Expert case by

Florian

ClaimSmart: AI for Next-Gen Health Insurance

Case information

Difficulty: Advanced

Interviewer-led

< 100 Ratings

Times solved: 2.1k

Digital transformation Non-conventional Operations strategy

Case Prompt:

Our client is a mid-sized health insurance provider called SureHealth with annual revenues of $2.4 billion. The company has experienced a steady rise in administrative claims processing costs and has seen a recent uptick in fraudulent claims. Senior leadership believes that leveraging AI in the claims assessment process could have significant benefits.

They have allocated a budget of $10 million over the next three years to implement AI solutions, but they are uncertain how to prioritize to reduce fraudulent claims and administrative processing costs.

Overview of All Exhibits

Practice makes the difference

Practicing alone helps – with a partner it’s even better. Solve this case in a realistic mock interview.

Instructions --> Please read before going through the case

This case is a typical McKinsey-style case in terms of design, difficulty, and evaluation.
However, it is also well-suited for candidate-led practice. If you are the interviewer, simply start with the prompt, then stop providing guiding questions unless the candidate gets stuck.
The case should be completed within 25 minutes. As the interviewer, ensure the discussion stays on track to meet this timeframe and provide guidance and shortcuts if necessary.
If you are the interviewer, make sure to work through the case yourself beforehand to ensure you’re fully prepared to guide the candidate.
The case includes very detailed interviewer notes. Please reach out if you want to use a real MBB feedback sheet with some explanations to score the performance realistically.

Further Case Information (for follow-up questions)

If the candidate asks for clarification of the objective and more details on the business model, provide the following information:

Clarify objective: Invest $10 million into AI to generate a high return in terms of reducing claims processing cost and fraudulent claims
Learn more about the business model of the company:
- Customers: Primarily small-to-medium businesses (employer-sponsored plans), plus some individual/family policies.
- Revenue and cost model:
  - Revenue: Monthly premiums (main driver) and administrative fees for self-insured employer plans.
  - Cost: Medical claims (largest expense) and administrative overhead (claims processing, compliance, etc.).
- Distribution channels: Direct online/phone sales and broker networks for employer groups.
- Competitive edge: Emphasis on flexible coverage options, broad provider networks, and customer-centric service.

Q1 - Brainstorming

You arrive at the client site on your first day with the new engagement, get your badge at the reception desk, then enter the elevator to find your team room high up in the building.

A senior employee enters the elevator, looks at your badge and says: “Ah, you must be from the consulting team – my name is Josh, I am the CTO here.” Your pulse goes up and you introduce yourself. “Pleasure to meet you” he replies and immediately proceeds to ask you a question: “Looking forward to the collaboration. Well, actually, what do you think? Considering the entire claims lifecycle – from initial submission to payout - where do you believe AI can have the greatest impact on driving efficiency and reducing fraud?”

In a nutshell, you are asked to brainstorm ideas on AI use cases in the client’s context.

Show solution Hide solution

Note for Interviewer

The candidate should play back the question, then take 1.5-2 minutes to draft their brainstorming answer.

When evaluating a candidate’s response, the interviewer should ensure it meets several criteria:

Content coverage
- Broadness: The answer should address all key aspects of the question at the top level, providing a thorough first-principles approach.
- Depth: It should go 2–3 layers deeper with specific examples at the lowest level.
- Insightful: The ideas need to be tied directly to the case scenario and should consist of a mix of common sense as well as creative ideas.
- Forward-thinking: Evaluate how the candidate drives the case forward by considering a prioritization of their ideas and next steps based on their answer (this is also relevant in an interviewer-led case).
- Keep in mind that there is never just one way to approach a brainstorming question in a case interview. Interviewers evaluate broadness, depth, and insightfulness and many different answers would be able to achieve that. As an example, I provide two potential brainstorming answers based on a process view and a component perspective, which would be equally well-received. There are other potential answers that would score equally well (for instance by breaking down the different stakeholders, etc.)
Communication style
- Structured approach: The candidate should start with an overview by presenting the top-level categories first (signposting), then move step by step into each top level category, explaining the answer in a logical sequence. All ideas need to be numbered.
- Detail and reasoning: Make sure the candidate clarifies why they think what they think, as interviews also focus on the reasoning process besides the actual content.
- Confidence and professionalism: Evaluate if the candidate presents insights in a way that sounds knowledgeable and assured.

If the candidate’s answer doesn’t seem wide-ranging or detailed enough, prompt them with a question such as, “Is there anything else you might be missing?” and offer gentle guidance to help them expand their thinking.

Check how coachable they are.

PROCESS APPROACH (looking at the underlying process of the problem)

1. Claims submission and initial review: How claims enter the system and how they’re prioritized.

Automatic data capture: AI can read and extract important information from documents (like medical bills or claim forms).
Sorting and prioritizing: AI can label or route claims based on urgency or complexity, so serious claims get immediate attention.

2. Finding and preventing fraud: Identifying suspicious activity and stopping fraudulent payouts.

Spotting unusual patterns: AI looks for claims that deviate from normal patterns, such as the same address repeatedly submitting high-cost procedures.
Mapping connections: AI can detect suspicious links between claimants, healthcare providers, or repeated diagnoses by analyzing how they’re connected.

3. Approving and paying out claims: Checking if claims are valid according to the policy and processing payments more efficiently.

Matching claims to policies: AI can quickly check if a claim is covered by a particular insurance plan.
Streamlining paperwork: Repetitive tasks (like updating claim status or confirming codes) can be automated, freeing people to focus on more complex cases.

4. Customer support and communication: Using AI to help policyholders get answers quickly and stay informed.

Chatbots and helpdesks: Customers can ask questions about their policy or claim status and get quick responses from AI-powered assistants.
Personalized messaging: Based on past behavior, AI can send reminders or updates before customers even ask.

5. Ongoing analysis and compliance: Using data to continually make the claims process more efficient and effective.

Predicting costs and trends: AI can analyze large amounts of data to guess how many claims might come in certain months (flu season, for example).
Learning from past outcomes: AI tracks which claims are denied or flagged for fraud and updates its models so it gets smarter over time.
AI governance: Set clear policies and meet legal requirements for data usage, privacy, and ethical considerations around AI-driven decisions.

COMPONENT PERSPECTIVE (looking at the relevant components of the problem)

1. Improve staff support:

Optimized decision making: By integrating AI at the front end, insurers can auto-approve low-risk, low-cost claims while routing complex ones to specialized teams.
AI support for claims teams: Administrative staff can rely on AI tools to quickly extract data from claims and route them based on complexity or urgency.
Streamlined processes: Process automation can handle repetitive tasks (e.g., updating claim status in the system), reducing manual errors and freeing up staff time.
Fraud detection boost: Investigators receive automated alerts on suspicious claims, allowing them to focus on high-risk cases rather than sifting through every submission.
Chatbots for customer sales reps: AI-driven chat interfaces can handle common questions about coverage or claim status, reserving human reps for escalations.

2. Enhance customer experience

Proactive updates: Tailored automated emails or texts can notify customers about their claim’s progress, required documents, or next steps.
Faster resolution for simple claims: Straightforward claims (like routine check-ups) can be rapidly processed, enhancing customer satisfaction and lowering costs. Implementing such a system should generally decrease processing time, also for more complex claims, providing customers with faster payouts and quicker help.

3. Update technology and data analytics

Forecasting: AI can analyze past data and external factors (e.g., flu season, regional events) to predict upcoming claim surges, helping with resource planning.
Cost modeling: Advanced algorithms estimate potential payouts, guiding finance teams to set adequate reserves.
Continuous learning: Each newly flagged fraudulent claim further refines the fraud model, making it smarter over time.
Performance metrics & dashboards: Analytics teams track AI success rates, operational efficiencies, and identify areas needing further optimization.

4. Implement strategy and governance

AI governance: Set clear policies and meet legal requirements for data usage, privacy, and ethical considerations around AI-driven decisions.
Budget allocation: Direct funds toward the most impactful AI initiatives (e.g., anti-fraud systems, customer experience tools).
Long-term vision: Leadership uses AI-driven insights to decide on product enhancements, market expansion, or partnerships that improve competitive advantage.

As a next step and to drive the case forward, think about what is logical and sensible in the context of the case: since fraud seems to be a major issue, it would be wise to prioritize fraud detection efforts.

Q2 - Framework

Once the whole team has arrived and settled into their team room, they begin their first problem-solving session of the day in preparation for the upcoming client meeting after lunch. At this point, the Engagement Manager poses the following question:

“What key factors should SureHealth analyze to determine how to prioritize AI investments in order to reduce administrative claims processing costs and minimize fraudulent payouts?”

Show solution Hide solution

Note for Interviewer

The candidate should play back the question, then take 1.5-2 minutes to draft their analytical framework.

When evaluating a candidate’s response, the interviewer should ensure it meets several criteria:

Content coverage
- Broadness: The answer should address all key areas of the problem at the top level, providing a thorough first-principles approach.
- Depth: It should go 2–3 layers deeper with specific areas to investigate at the lowest level.
- Insightful: The categories and areas to analyze need to be tied directly to the case scenario and should consist of a mix of common sense as well as creative areas.
- Forward-thinking: Evaluate how the candidate drives the case forward by considering a prioritization of their analysis and next steps based on their answer (this is relevant even in an interviewer-led case).
- Keep in mind that there is never just one way to approach a framework in a case interview. Interviewers evaluate broadness, depth, and insightfulness and many different answers would be able to achieve that.
Communication style
- Structured approach: The candidate should start with an overview by presenting the top-level categories first (signposting), then move step by step into each top level category, explaining the framework in a logical sequence. All parts of the framework need to be numbered.
- Detail and reasoning: Make sure the candidate clarifies why they think what they think, as interviews also focus on the reasoning process besides the actual content.
- Confidence and professionalism: Evaluate if the candidate presents their planned analysis in a way that sounds knowledgeable and assured.

If the candidate’s framework doesn’t seem wide-ranging or detailed enough, prompt them with a question such as, “Is there anything else you might be missing?” and offer gentle guidance to help them expand their thinking.

Check how coachable they are.

1. Fraud risk and process issues

1.1 Current fraud patterns

What is the current number and percentage of frauds?
Where does it happen the most and which areas of the claims process are most susceptible to fraudulent activity?
What is the business impact on costs (e.g., illegitimate payouts) ?
What is the impact on operations (e.g., staff hours spent)?

1.2 Process bottlenecks and weak links

How are claims currently segmented and processed?
What is the average claim handling time and staff involvement (also comparing it to industry benchmarks)?
Where in the claims lifecycle are manual reviews or outdated checks causing inefficiencies?

2. Available tools and ROI

2.1 Current tools

What digital tools are we using today for fraud detection and claims management (e.g., are we already using AI)?
What are the issues with these current tools (e.g., integration, reliability)?

2.2 Available tools on the market

What are the current leading edge applications on the market?
How are competitors and other similar industries using AI tools that we could implement?
How would the available tools fit our organization and needs?

2.3 ROI of implementation

What will it cost to set up, integrate, and maintain AI-driven tools and workflows?
How quickly can AI reduce manual processing or lower fraud payouts to offset costs? What will be the annualized cost-savings impact?
Will the impact be realized immediately or over time and will the investment break even in the short term or require a longer horizon?

3. Technology readiness

3.1 Data

Does the client have consistent, high-quality data to train AI models effectively?

3.2 Infrastructure

Are there adequate current technology platforms to support AI deployment?
Do other technological bottlenecks or issues need to be addressed first?

4. Organizational readiness and compliance

4.1 Workforce skills

Do the teams have the necessary data and AI competencies?
What training is required?
Is outside hiring required?

4.2 Adoption and cultural acceptance

Will staff readily adopt AI tools?
What are potential barriers (e.g., fear of automation replacing jobs)?

4.3 Regulations

Will implementing AI require new compliance measures?
What areas will be affected by it (e.g., personal data, insurance rules)?

4.4 Ethical practices

What guidelines and checks are needed to ensure AI decisions (e.g., claim denials) are transparent and fair?

Bonus. Strategic alignment (not mission critical but provides effective additional value and considerations to the client within the context of the case)

Bonus.1. Current priorities and alternatives

Are administrative cost reduction and fraud prevention the top strategic goals?
Do they compete with other initiatives where AI could contribute more value?

Bonus.2 Impact on overall strategy

How can AI solutions support or enhance the client’s long-term vision and market positioning?
How can AI solutions enhance customer experience and satisfaction?

As a next step and to drive the case forward, think about what analysis is logical and sensible in the context of the case: since you are at the start of the project, you should first get an understanding of the problem and its magnitude (the what) before understanding the underlying issues that cause it (the why) and then working on the solutions (the how).

So first, it would be sensible to propose to investigate the category labelled fraud risk and process issues.

Q3 - Exhibit Interpretation

The engagement has progressed and several potential AI use cases have been identified.

Provided with the data from the chart, which AI initiatives should the client prioritize over the next three years to minimize fraud and reduce claims processing costs, given their $10 million budget and implementation constraints?

Show additional information Hide additional information

Show solution Hide solution

Note for Interviewer

The candidate should play back the question, briefly describe the exhibit (provide a high-level overview without going deeper into the data), ask clarification questions (if needed), then take 0.5-1 minutes to analyze the exhibit.

When assessing a candidate’s answer and analysis of any supporting exhibits, you should look for the following:

Substance of the content
- Do they identify the key insights – typically two to four critical data points or anomalies/outliers – that are relevant within the context of the case?
- Do they interpret what those findings mean for the case at hand and propose how to act on them? (the “so what?” or synthesis)
- Do they outline concrete next steps to move the case forward?
Style of communication
- Is the answer presented in a structured, top-down format, clearly signposted and easy to follow?
- Do they explain their reasoning in enough detail so you understand why they formed their conclusions?
- Do they speak with clarity, confidence, and professionalism?

These criteria help ensure that the candidate not only understands the data but can also convey its significance in the case context and propose practical actions.

If the answer misses relevant insights, fails to provide implications, or omits next steps, consider asking: “Is there anything else you have not considered here?, What does this mean in the context of the case?, What next?”

Provide guidance if needed. Check how coachable they are.

Insights:

Automated claim review and fraud detection offer the highest cost savings over the three-year period.
The impact of these initiatives is not immediate, varies early on, and requires 1.5 to 2 years for full implementation and benefits realization.
Both initiatives are challenging to implement for different reasons such as training requirements and regulatory considerations, consistent with their longer timeframes for full readiness and potentially high impact.

Implications:

With a $10 million budget and a three-year timeframe, the client should balance quick wins with long-term high-impact initiatives by prioritizing automated claim review and fraud detection:

Automated claim review can deliver substantial quick wins starting in the first year.
Fraud detection should begin in parallel, as it offers the greatest overall impact despite a slower ramp-up.

Next steps:

To finalize the decision, the following aspects should be evaluated:

Cost-benefit analysis: Compare cost savings with the required investments for both options.
Feasibility of implementation:
- For claim review, assess the organization’s ability to manage significant workflow changes to avoid delays or cost overruns.
- For fraud detection, verify whether the necessary data is both available and of sufficient quality. Check the regulatory requirements.
Additional evaluation: Consider potential quick wins from other initiatives, while keeping in mind that their smaller savings are unlikely to meet the client’s goals and should be deprioritized unless critical gaps are identified.

Q4 - Math Problem

The health insurance provider processes $1.6 billion in total claims each year. Approximately 5% of these claims are fraudulent, resulting in millions of fraudulent payouts annually.

The team and the client have jointly agreed to implement a new AI-driven fraud detection system and focus on this first (due to staff training issues related to the claim review). This newly developed system, which was not on our radar so far, costs $12 million in upfront implementation expenses (thereby exceeding the initially planned budget) and promises to reduce the fraudulent claims rate from 5% to 4.5%.

Assume that after the system is deployed there are $2 million fixed costs and around $0.3m variable costs per year and the fraud reduction benefits will be fully realized starting in the first year of operation.

How long does it take for the company to recoup its $12 million investment, based on these savings?

Show solution Hide solution

Note for Interviewer

The candidate should first summarize the given information and request one minute to outline their calculation approach. During this time, they may ask clarifying questions if needed.

Once prepared, they should clearly present their approach before starting the calculation.

Assess their calculation approach for efficiency and clarity, their calculation speed and accuracy, and their ability to communicate confidently and calmly, avoiding any signs of stress.

Smart rounding within a 5-10% margin is encouraged (directionally accurate and efficient > exact precision).

The candidate may perform a quick double-check of their numbers before delivering their final answer.

Afterward, they should provide the answer and interpret the results in the context of the case, tying them back to the broader problem. They should then suggest logical next steps to continue driving the case forward.

Throughout all steps, pay attention to their structured communication, logical reasoning, and proactive execution. Offer guidance when necessary and check how coachable they are.

1. Fraudulent payments per year

$1.6bn x 5% = $80 million

2. Annual savings

Quick way: 0.5% x $1.6bn = $8 million
Long way:
- Fraudulent payouts before implementation: 5% x $1.6bn = $80 million
- Fraudulent payouts after implementation: 4.5% x $1.6bn = $72 million
- Difference: $8 million

3. Annual operating costs

Fixed cost + variable cost = $2 million + $0.3 million = $2.3 million

4. Net annual savings

Annual savings – annual operating costs = $8 million – $2.3 million = $5.7 million

5. Payback period

Payback period = Initial investment / net savings = $12 million / $5.7 million ≈ 2.11 years

The company will recoup its $12 million investment in approximately 2.1 years (about 25 months) after implementing the AI system.

The implementation of the AI-driven fraud detection system presents a compelling business case with both financial and operational implications:

Financial justification:
- The system will generate $8 million in fraud reduction savings annually, with net annual savings of $5.7 million after accounting for operational costs.
- The investment will be recouped within approximately 2.1 years, which is a strong return on investment (ROI) timeline for a project of this scale.
Operational justification:
- Beyond direct financial benefits, the system is expected to streamline claims processing, reduce the burden on human reviewers, and enhance detection accuracy, potentially improving customer trust and satisfaction.
- As fraud rates decrease, resources can be redirected to higher-value activities, such as enhancing customer service or developing new insurance products.
Challenges:
- The upfront cost of $12 million exceeds the initial budget, requiring stakeholder alignment and confidence in the projected benefits.
- Potential resistance from employees and customers to adopting AI-driven processes must be addressed through clear communication and training.
- Ongoing fixed and variable costs must be closely monitored to ensure they do not erode the expected savings.
Opportunities:
- If successful, the system could serve as a model for other areas of the business, such as claims forecasting.
- This initiative could also position the company as an industry leader in leveraging AI for operational efficiency, providing a competitive advantage.

Next steps

Consider the implementation: Negotiate with AI system providers and define key project milestones. Develop an implementation roadmap to deploy the system, including timelines, resource allocation, and training plans for relevant teams.
Engage stakeholders: Present the business case to senior management and board members to secure buy-in for the revised budget. Communicate the benefits of the system to employees, especially claims teams, to build trust and reduce resistance.

Q5 - Recommendation

On the night before the final SteerCo meeting, you're in the elevator at 10 PM, finally heading home. As the doors start to close, someone calls out, "Wait for me!" It's Josh, the CTO. He steps in and says, "I’m looking forward to your findings tomorrow. Can you give me a sneak peek of what you’re going to talk about?"

Provide a brief recommendation based on your findings for the 60 second elevator journey.

Show solution Hide solution

Practice This Case With Peers Who Are Currently Looking for Interview Partners.

All Meeting Slots

Do you have questions on this case?

Ask our community and receive answers and tips directly from our experts.

ClaimSmart: AI for Next-Gen Health Insurance i

Case Prompt:

Instructions --> Please read before going through the case

Further Case Information (for follow-up questions)

Q1 - Brainstorming

Q2 - Framework

Q3 - Exhibit Interpretation

Q4 - Math Problem

Q5 - Recommendation

ClaimSmart: AI for Next-Gen Health Insurance