Services

Link AI to your business strategy

AI pays off when it changes a business result, not when a pilot works. BCG estimates that only 10% of the value of an AI transformation comes from the algorithms, 20% from data and technology, and 70% from changing workflows, people and processes. McKinsey's 2026 survey finds that nearly three-quarters of the companies getting the most from AI have redesigned their workflows around it, against about a quarter of the rest. I start from your strategy and your numbers, and end with a plan your leadership can commit to and track.

Value discovery

I start from your business strategy and P&L, not from the technology. Through interviews with your leadership and process owners, I identify the few areas where AI can move a measurable result, such as revenue, margin, cost to serve, working capital or cycle time. Each area gets a baseline and a target.

Use case portfolio

I score candidate use cases on value and feasibility, using evidence on data readiness rather than assumptions, and match each to the right technique:

• predictive models for forecasting and pricing
• optimisation for planning and allocation
• generative AI for documents and language
• AI agents for multi-step workflows

Many problems are better solved by a well-tested statistical model than by a language model, and some do not need AI at all. I will tell you which is which.

Roadmap and business case

A sequenced plan with early results within months and larger changes over one to three years. Each use case has an owner, a KPI, a baseline, an investment estimate and a decision point to continue or stop.

Operating model and governance

Who owns AI, which skills to hire or develop, and what to build, buy or partner for. I give independent advice on vendors and platforms, with no reseller ties. Governance covers:

• review gates for models before release (testing, approval and documentation)
• explainability and bias controls where decisions affect people
• an approved-tools policy for staff
• your obligations under the EU AI Act

Adoption and value tracking

I help redesign workflows around the new tools, train your teams, and measure results against the agreed baseline. AI programmes rarely fail because of the model; they fail when nobody owns the outcome.

Ongoing option: part-time AI leadership

I can act as your senior AI lead a few days a month: running the roadmap, overseeing vendors and delivery, and reporting to management or the board, until you are ready to hire your own team.

Related work: MBC: an eight-stage deal workflow mapped end to end, with human approval at every stage. Digital Egypt: measurable phase-1 targets agreed with stakeholders before the build began.

Make your data fit for AI

Data is the raw material of every analytics, machine learning and AI project, and one of the most common reasons they stall. I tell you what your data can support today, what it would take to fix, and how to manage it so the problem does not return.

Data value assessment: know what your data can support before you invest

• Avoid costly dead ends. Data problems are among the most common reasons AI projects fail. Gartner expects organisations to abandon 60% of AI projects that lack AI-ready data through 2026, and RAND's interviews with experienced practitioners name inadequate data as one of five root causes of failure. The assessment shows what your data can and cannot support before you commit a budget.
• Fix what matters first. Every finding traces back to the failing checks and sample records, and I rank fixes by their effect on your planned use cases. Better data often beats a better model: in one manufacturing case Andrew Ng described, two months of model work gained nothing, while two weeks of fixing the data lifted accuracy from 76% to 93%.
• Find the value already in your data. I propose use cases your data could support, such as forecasting, churn prediction or segmentation, and test each one with a quick model before recommending it. Opportunities come with evidence, not assumptions.

How it works. I score your data against the use you have in mind, because the same table can be ready for one task and unfit for another. The assessment covers three layers:

• Data quality on seven standard dimensions: accuracy, completeness, coverage, conformity, consistency, timeliness and uniqueness. Scores roll up from column to table to your whole data estate, and the critical data elements weigh the most.
• Task readiness for the analytics you plan: forecasting, prediction, classification, segmentation or anomaly detection. A quick test model measures how much real signal the data holds compared with a simple baseline, and I check for "leakage", where a field looks predictive only because it is recorded after the outcome.
• Document readiness for generative AI assistants: whether your documents can be extracted, are current, consistent and free of duplicates, and whether the right document is actually found when a user asks a question.

Each use case gets a clear verdict, Ready, Ready with fixes or Not feasible, with a list of fixes ranked by their impact. I work inside your environment, and your data is not sent to outside AI services.

Data management capability assessment

A structured review of how your organisation manages data, assessed against recognised good practice in eight areas:

• data strategy and the business case for data
• organisation, funding and programme management
• data definitions, ownership and authoritative sources
• data platforms and technology
• data quality management
• governance, policies and standards
• controls, lineage and day-to-day data operations
• management of analytics and AI

Data strategy and governance

I design what keeps your data fit for use once it is fixed:

• the data domains and authoritative source for each critical data element
• clear ownership of each data set
• quality metrics with a reporting routine
• policies aligned with your privacy and security rules
• a business case that ties data investment to the outcomes it enables

Related work: Digital Egypt: one data warehouse for services from 12 ministries, held on premises under strict privacy rules. MBC: a full audit of a 122-table AI platform, mapping every workflow step to the data it reads and writes.

Process automation build

I take one process, such as supplier invoices, month-end reporting or customer queries, and automate it end to end with AI agents connected to your existing systems. Fixed scope, working software at the end.

• Business and technical in one person. 22 years in enterprise technology, 12 of them in hands-on data science and analytics, and an INSEAD MBA. I can discuss strategy with your board and review designs with your engineers.
• Predictive and generative AI. Deep experience in forecasting, pricing and recommendation models, combined with modern AI agents and assistants. I choose the technique for the problem.
• Evidence before investment. I test data readiness before recommending a build, and measure every engagement against a baseline.
• Your data stays with you. I work inside your environment and on tools you approve.