clients: Allianz, Airbus, Deutsche Bank, Vodafone, Kuehne+Nagel, Firmenich, Origin Systems and Liberty Global, including its markets VodafoneZiggo, Virgin Media O2, Virgin Media Ireland, Liberty Blume and Sunrise
Languages: English and Arabic

One seat, three disciplines

Value is rarely lost inside a discipline. It is lost in the handovers between them.

Commercial × Technology

Commercial without technology sells what cannot be built. With technology, every request is checked against the agreement and against what delivery can carry, before a price is given.

Technology × Quality engineering

Technology without quality engineering ships late or ships wrong. With it, the slip is visible before it happens. On an insurance programme, a predictive model built from development, testing and defect data showed which team to focus on to hold the launch date.

Quality engineering × Commercial

Quality engineering without commercial protects the product but not the money. With commercial, estimate, quote, delivery and invoice are reconciled as soon as they diverge. More than €10M a year of client and partner needs became commercial value beyond contracted revenue, and a billing-rate gap closed within 24 hours of being raised.

Together, the three close the loop: what is sold can be built, what is built holds, and what holds gets paid. I built all three on the supplier side and have used them from the client side for seven years. I know what each side holds back, and what each handover costs.

Range

From avionics to telecoms, from the delivery floor to the commercial table.

The same person in the delivery seat and at the commercial table, in seven industries, across Europe, the Middle East, Africa and Asia.

Beyond the contract. Beyond the technology.

Both sides of the table.

Commercial and partnerships

Quality engineering

Tech strategy

Beyond the plan

Practice

Work

About

What makes things work is what no plan can specify. Nobody owns it, and that is where value is won or lost.