What an AI-Focused Management Consultant Does at MBB and the Big Four

A management consultant helps senior leaders solve important business problems, make difficult decisions and implement organisational change.

In an AI engagement, the consultant’s job is not simply to recommend an AI model or build a chatbot. The consultant must answer a broader set of questions:

Where can AI create measurable business value, which use cases should we invest in, how should the solution operate, what technology is required, what risks must be controlled, and how do we persuade people to adopt it?

The role therefore sits between:

• Business strategy
• Process transformation
• Data and technology
• Financial analysis
• Risk and regulation
• Organisational design
• Programme delivery
• Change management

MBB firms increasingly combine strategy consultants with engineers, data scientists, designers and product managers. McKinsey does this through QuantumBlack, BCG through BCG X, and Bain through its AI, Insights and Solutions capabilities. The Big Four similarly combine business consulting with technology implementation, governance, risk, workforce transformation and managed services. (McKinsey QuantumBlack)

---

1. The fundamental job of a management consultant​

A client normally hires consultants because one or more of the following is true:

• The problem is strategically important.
• The problem crosses several departments.
• Management does not have sufficient internal expertise.
• Executives disagree about what should be done.
• The company needs an independent assessment.
• The organisation needs to move faster.
• The decision involves substantial investment or risk.
• The company needs additional delivery capacity.
• Senior management needs evidence before approving a programme.

The consultant brings structure to this ambiguity.

For example, a bank may approach a consulting firm saying:

“We need to use generative AI in customer service.”

That is not yet a well-defined problem. A consultant would convert it into a series of decision questions:

1. What business problem is the bank trying to solve?
2. Is the priority cost reduction, customer satisfaction, revenue or regulatory compliance?
3. Which customer-service journeys are suitable for AI?
4. Should AI assist employees or communicate directly with customers?
5. What data can the system access?
6. What decisions may the AI make autonomously?
7. What must be approved by a human?
8. Which models and platforms are appropriate?
9. What would the solution cost?
10. What operational benefit would it generate?
11. What regulatory and reputational risks exist?
12. How would the bank scale the solution safely?

The consultant does not begin with the technology. The consultant begins with the business decision.

---

2. The end-to-end AI consulting lifecycle​

Stage 1: Opportunity identification and business development​

Consulting often begins before the client formally launches a project.

Partners, directors and senior managers speak with client executives to understand their priorities. They may identify an opportunity through:

• A conversation with the CEO, CIO or Chief Data Officer
• A previous consulting engagement
• An industry event
• A regulatory development
• An unsuccessful AI pilot
• A request for proposal, or RFP
• A strategic partnership with a cloud or software company
• A board mandate to develop an AI strategy

The consulting team then develops a proposition.

Activities​

The consultant may:

• Research the organisation and its industry.
• Study its annual reports and strategic priorities.
• Identify major cost, revenue and risk drivers.
• Develop hypotheses about where AI could create value.
• Prepare credentials and case studies.
• Design a proposed project approach.
• Estimate the team, duration and price.
• Identify delivery partners.
• Present the proposal to the client.
• Respond to commercial and technical questions.

Deliverables​

Typical outputs include:

• Proposal deck
• Statement of work
• Project approach
• Team structure
• Delivery timeline
• Pricing model
• Assumptions and dependencies
• Initial AI opportunity map
• Relevant case studies
• Risk and quality review documentation

At manager level, consultants are often heavily involved in writing proposals, shaping the methodology, estimating resources and preparing senior leaders for client presentations.

---

Stage 2: Mobilisation​

Once the client approves the engagement, the team must establish how the work will operate.

The consultant defines:​

• The project objectives
• The scope and exclusions
• The workstreams
• The governance structure
• Decision-making responsibilities
• Meeting cadence
• Required data and system access
• Stakeholder responsibilities
• Risks, assumptions, issues and dependencies
• Quality assurance process

Common deliverables​

• Project charter
• RACI matrix
• Stakeholder map
• Integrated project plan
• RAID log
• Communication plan
• Data request
• Interview schedule
• Steering committee terms of reference
• Definition of success

A weak mobilisation can damage the entire engagement. For example, an AI pilot may fail not because of the model but because the client did not provide data access, security approvals or subject-matter experts.

---

Stage 3: Diagnose the current state​

The consultant must understand how the organisation currently operates before recommending AI.

This is sometimes called:

• Current-state assessment
• Discovery
• Diagnostic
• Baseline assessment
• As-is analysis

What consultants investigate​

Business strategy​

• What are the organisation’s priorities?
• What competitive pressures does it face?
• Where is growth expected?
• Which costs are increasing?
• What does senior leadership expect from AI?

Processes​

• How is work performed today?
• Which activities are manual?
• Where do delays occur?
• Where are errors introduced?
• Which decisions require expert judgement?
• Which tasks are repetitive but information-intensive?

Data​

• What data exists?
• Who owns it?
• Is it accurate and accessible?
• Does it contain sensitive information?
• Can it legally be used for the proposed purpose?
• Is the data sufficiently representative?

Technology​

• What cloud platforms are used?
• Which business applications must be integrated?
• What APIs are available?
• Are legacy systems involved?
• What identity, access and security controls exist?
• Is the organisation already using approved AI models?

People and organisation​

• Who performs the work?
• What skills do employees have?
• How are teams structured?
• Who owns AI decisions?
• How open are employees to AI-enabled working?

Governance and risk​

• What policies already exist?
• How are models approved?
• Who is accountable when an AI system fails?
• What regulations apply?
• How are incidents identified and escalated?

Methods used​

Consultants gather this information through:

• Executive interviews
• Employee interviews
• Workshops
• Process observation
• System demonstrations
• Data analysis
• Document reviews
• Surveys
• Benchmarking
• Customer research
• Process mining
• Architecture reviews
• Control assessments

The objective is not merely to collect information. It is to identify the root causes of poor performance.

---

Stage 4: Frame the problem​

Consultants convert the discovery findings into a structured problem statement.

A useful problem statement normally includes:

• The affected stakeholder
• The current difficulty
• The scale of the problem
• The underlying causes
• The desired outcome
• The relevant constraints

Poor problem statement​

“The bank needs a generative AI chatbot.”

Better problem statement​

“The bank’s contact centre receives approximately 2 million annual enquiries. Nearly 40% concern a limited number of repetitive information requests, but agents spend substantial time searching fragmented policy documents. This increases average handling time, creates inconsistent answers and delays complex cases. The bank needs a controlled AI assistant that helps employees retrieve approved information while preserving human accountability.”

The second statement leaves open the possibility that the correct solution is an employee assistant, not a customer-facing autonomous chatbot.

Management consultants frequently use issue trees and hypothesis-driven analysis to break one large problem into smaller questions.

For example:

---

3. AI strategy development​

An AI strategy explains how the organisation will use AI to support its wider business strategy.

It should not be a list of interesting technologies.

A serious AI strategy usually covers six areas.

3.1 Strategic ambition​

The consultant helps leadership decide what role AI should play.

Possible ambitions include:

• Improve employee productivity
• Reduce operational cost
• Increase revenue
• Improve customer experience
• Reduce risk
• Accelerate product development
• Create AI-enabled products
• Transform the organisation’s operating model
• Develop an entirely new AI-driven business

The ambition must be sufficiently specific to guide investment.

---

3.2 Where to play​

The consultant identifies where AI should be deployed.

This may involve:

• Customer service
• Sales
• Marketing
• Finance
• Human resources
• Procurement
• Supply chain
• Software engineering
• Legal operations
• Risk and compliance
• Research and development
• Internal knowledge management

The consultant evaluates whether value is concentrated in a small number of functions or distributed across the enterprise.

---

3.3 How to win​

The organisation must decide what capabilities will create an advantage.

Questions include:

• Will the company build proprietary AI?
• Will it use commercial foundation models?
• Will it rely on open-weight models?
• Does it possess unique data?
• Should AI capabilities be centralised?
• Which capabilities should be outsourced?
• Which technology alliances are strategically important?
• How quickly must it scale?

The appropriate answer depends on economics, risk, existing capabilities and competitive differentiation.

---

3.4 Capability requirements​

The consultant assesses the capabilities required across:

• Data
• Cloud infrastructure
• AI engineering
• Product management
• Cybersecurity
• Responsible AI
• Legal and compliance
• Procurement
• Change management
• Model operations
• Financial management
• Vendor management

The team then compares required capabilities with the current state and identifies the gaps.

---

3.5 Investment roadmap​

The strategy is translated into a sequenced roadmap.

A roadmap might include:

First 90 days

• Establish AI governance.
• Select priority use cases.
• Confirm approved technology patterns.
• Launch two controlled pilots.
• Develop an AI literacy programme.

Three to twelve months

• Productionise successful use cases.
• Establish reusable data and AI platforms.
• Create an AI product-management function.
• Introduce model evaluation and monitoring.
• Redesign selected business processes.

Twelve to thirty-six months

• Scale AI across functions.
• Introduce agentic workflows where justified.
• Consolidate platforms and vendors.
• Redesign roles and workforce structures.
• Launch AI-enabled products and services.

---

3.6 Value measurement​

The strategy must define how value will be measured.

Typical metrics include:

• Revenue generated
• Cost avoided
• Employee hours released
• Average handling time
• Conversion rate
• Error rate
• Customer satisfaction
• First-contact resolution
• Cycle time
• Risk loss reduction
• Model accuracy
• Adoption rate
• Percentage of AI outputs accepted by employees
• Number of incidents
• Unit cost per AI interaction

Current consulting offerings consistently emphasise measurable business impact rather than experimentation alone. Bain describes evaluating use cases through value, feasibility, risk and differentiation; KPMG emphasises business cases and value tracking; PwC describes measurable and sustained outcomes; BCG frames AI as an enterprise transformation rather than an isolated technology deployment. (BCG AI @ Scale)

---

4. AI use-case identification and prioritisation​

Most large organisations can identify hundreds of possible AI use cases. They cannot fund all of them.

The consultant creates a structured prioritisation process.

Step 1: Generate use cases​

The team conducts workshops with business functions.

For a bank contact centre, possible use cases include:

• Customer chatbot
• Agent knowledge assistant
• Automated call summarisation
• Email classification
• Complaint-routing agent
• Quality-assurance monitoring
• Vulnerable-customer identification
• Compliance checking
• Next-best-action recommendations
• Workforce forecasting

Step 2: Define each use case​

Each use case should specify:

• User
• Business problem
• Current process
• Proposed AI capability
• Required data
• Human role
• Expected benefit
• Risks
• Dependencies
• Success metrics

Step 3: Score the use cases​

A consultant may score each use case against:

Value​

• Revenue potential
• Cost reduction
• Customer impact
• Risk reduction
• Strategic importance

Feasibility​

• Data availability
• Technical complexity
• Integration complexity
• Skill availability
• Delivery time

Risk​

• Regulatory exposure
• Customer harm
• Privacy
• Bias
• Explainability
• Cybersecurity
• Reputational risk

Adoption​

• User willingness
• Process maturity
• Leadership sponsorship
• Training requirements
• Change complexity

Step 4: Create the portfolio​

Use cases are commonly classified as:

• Quick wins
• Strategic bets
• Foundational capabilities
• Experiments
• Do not pursue
• Reassess later

The consultant is expected to recommend what the organisation should not do, not merely create a long list of opportunities.

---

5. Developing the business case​

Senior executives need to understand whether an AI investment is financially justified.

The consultant develops a baseline and calculates the expected effect of the proposed change.

Example​

Suppose a contact centre has:

• 1,000 agents
• 220 working days annually
• 6 productive hours per agent per day
• An estimated fully loaded cost of $35 per hour

Annual productive hours:

$$1{,}000 \times 220 \times 6 = 1{,}320{,}000$$

Annual labour cost represented by those hours:

$$1{,}320{,}000 \times \$35 = \$46.2\text{ million}$$

Assume an AI assistant reduces time spent searching for information by 8%.

The theoretical gross capacity released would be:

$$\$46.2\text{ million} \times 0.08 = \$3.696\text{ million}$$

However, the consultant should not immediately claim a $3.696 million saving.

They must distinguish among:

• Time released: Employees have more available time.
• Productivity improvement: More work is completed with the same workforce.
• Cost avoidance: Future hiring is prevented.
• Cashable saving: Actual expenditure is removed.
• Revenue benefit: Employees use the released capacity to generate revenue.

The consultant then adds:

• Implementation cost
• Cloud and model cost
• Software licensing
• Integration cost
• Support and maintenance
• Training
• Governance
• Security
• Contingency
• Ongoing product team cost

The business case should include:

• Base, upside and downside scenarios
• Benefit timing
• Adoption assumptions
• Cost assumptions
• Net present value
• Payback period
• Sensitivity analysis
• Risks to value realisation

This is where a management consultant adds value beyond an engineer: the consultant translates technical performance into an economically credible investment decision.

---

6. Designing the target operating model​

Deploying AI changes how an organisation operates. A target operating model defines how the future organisation will work.

Components of an AI operating model​

Governance​

• Who approves AI use cases?
• Who accepts the residual risk?
• Who can stop a system?
• Who owns the model?
• Who owns the business outcome?
• Who investigates incidents?

Organisation​

• Central AI centre of excellence
• Federated business AI teams
• Embedded product teams
• Responsible AI function
• AI platform team
• Data ownership model

Roles​

• Executive sponsor
• AI product owner
• Business process owner
• Data owner
• Solution architect
• AI engineer
• Model-risk specialist
• Security lead
• Legal and privacy adviser
• Change lead
• Service manager

Processes​

• Use-case intake
• Prioritisation
• Model selection
• Development
• Testing
• Risk assessment
• Approval
• Deployment
• Monitoring
• Incident management
• Retirement

Technology​

• Approved model catalogue
• AI gateway
• Data platform
• Evaluation platform
• Observability
• Identity and access management
• Prompt and model registry
• Deployment pipeline
• Cost monitoring

Performance management​

• Business KPIs
• Model KPIs
• Risk indicators
• Adoption measures
• Financial benefits
• Service-level objectives

The consultant may create the design and then help the organisation implement it.

---

7. Solution shaping and architecture​

An AI-focused management consultant does not necessarily write all the code. However, they must understand the solution sufficiently to challenge technical decisions and connect architecture to business requirements.

Typical architecture decisions​

• Predictive model or generative AI
• Single agent or multi-agent workflow
• Retrieval-augmented generation or model fine-tuning
• Cloud API or self-hosted model
• Open-weight or proprietary model
• Real-time or batch processing
• Human approval or autonomous execution
• Central or decentralised deployment
• Shared enterprise platform or use-case-specific solution

Questions the consultant helps answer​

• Does the use case genuinely require an LLM?
• What level of accuracy is acceptable?
• What latency is required?
• Which data sources must be connected?
• What information may the model retain?
• How will confidential data be protected?
• How will users authenticate?
• How will outputs be evaluated?
• What happens when the model is uncertain?
• How will the system be monitored?
• How will the cost per transaction be controlled?
• Can the proposed design scale?

The detailed design is usually produced collaboratively by:

• Management consultants
• AI solution architects
• Data architects
• Product managers
• AI engineers
• Security specialists
• Risk specialists
• User-experience designers
• Client subject-matter experts

EY describes its AI services as combining strategy, design, architecture, data, systems integration, programme operations and risk. Deloitte similarly describes supporting clients from AI strategy through bespoke solution delivery, while Bain highlights integrated teams containing consultants, product managers and engineers. (Bain AI Consulting)

---

8. Building and testing AI solutions​

Some consulting engagements end with a strategy. Increasingly, however, consulting firms are also expected to build and deploy solutions.

McKinsey’s QuantumBlack combines strategy and domain expertise with AI delivery. BCG X describes itself as a technology build and design unit containing technologists, scientists, programmers, engineers and designers. Bain describes end-to-end capabilities across design, engineering, product management, data science and machine-learning engineering. (McKinsey QuantumBlack)

During delivery, the management consultant may:

• Translate business requirements into product features.
• Maintain the product backlog.
• Define acceptance criteria.
• Coordinate engineers and business users.
• Organise design workshops.
• Track dependencies.
• Escalate delivery risks.
• Prepare demonstrations.
• Manage user testing.
• Coordinate security and legal reviews.
• Document decisions.
• Report progress to executives.
• Ensure the product is creating the expected value.

AI testing areas​

An AI solution requires more than conventional software testing.

The project may test:

• Answer correctness
• Groundedness
• Relevance
• Hallucination rate
• Retrieval quality
• Tool-selection accuracy
• Task-completion rate
• Bias
• Toxicity
• Data leakage
• Prompt injection resistance
• Human escalation
• Latency
• Availability
• Cost per transaction
• Behaviour under unusual inputs
• Performance across customer groups

The consultant ensures that technical evaluation is connected to business requirements.

For example:

“The model achieved an 86% score” is not sufficient.

The consultant should ask:

• Which test set was used?
• Does it represent production traffic?
• What types of error account for the remaining 14%?
• Could those errors cause customer harm?
• Is human review available?
• How does performance compare with employees?
• What is the commercial effect of the errors?
• Who approved the acceptance threshold?

---

9. Responsible AI, risk and governance​

AI consultants must consider risk throughout the lifecycle rather than adding governance immediately before deployment.

Important risk categories​

Accuracy and reliability​

• Hallucinated answers
• Incorrect classifications
• Unstable behaviour
• Poor generalisation
• Inadequate testing

Fairness​

• Discrimination
• Unequal error rates
• Under-representation
• Proxy variables

Privacy​

• Personal-data leakage
• Inappropriate data use
• Excessive retention
• Lack of lawful basis

Security​

• Prompt injection
• Data exfiltration
• Poisoned knowledge sources
• Insecure tools
• Excessive system permissions

Transparency​

• Users not knowing they are interacting with AI
• Inadequate explanation
• Poor documentation
• Lack of traceability

Human accountability​

• Unclear decision ownership
• Automation bias
• Inadequate escalation
• Removal of meaningful human review

Third-party risk​

• Model-provider dependency
• Data-location concerns
• Unclear contractual rights
• Service discontinuation
• Model changes outside the client’s control

Operational risk​

• Poor monitoring
• No rollback mechanism
• Inadequate incident response
• Uncontrolled model or prompt changes

Big Four practices are particularly prominent in linking AI transformation to controls, regulation, assurance and technology risk. For example, PwC describes Responsible AI controls and governance across the AI lifecycle; KPMG emphasises trusted AI, risk and regulatory expectations; EY includes “trust” as one of its AI transformation domains. (PwC AI)

A PwC UK Responsible AI manager role, for example, describes work at the intersection of emerging technology, risk management and regulation, including operationalising AI governance and embedding trust by design across generative and agentic AI lifecycles. (PwC UK careers)

---

10. Change management and adoption​

A technically successful AI system can still fail if employees do not use it.

Management consultants therefore work on the human side of AI transformation.

Questions addressed​

• Which roles will change?
• Which tasks will be automated?
• Which tasks will be augmented?
• What new skills are needed?
• How will employees be trained?
• How should performance expectations change?
• What concerns will employees have?
• How will the organisation build trust?
• What incentives encourage adoption?
• How will feedback reach the product team?

Common interventions​

• Leadership alignment
• Stakeholder engagement
• Employee communications
• AI literacy programmes
• Role-based training
• Champion networks
• Updated standard operating procedures
• Revised performance metrics
• User-support channels
• Feedback mechanisms
• Adoption dashboards
• Change-impact assessments

The consultant must distinguish between:

• Technical deployment: The software is available.
• Adoption: Employees use it.
• Behaviour change: Employees change how work is performed.
• Value realisation: The changed behaviour produces measurable outcomes.

BCG, Bain, PwC and KPMG all explicitly position workforce capability, operating-model change or adoption as central to scaling AI rather than as secondary activities. (BCG AI @ Scale)

---

11. Programme management and implementation​

Large AI transformations may contain dozens of use cases, multiple vendors, cloud programmes, data migrations and regulatory approvals.

Consultants establish the mechanisms needed to coordinate them.

Typical programme-management activities​

• Maintain the integrated delivery plan.
• Track milestones and dependencies.
• Manage the budget.
• Monitor resource requirements.
• Coordinate workstreams.
• Maintain the RAID log.
• Prepare steering committee reports.
• Manage scope changes.
• Track executive decisions.
• Escalate blockers.
• Coordinate vendors.
• Monitor quality.
• Track benefits.
• Ensure operational handover.

A consultant should not act merely as an administrative project manager. They must understand the content well enough to challenge workstream leaders.

For example:

“The data pipeline is delayed” is only a status update.

A strong consultant asks:

• Which use cases depend on the pipeline?
• What caused the delay?
• Can the dependency be redesigned?
• What decision is required?
• What is the effect on the critical path?
• Does this change the business case?
• Who has authority to resolve it?

---

12. Executive communication​

A major part of consulting is communicating complicated findings clearly.

Senior executives normally do not want a detailed explanation of embeddings, vector databases or attention mechanisms. They want to understand:

• What decision must be made?
• Why is it needed now?
• What value is available?
• What will it cost?
• What could go wrong?
• What alternatives exist?
• What do you recommend?
• What happens next?

Example of poor communication​

“We recommend implementing GraphRAG using a hybrid vector and graph retrieval layer with agentic orchestration.”

Better executive communication​

“Customer-policy information is fragmented across five systems, causing agents to search manually and provide inconsistent answers. We recommend a controlled employee assistant that retrieves approved information and cites its source. It could reduce search time while maintaining human accountability. A twelve-week pilot would test accuracy, employee adoption, security and financial value before wider deployment.”

The technical details remain important, but they belong in the supporting material.

A consultant must be able to communicate the same solution differently to:

• CEO
• CFO
• CIO
• CISO
• Legal team
• Risk committee
• Operations director
• Data scientist
• Software engineer
• Front-line employee

---

13. What consultants produce​

Consulting work often becomes visible through presentations, but “making slides” is not the real purpose. Slides are the medium through which evidence, decisions and recommendations are communicated.

Common AI consulting deliverables​

Strategy and value​

• Enterprise AI strategy
• AI ambition and vision
• Use-case portfolio
• Prioritisation model
• Business case
• Investment roadmap
• Competitor benchmark
• Executive decision paper

Operating model​

• AI governance model
• Organisation design
• Centre-of-excellence design
• RACI
• Process model
• Capability assessment
• Skills strategy
• Workforce plan

Technology​

• Target architecture
• Platform strategy
• Model-selection framework
• Data architecture
• Integration design
• Security architecture
• Build-versus-buy assessment
• Vendor evaluation

Responsible AI​

• AI policy
• Risk-classification framework
• Control library
• Impact assessment
• Model documentation
• Evaluation framework
• Approval process
• Monitoring and incident process

Delivery​

• Project charter
• Product roadmap
• Backlog
• RAID log
• Test strategy
• Deployment plan
• Change plan
• Benefits-realisation dashboard

Commercial and procurement​

• Request for proposal
• Vendor scorecard
• Total-cost-of-ownership model
• Contract requirements
• Service-level requirements
• Exit strategy

---

14. What a consultant does each day​

The daily work depends on the project phase.

Example day for an AI consultant​

8:30 — Internal team check-in​

The team discusses:

• What was completed yesterday
• Important findings
• Client meetings
• Delivery risks
• Required decisions
• Priorities for the day

9:00 — Client process interview​

The consultant interviews contact-centre managers to understand:

• How customer enquiries are routed
• Where agents search for information
• Which cases require escalation
• Which errors are most damaging
• How performance is measured

10:30 — Analysis​

The consultant may:

• Analyse contact-volume data
• Review process documentation
• Calculate average handling time
• Segment queries
• Assess the financial baseline
• Compare performance across teams

12:00 — Working session with engineers​

The consultant reviews:

• Data availability
• Integration requirements
• Proposed architecture
• Security constraints
• Testing requirements
• Estimated model cost

13:30 — Client workshop​

The team facilitates a use-case prioritisation workshop with business, technology, legal and risk stakeholders.

15:00 — Synthesis​

The consultant converts the findings into:

• Key messages
• Supporting evidence
• Implications
• Recommendations
• Decisions required

17:00 — Manager or partner review​

Senior colleagues challenge the work:

• Is the recommendation supported by evidence?
• What is the commercial value?
• What could the client disagree with?
• Is the message clear?
• What decision do we need from the executive?

18:00 — Revision and planning​

The consultant incorporates feedback, prepares the next meeting and assigns follow-up actions.

In practice, the work involves a mixture of meetings, analysis, problem solving, writing, reviewing and coordinating.

---

15. Responsibilities by level​

Titles differ by firm, but the progression is broadly similar.

Analyst or business analyst​

The analyst performs much of the detailed research and analysis.

Responsibilities include:

• Collecting data
• Conducting interviews
• Building financial models
• Analysing processes
• Researching competitors
• Preparing slides
• Testing hypotheses
• Documenting findings
• Supporting workshops

At McKinsey, business analysts and associates typically own defined portions of the analysis or workstream within a client-service team. (McKinsey consulting roles)

Consultant or associate​

The consultant owns a defined workstream.

For example:

• AI use-case prioritisation
• Business-case development
• Responsible AI design
• Customer-service process redesign
• Technology-vendor assessment

The consultant is expected to manage their own analysis, interact directly with client stakeholders and present findings.

Manager, engagement manager or project leader​

The manager is responsible for the integrated project.

They:

• Convert the problem into a workplan.
• Allocate work across the team.
• Maintain the client relationship.
• Review analyses and deliverables.
• Resolve conflicts between workstreams.
• Coach junior consultants.
• Manage deadlines and quality.
• Prepare executive meetings.
• Coordinate specialists.
• Control scope and economics.
• Identify follow-on opportunities.

This role requires less individual analysis than a junior role but substantially more judgement, synthesis and stakeholder management.

Director, principal or associate partner​

At this level, the consultant typically:

• Oversees several projects or a large programme.
• Maintains senior client relationships.
• Shapes major recommendations.
• Brings industry or functional expertise.
• Ensures project quality.
• Supports sales.
• Develops the consulting team.
• Identifies future opportunities.

McKinsey describes associate partners as owning the overall delivery of multiple client projects and ensuring the quality of the work. (McKinsey consulting roles)

Partner​

The partner is responsible for:

• Building trusted executive relationships
• Winning work
• Setting the intellectual direction
• Challenging the client’s thinking
• Ensuring quality and risk management
• Resolving major problems
• Developing the firm’s market position
• Owning the commercial relationship

Partners may not attend every working meeting, but they are accountable for the engagement and the client relationship.

---

16. MBB versus Big Four​

The following distinctions are tendencies, not absolute rules. All seven firms now operate across strategy, technology, implementation and transformation.

Dimension	MBB	Big Four
Traditional strength	Corporate strategy and senior executive decisions	Broad business transformation, technology, risk, finance and implementation
Typical entry point	CEO, business-unit leader, strategy leader	CIO, CFO, COO, CDO, risk, technology or functional leaders
Project style	Often hypothesis-led, strategically concentrated	Often multidisciplinary and implementation-heavy
Team structure	Frequently smaller core consulting teams supported by specialists	Frequently larger teams spanning business, technology, risk and operations
AI focus	Competitive strategy, enterprise reinvention, value pools, business-model change and increasingly solution building	Strategy, process transformation, platforms, systems integration, governance, controls and managed operations
Implementation	Growing rapidly through specialist build units	Historically a major component of the model
Risk and controls	Included, especially in regulated engagements	Often deeply integrated due to strong risk, assurance and regulatory practices
Engagement duration	Can be relatively focused, although major transformations may be long	Frequently includes large multiyear implementation programmes
Commercial scale	Smaller strategic teams can command high fees	Larger programmes may involve substantial delivery teams

MBB AI model​

McKinsey​

McKinsey combines management consulting with QuantumBlack, its AI consulting arm. Its positioning emphasises strategic thinking, domain expertise, data, technology and real-world implementation. (McKinsey QuantumBlack)

BCG​

BCG combines traditional consulting with BCG X, its technology build and design division. BCG X brings together engineers, scientists, designers and entrepreneurs to develop and scale AI-enabled products, services and businesses. (BCG AI @ Scale)

Bain​

Bain combines business strategy with AI product management, engineering, data science and ecosystem partnerships. Its AI approach explicitly covers strategy, use-case selection, technology stacks, operating models, talent and change management. (Bain AI Consulting)

Big Four AI model​

PwC​

PwC’s AI work includes strategy, workflow and application implementation, integrated technology solutions, security, responsible AI, governance and business-function transformation. PwC’s current positioning emphasises purpose, trust and measurable business outcomes. (PwC AI)

Deloitte​

Deloitte combines AI strategy with analytics, data modernisation, cloud platforms, intelligent automation, systems and operational transformation. (Deloitte AI & Data)

EY​

EY combines strategy, architecture, data, systems integration, programme operations, intelligent automation and risk. Its AI framework emphasises insight, performance, automation, experience and trust. (EY AI Consulting)

KPMG​

KPMG’s AI positioning centres on value identification, trusted AI, workforce transformation, data foundations, alliances and scaling solutions into core operations. (KPMG AI)

---

17. Management consultant versus AI engineer​

These roles overlap, but their primary accountabilities differ.

Management consultant	AI engineer
Defines the business problem	Builds the technical solution
Develops the business case	Develops models, services and pipelines
Facilitates executive decisions	Implements technical requirements
Designs the operating model	Designs or implements technical components
Prioritises use cases	Optimises solution performance
Coordinates stakeholders	Integrates systems
Manages transformation	Maintains engineering quality
Develops change strategy	Creates tests and deployment pipelines
Tracks business value	Tracks technical performance

A strong AI programme requires both.

The management consultant should understand enough technology to avoid recommending an unrealistic solution. The engineer should understand enough business context to avoid building something that does not solve the actual problem.

---

18. The AI solution-engineering consultant​

A role such as AI Solution Engineer, AI Solution Architect or AI Transformation Manager is often positioned between management consulting and engineering.

The person may be responsible for:

1. Conducting client discovery.
2. Identifying high-value AI opportunities.
3. Translating requirements into solution concepts.
4. Selecting appropriate architecture patterns.
5. Developing prototypes or directing prototype delivery.
6. Estimating cost, effort and performance.
7. Incorporating security and Responsible AI.
8. Preparing proposals and executive presentations.
9. Coordinating engineers, cloud providers and client teams.
10. Supporting implementation and handover.
11. Developing reusable assets.
12. Converting one project’s lessons into wider consulting propositions.

Current AI consulting vacancies illustrate this hybrid model. EY roles combine consulting, business analysis and AI understanding with collaboration across engineering teams, while PwC roles include embedding engineers within consulting teams to design and deploy AI solutions and convert recurring delivery patterns into reusable capabilities. (EY Careers)

---

19. Worked example: Banking customer-service AI​

Consider a bank with high contact-centre costs and inconsistent customer answers.

Phase 1: Diagnose​

The consulting team discovers:

• Customers frequently call about routine policy questions.
• Agents search across several systems.
• Knowledge articles are inconsistent.
• After-call documentation is manual.
• Complex complaints are not always escalated correctly.
• Management lacks reliable information about why customers are calling.

Phase 2: Identify options​

The team considers:

1. Customer-facing chatbot
2. Agent knowledge assistant
3. Automated call summarisation
4. Complaint classification
5. Real-time compliance monitoring
6. Next-best-action recommendations

Phase 3: Prioritise​

The customer chatbot is potentially valuable but high risk because it communicates directly with customers.

The agent assistant is selected first because:

• Employees remain accountable.
• The AI can cite approved sources.
• The bank can test the solution in a controlled environment.
• Errors can be reviewed before reaching customers.
• Adoption and performance can be measured.

Phase 4: Design the solution​

The proposed solution includes:

• Retrieval from approved policy documents
• Source citations
• Role-based access
• Personal-data controls
• Prompt-injection protection
• User feedback
• Human escalation
• Audit logging
• Answer-quality evaluation
• Cost and latency monitoring

Phase 5: Define the pilot​

The pilot is limited to:

• 100 employees
• Three enquiry categories
• Approved internal knowledge
• Eight weeks of live testing
• Human review of all recommendations

Phase 6: Set success criteria​

Business​

• Reduced search time
• Lower handling time
• Higher first-contact resolution
• Improved employee satisfaction

Technical​

• Retrieval accuracy
• Groundedness
• Response latency
• Service availability

Risk​

• No critical data leakage
• Acceptable error severity
• Effective escalation
• Complete audit records

Adoption​

• Weekly active users
• Recommendation acceptance rate
• Employee feedback
• Training completion

Phase 7: Scale decision​

At the end of the pilot, the consultant prepares an executive recommendation:

• Scale
• Scale with conditions
• Redesign and retest
• Stop the programme

The recommendation combines financial performance, technical results, risk findings and employee adoption.

This is the essence of AI management consulting: turning an ambiguous idea into an evidence-based, executable and governable business decision.

For a full worked banking series covering strategy, commercial case, architecture, Responsible AI and delivery, see the Banking Customer-Service AI playbook.

---

20. Skills required​

Structured problem solving​

The consultant must break complicated problems into manageable components and identify the analyses required to reach a decision.

Business and financial understanding​

They must understand:

• Revenue
• Cost
• Margin
• Capital investment
• Operating expenditure
• Productivity
• Return on investment
• Risk-adjusted value

AI and technology fluency​

An AI consultant should understand:

• Machine learning
• Generative AI
• Agentic systems
• RAG
• Model evaluation
• Data architecture
• APIs
• Cloud platforms
• Security
• MLOps and LLMOps
• Model economics

Not every consultant must be an expert programmer, but they must be able to identify weak technical assumptions.

Communication​

They must:

• Interview stakeholders
• Facilitate workshops
• Write clearly
• Present recommendations
• Handle objections
• Communicate bad news
• Adapt communication to different audiences

Stakeholder management​

AI programmes frequently create disagreement between:

• Business and technology
• Innovation and risk
• Central and local teams
• Employees and management
• Internal teams and vendors

The consultant must create alignment without ignoring legitimate disagreement.

Delivery leadership​

The consultant must understand:

• Project planning
• Product management
• Agile delivery
• Dependency management
• Risk management
• Quality assurance
• Benefits realisation

Commercial judgement​

Senior consultants need to understand:

• How consulting work is sold
• How projects are priced
• How scope is controlled
• How delivery affects profitability
• How new opportunities are identified
• How client trust is maintained

---

21. What makes an excellent AI management consultant​

An average consultant can produce a polished presentation.

An excellent consultant can:

• Identify the real problem behind the client’s request.
• Separate attractive technology from genuine business value.
• Challenge unsupported assumptions.
• Translate technical issues into executive decisions.
• Translate business needs into implementable requirements.
• Balance innovation with risk.
• Distinguish a pilot from a production solution.
• Create recommendations that employees can actually implement.
• Make uncertainty visible.
• Explain what should not be automated.
• Demonstrate measurable value.
• Earn trust across business, technology, engineering and risk teams.

The most important mindset is:

Do not ask only, “What can AI do?” Ask, “What outcome does the organisation need, what must change to achieve it, and where does AI provide a defensible advantage?”

---

Sources​

• McKinsey QuantumBlack
• BCG AI @ Scale
• Bain AI Consulting
• PwC AI
• PwC UK Responsible AI Consultant role
• McKinsey consulting roles
• Deloitte AI & Data
• EY AI Consulting
• KPMG AI
• EY AI & Data Technology Consulting careers