Capgemini Invent Case Interview: 45 Minutes, Five Criteria

The five things Invent says the case assesses

The students and graduates page sets the format first. The case is guided. It runs about 45 minutes. You solve a business problem in real time. The interviewer plays the client and holds the facts you need, which Invent states in so many words.

Invent marks that list "Not exhaustive", so treat it as the published floor.

The fifth one is worth rehearsing, because few firms test it on purpose. Invent asks for risks and ways to reduce them, in the same breath. A risk named with no fix is half the answer. The other half is free.

Its six tips run in order. Take your time. Ask for clarification. Take notes throughout. Sum up the issues and findings. State your recommendations. Then give next steps and the impact you expect. That last one shapes the whole answer: the case ends on impact, after the conclusion.

The prompt Invent publishes, and how to size it

Invent prints a full example. A car maker wants to know what it should do about electric and self-driving vehicles, and wants a view on its future in those markets. Four questions follow. The first is a sizing: what is the market size opportunity in North America.

Everything below is our own chain of assumptions, built for the exercise. A sizing question tests assumptions, so none of these numbers is a fact we looked up. Say each one out loud as you use it.

1. Start from vehicles. Assume 17 million new light vehicles sold in North America a year.
2. Take an electric share. Assume 30% by the target year, which is 5.1 million electric vehicles.
3. Price them. At an average $48,000, that is $244.8bn of vehicle value.
4. Cut to the client's slice. If the client supplies the electric drivetrain at 22% of vehicle value, its opportunity is $53.86bn, which we will carry as roughly $54bn.
5. Size the second business separately. If 8% of those electric vehicles ship with an autonomy system carrying $3,500 of content, that is 408,000 units and $1.4bn.

Step five answers the question Invent asked. Electric and self-driving are one line in the prompt. In the math they are two very different businesses: the self-driving number is about 2.7% of the drivetrain number. Size them as one and you have answered a question the client never asked.

That also gives you the ending Invent's fifth criterion wants. Go where the money is. Size the smaller bet honestly. Then name the risk: the electric share carries the whole answer. Move 30% to 20% and the drivetrain number drops by a third.

Three Invent routes, three different days

Set the published routes next to each other and the differences are structural.

The Swedish postings are the only Invent papers we found that name the whole funnel. CV and grade review. Phone interview. Video interview. On-site case. Final interview. Offer. Two near identical postings differ on one word. One says "On-site case". The Data and AI one says "On-site group case". So the format follows the track, and the country decides nothing here.

Sweden also carries the hardest application rule in this section. You send a CV in English. You also merge your high school, bachelor and master transcripts into one file. Invent states that incomplete applications will not be considered. Identity is checked against your ID card.

The UK route, which has no case

Invent's Accelerate programme publishes a three step process, and a candidate preparing cases for it is preparing the wrong thing.

Step one is the application, reviewed on a first come, first served basis. Applying early is a real advantage here in a way it is not everywhere.

Step two is a Digital Interview, where you answer a series of questions on camera. Invent provides an unassessed practice demonstration beforehand, which is worth using because the recorded format punishes a first take.

Step three is a half day assessment centre. It holds a group exercise and a one to one strengths based interview, plus ice breakers and a short talk about the pathway you applied to. A strengths interview asks what you enjoy and what you do naturally. Rehearsed achievement stories answer a different question.

Two rules bind the UK route. You may apply for one role per recruitment season, which runs September to August. And Invent says that grades count toward the decision without being used to screen people out on their own. The firm also tells candidates they are not in competition and will need to work together on the day. Read that as an instruction for the group exercise.

What the group process page leaves out

Capgemini's group recruitment process page is the page most candidates find first. It is the least useful one here. It says the application takes three to five minutes. It says interviews may be by phone, video or in person. It says you might get a mix of behavioral, technical and cultural fit interviews and assessments. It gives no round count and no duration, names no test, and never mentions a case.

The national pages carry what it leaves out. Capgemini India publishes proctored online technical and language tests. It names an aptitude test covering logical reasoning, verbal ability and quantitative skills. It adds a group discussion for select roles, and one or two technical interviews. It also warns that video interviews may be recorded, with consent asked for first.

The UK graduate FAQ holds the only candidate AI guidance we found at Capgemini, and it is a relaxed one. The firm says it is comfortable with AI for finding roles and drafting. It asks you not to use it for mass generic applications. It states that it uses AI to streamline some steps, and that its selection decisions are always made by its team. One hard rule sits on the same page. Fail at any stage and you wait for the next cycle to apply again.

Where our weights land on an Invent case

Invent lists five criteria and attaches no weights, and marks its own list as not exhaustive. No firm we have read publishes a marking scheme. The weights below are ours, read from the code that runs our grader, each out of a hundred, with the reasoning on how an answer is assessed.

What Capgemini does not publish

Where the case sits in the US process. Invent describes the case in detail and never says which round it is, or how many rounds there are.

Whether the US case exists outside the US. Sweden has one. The UK graduate route does not. Nothing states a rule.

Group sizes, either for the Swedish group case or the UK assessment centre.

The length of the UK digital interview, or how many questions it contains.

The size of the UK assessment centre pool, or what share of it is invited.

Any threshold, on the case, on the Indian tests, or on the digital interview.

A single entry title. Management consultant in Sweden, Associate Consultant in Belgium and Germany, Junior Strategy and Transformation Consultant in Switzerland, Junior Consultant in the Netherlands.

Pay. No Invent page or posting we read publishes a range.

Everything above that is attributed to Capgemini was read on the page or posting it is credited to, with the access route recorded in the ledger. The sizing chain is ours and says so at the top, and the scoring weights are ours, as the section carrying them says.

One firm runs the same forty-five minutes on a different object: West Monroe builds its case on one real project rather than a published criteria list.

What changed on this Capgemini Invent guide

Each entry carries the date of the change.

29 September 2026. A claim that the steps could not be verified. The earlier version said it would not describe Capgemini Invent's screening steps, for lack of any way to check them. Invent US publishes a case page with a length, five criteria and a full sample prompt. Its Swedish postings name six steps in order. That sentence was wrong about the best documented firm in this batch, and it has gone.

29 September 2026. One Invent process. The old page treated Invent as one pipeline, and described its cases in two sentences. Three routes are published. One of them has no case at all. The table above sets them out, with the page each came from.

29 September 2026. The published prompt was never used. Invent gives away a full case prompt, and nobody had worked it, ourselves included. This version sizes it. Every assumption is labeled as ours. The last line says what the math shows about the question Invent asked.